Why FIRC Should Increase Trade

The argument for why FIRC would result in increased economic ties between states, such as heightened bilateral trade, has four parts. First, intervening countries promote their own interests when they use force abroad. Secondly, in some cases the interests they seek to promote are economic. Thirdly, interveners empower leaders in target states who will promote those interests. Finally, imposed leaders are likely to retain close ties to their patrons, resulting in heightened levels of imports, exports and overall bilateral trade. We take the first of these propositions as a given, but flesh out the remaining three parts of the argument in some detail.

States intervene militarily in other countries to protect and advance their interests, including economic ones. The history of US interventions in Latin America is often cited as evidence of such motives. In his overview of regime change operations by the United States in the twentieth century, Stephen Kinzer argues that economic motives underlie almost all such US interventions. In Nicaragua, for example, Kinzer attributes US hostility to Nicaraguan strongman José Santos Zelaya in 1909 to Secretary of State Philander Knox, a successful corporate lawyer with extensive ties to big business. One of Knox’s former clients was La Luz and Los Angeles Mining Company, a Philadelphia-based firm owned by the Fletcher family that operated a gold-mining operation in Nicaragua. Knox was friendly with the Fletchers, who apparently lobbied him to get rid of Zelaya.Footnote ²⁷ After the United States helped topple Zelaya and his successor José Madriz, it imposed the Dawson Agreement on Managua, the terms of which included a multi-million dollar loan from US banks, guaranteed by US control over Nicaraguan customs receipts.

Economic considerations have also been postulated as motives for US regime changes in Latin America during the Cold War. The two cases most frequently mentioned as economically driven US interventions are Guatemala (1954) and Chile (1973). The day before he dictated his famous order to the CIA to ‘save Chile’,Footnote ²⁸ President Richard Nixon was visited by close friend Donald Kendall, chief executive of PepsiCo. Motivated by the fear of losing his company’s Chilean bottling plant to expropriation, as well as the broader threat socialism posed to PepsiCo’s foreign sales, Kendall implored Nixon to block Allende’s inauguration.Footnote ²⁹ Nor was PepsiCo the only concerned US corporation. A board member of International Telephone and Telegraph, worried that Allende would nationalize its Chilean subsidiary, offered $1 million during a meeting with Nixon’s National Security Advisor Henry Kissinger to channel to Allende’s opponents ahead of the presidential run-off in September 1970.Footnote ³⁰ Representatives of Anaconda Copper similarly met with US Ambassador to Chile Edward Korry to lobby against Allende.Footnote ³¹ This is just a sample of the business leaders that urged Nixon to act, which he did in 1970 by supporting an attempted coup and continuing until General Augusto Pinochet succeeded in ousting Allende in 1973. In short, some scholars and primary documents suggest that US regime changes in Latin America were economically motivated.

The next link in the argument is that interveners empower leaders who are friendly to their nations’ interests and who will act to safeguard those interests even at the expense of their own citizens. After a democracy overthrows a foreign leader, Bueno de Mesquita and Downs argue that installing an autocrat maximizes the intervening leader’s chances of re-election.Footnote ³² Democratic interveners have few incentives to empower democratic institutions abroad, because leaders of democracies are most concerned with retaining office by winning the next election. Installing a democracy makes it harder for interveners to control foreign leaders, since those leaders must heed the wishes of their domestic public to retain power rather than the intervener’s wishes. As they do not rely upon public support to stay in power, autocrats are free to implement policies that benefit their foreign patrons with little fear that this will result in a loss of office. This arrangement ensures that the intervention achieves its objectives, which pleases the public in the intervening state and helps keep the intervening leader in power.

The final step in the argument is that imposed leaders ‘deliver the goods’. In Honduras in 1911, US officials concluded that (unlike Miguel Dávila) their preferred puppet Manuel Bonilla ‘was eager to lead Honduras into what would necessarily be a highly unequal partnership with the United States’.Footnote ³³ After Bonilla took office, he thanked his patron Sam Zemurray by giving him nearly 50,000 acres of land.Footnote ³⁴ After the CIA ousted Mohammad Mossadeq in Iran and replaced him with Mohammad Reza Shah in 1953, US oil companies received 40 per cent of the shares in the New National Iranian Oil Company, a market previously dominated by the British.Footnote ³⁵

Few scholars have examined the broader effects of intervention in a systematic way. One exception is Berger et al.,Footnote ³⁶ which investigates the effect of CIA interventions during the Cold War on imports from the United States. The study finds that imports from the United States are roughly 28 per cent greater (as a share of country GDP) in country-years coded as having a CIA intervention than those without an intervention. Berger et al. argue that the United States has greater influence in countries where the CIA has helped to install a new government or is engaged in propping up an existing government. This increased level of influence causes the supported government to shift imports away from suppliers in other countries and toward firms based in the United States.Footnote ³⁷ One reason for this shift is that the new leader owes his position to the intervener, and may have agreed to provide economic concessions as a condition for the foreign country supporting his bid for power. Imposed leaders may also lack a solid base of support inside the country. Such leaders frequently depend heavily on their foreign patrons for protection against their domestic opponents. Finally, the newly installed leader may seek to restock his military supplies to help secure himself, which may create a demand for imports that is likely to be fulfilled by firms from the intervening state.

In short, because imposed leaders rely on a foreign power for their political survival, imports from the intervener should increase. This should cause an increase in total bilateral trade. Whether FIRC should also increase exports is an interesting question. Following FIRC, it is reasonable to expect that the imposer would become the preferred destination for products from a country, as the target country’s new leader is likely to offer preferential trade terms. Moreover, officials in the imposing country may choose to accept more imports from a country after FIRC to generate support for their new protege among business interests in his country. Drawing upon this analysis, we posit the following hypothesis about trade between imposers and targeted states:

The effect of FIRC on trade between the imposing country and the target might depend on the type of government that is installed. Previous studies of regime type and trade, for example, find that democracies trade more among themselves than they do with non-democracies, or than non-democracies do among themselves.Footnote ³⁸ These studies suggest that only those US regime changes that result in the creation of democratic regimes should increase bilateral trade. FIRCs that bring non-democratic regimes to power, by contrast, should decrease bilateral trade. An alternative logic, however, suggests the opposite: leaders in non-democracies are less encumbered by public opinion than in democracies, and are freer to provide benefits to external patrons. Imposed autocrats may also remain dependent on the imposing country for internal and external security, and thus feel compelled to make economic concessions in return for security benefits.Footnote ⁴¹ In democracies, voters are likely to oppose economic favors that benefit foreigners and have the ability to punish leaders who fail to heed their views. These opposing logics generate the following conditional hypotheses:

Why FIRC Should Have No Impact on Trade

The assertion that military interventions are motivated by economic concerns is quite controversial in the historical literature. This debate is especially intense concerning the cases that are the subject of this article: US interventions in Latin America. This section briefly lays out the case that economic motives do not feature prominently when US policy makers intervene in the region.

Most scholars who make the economic argument rely on two strategies for demonstrating that economics shape intervention. First, they document the business backgrounds of key government officials, the ties between officials and firms doing business in the target of intervention, and attempts by officers of those firms to lobby policy makers.Footnote ⁴² Secondly, these scholars point out that certain companies benefited from intervention and infer that those benefits must have been the reason for intervention in the first place.Footnote ⁴¹ Yet proponents of economic interpretations have rarely, if ever, been able to demonstrate definitively that the desire to expand markets or protect the profits of US corporations drove policy makers to act. As one historian comments about the Guatemalan case, ‘no hard evidence has yet come to light that any U.S. officials made policy on the basis of UFCO’s [United Fruit Company] interests. It appears, in fact, that rather than being the tool of UFCO, Eisenhower sought to “use” the company to contain communism in Central America’.Footnote ⁴²

Underlying most US interventions in Latin America is the long-held principle – originally articulated in the Monroe Doctrine – that the United States will not tolerate foreign influence in its hemisphere. Early US interventions, for example, were driven by Washington’s fear of Europeans gaining control over the territory or finances of small states in Central America and the Caribbean. For example, the key factor that made US officials sympathetic to regime change in Nicaragua and Honduras in 1909, 1910 and 1911 was not the plight of US businesses but rather the fact that Zelaya and Dávila had contracted sizable loans from European banks. Even as ardent a proponent of economic arguments as Stephen Kinzer concedes that US Secretary of State Philander Knox ‘understood perfectly well that by borrowing money from European rather than American banks, Zelaya was trying to make his country less dependent on the United States. This he could not abide’.Footnote ⁴³ Knox and President Taft similarly ‘disapproved’ of Honduras’s practice of taking European loans.Footnote ⁴⁴

Economic motives for intervention in Chile are also far from clear-cut. In his memoirs, Secretary of State Henry Kissinger subsequently denied that economics figured in US decisions on Chile. While somewhat self-serving, this statement is corroborated by contemporaries who characterized Kissinger’s views toward US multinational corporations as ‘contemptuous’. One such observer said of Kissinger, ‘he never gave a shit about the business community’.Footnote ⁴⁵ Declassified documents support the view that Nixon and Kissinger interpreted Allende’s Chile as a challenge to US credibility. In a meeting of the National Security Council on 6 November 1970, President Nixon remarked, ‘If Chile moves as we expect and is able to get away with it it gives courage to others who are sitting on the fence in Latin America’, such as Brazil and Argentina. The president continued: ‘If we let the potential leaders in South America think they can move like Chile and have it both ways, we will be in trouble. Latin America is not gone, and we want to keep it’.Footnote ⁴⁶ A review of US policy in the event Allende became president written in August 1970 similarly noted that an ‘Allende victory would inevitably be seen around the world and within the United States as a definite set-back to U.S. interests and aspirations and would be exploited as such by our adversaries’.Footnote ⁴⁷ According to Grow, ‘The administration feared that a perception of U.S. weakness in Chile would damage its credibility in the eyes of communist bloc governments’.Footnote ⁴⁸ This evidence that US interventions are driven by a variety of non-economic goals leads us to posit the following hypothesis:

Why FIRC Should Decrease Trade

A third argument decouples the effects of interventions from their causes. According to this view, regardless of why state leaders intervene to overthrow foreign governments – to promote the economic welfare of their national firms or to eliminate threats (or potential threats) to their national security – FIRC has unforeseen side effects that destabilize the target country and foster violence and instability. Increased political instability following regime change has two effects. First, it causes foreign businesses to depart and deters others from coming in, which drives down imports and foreign investment. For example, the OECD reports that American firms did not invest at all in the Haitian economy for eight years after its 1994 FIRC.Footnote ⁴⁹ Similarly, following the advance of fighters from the Islamic State of Iraq and Syria (ISIS) into Iraq in 2014 and their capture of the Baiji oil refinery – part of the continuing fallout from the 2003 US FIRC in that country – US oil companies began to withdraw personnel from the country.Footnote ⁵⁰ Secondly, instability decreases exports by destroying property and infrastructure, hindering travel and spreading fear among the population. This argument thus anticipates that trade between the imposing country and the target will decrease in the aftermath of FIRC.

The argument is based on the assumption that international trade is driven by the principle of comparative advantage; for serious and lasting increases in trade between nations to occur, external shocks like FIRC must enhance the complementarity of their respective economies or of various firms within those economies. Although this may sound basic, it is worth noting that the economic argument spelled out above does not follow this logic. Rather, it simply asserts that political leaders can directly influence their domestic firms’ behavior, ordering or coercing them to trade with firms from one country over another. While this conjecture might be plausible when significant portions of the economy are state owned, this is not the case in either the United States or Latin American countries targeted for FIRC. Rather than promote nationalization, the United States accepted only pro-free trade and capitalist economic systems and actively dissuaded countries from engaging in economic reform. It is thus not clear that the economic argument has the right model of leader influence. It seems at least as plausible that leaders exercise their influence indirectly in ways that affect comparative advantage, such as by subsidizing the production of some commodity for export, thus making it cheaper and more attractive in foreign countries. Leaders could also help their firms by investing in infrastructure upgrades and educating the workforce so that manufacturers can get goods to port more quickly.Footnote ⁵¹ But the economic approach does not make these sorts of arguments.

We argue, therefore, that the correct locus of a theory about the economic consequences of intervention ought to be the firm. As profit-maximizing agents that depend upon shareholders’ support, firms are unlikely to trade when they do not expect to earn a profit.Footnote ⁵² Domestic firms are likely to expect a profitable return on trade when they have a comparative advantage over foreign rivals. In those conditions, firms in countries targeted for FIRC will import cheaper goods rather than manufacture them more expensively at home. Similarly, domestic firms will only increase their imports from manufacturers in the country targeted for FIRC if regime change improves their comparative advantage. The key question is thus how FIRC affects the comparative advantage of firms in the imposing and target countries. We argue that FIRC harms the competitiveness of firms in both places, and that this effect overwhelms any positive inducements to trade that newly empowered leaders might implement. FIRC thus reduces imports, exports and overall trade.

An emerging literature has shown that FIRC is often followed by political instability, violence, coups and even civil war. Peic and Reiter and Downes, for example, document a significant increase in the probability of civil war in the years immediately following FIRC.Footnote ⁵³ Other studies have found that foreign-imposed leaders face an elevated risk of violent removal from office compared to leaders that come to power in elections.Footnote ⁵⁴ Both of these phenomena are on display in Latin America, as highlighted by the American intervention in Nicaragua. As noted above, after US Marines helped oust Liberal President José Madriz, the United States imposed the Dawson Agreement, which forced Nicaragua to accept a large loan from US banks, guaranteed by US control over Nicaraguan customs revenues. This deal enraged Liberals, who demonstrated against it and began plotting a coup, and also angered key Conservatives (the party the United States installed). One such Conservative, Minister of War Luis Mena, joined with Liberals in launching a rebellion in protest. The ensuing civil war killed between 2,000 and 5,000 Nicaraguans and prompted another US intervention to rescue its imposed government.Footnote ⁵⁵ US forces occupied Nicaragua nearly continuously until 1933, and had to help quell a second civil war that started after another US FIRC in 1926.

More recent cases similarly exemplify this dynamic. After the US regime change in Guatemala in 1954 curtailed that country’s 10-year experiment with democracy, Guatemala entered a tailspin of political violence, military rule, civil war and genocide. Rather than transform into a capitalist hub of manufacturing, Guatemala gained a reputation for brutality and internal instability that was consolidated during the resulting thirty-six year long civil war. Even though the war formally ended in 1996, Guatemala continues to be plagued by violence associated with drugs and gangs; indeed, the country has one of the highest violent crime rates in Latin America.Footnote ⁵⁶ As a result of this instability, Guatemala’s per capita income is roughly half that of the average Latin American or Caribbean country; 54 per cent of the population lives below the poverty line, and nearly one-half of children under five are chronically malnourished.Footnote ⁵⁷ According to one source, ‘concerns over security, the lack of skilled workers, and poor infrastructure continue to hamper foreign direct investment’.Footnote ⁵⁸

As mentioned above, this post-FIRC turmoil decreases trade flows through two mechanisms. First, concerns about domestic stability and security make the targeted country a less attractive destination for foreign investment. If foreign firms come to believe that their investments could be destroyed by riots and civil war, seized by nationalist campaigns, or their workers killed and maimed, they are unlikely to invest. Following the 1916 US FIRC in the Dominican Republic, for example, a low-level insurgency developed in the eastern, sugar cane-growing region of the country. Insurgent raids on sugar plantations frightened the workers so much that the production of sugar on one estate fell by 75 per cent in 1921, and the kidnapping of foreign personnel by insurgents caused some workers to flee the country in fear.Footnote ⁵⁹ Foreign firms already present in the country could curtail their operations, or even sell their investments and leave if they feel threatened by any of these phenomena. This should decrease imports from abroad.

Additionally, post-FIRC violence and instability may suppress consumer demand in the target country. It is well known, for example, that low income increases the risk of civil war, but it is also true that once conflict breaks out, it further reduces income. One study puts the annual loss in GDP from civil war at 2.2 per cent.Footnote ⁶⁰ Taking into account the average length of civil wars (seven years), the amount of time it takes the economy to return to pre-war rates of growth (ten years from war termination), and the cost of increased military spending during and after the conflict, the typical civil conflict costs 123 per cent of the country’s GDP.Footnote ⁶¹ Although FIRC does not always lead to full-blown civil wars, it sometimes does, and even when it does not, other forms of violence – such as coups, assassinations and government repression – are relatively common. The reduction in national income from this instability is bound to translate into reduced consumption of all goods, including imports.

Secondly, this turmoil also makes domestic firms less internationally competitive. US FIRCs in Latin America often maintained the power of conservative, upper-class elites at the expense of the working class. The resulting oligopolies likely reduced entrepreneurship and blocked the market entry of more productive firms, which resulted in delayed economic development.Footnote ⁶² Low investment in public goods resulted in a less-educated and less-healthy workforce, which also slowed economic growth. Growth aside, it is dubious whether firms can maintain historical productivity levels when workers cannot leave their homes due to security concerns, producers cannot obtain replacement parts and leaders impose regressive economic policies. Images of turmoil and violence, moreover, likely decrease the perceived quality of domestically manufactured goods abroad. In sum, FIRC should also decrease the targeted country’s exports.

The dynamics of this argument hold even absent a full civil war. FIRCs give rise to a panoply of violence as new rulers try to consolidate their power, ousted elites struggle to get it back, the military intervenes in politics and groups long out of power take revenge on their former oppressors. All of these forms of political instability generate fear that keeps consumers at home, failure to invest in infrastructure and human capital that undercuts the ability to produce for export, and perceptions among the foreign business community that the country is a poor investment risk. Our argument leads us to posit our final hypothesis:

Case Selection

FIRC has removed 111 leaders from power since 1816, thirty-one of them (in twenty-eight separate episodes) by the United States.Footnote ⁶³ Rather than study the entire universe of cases, we focus on US-led FIRCs in Latin America for methodological as well as substantive reasons. Methodologically, US FIRCs in Latin America are widely considered archetypal cases of economic intervention, which suggests that this group of cases is a ‘most likely’ setting for economic arguments.Footnote ⁶⁴ If the economic argument is not supported here, in other words, it is unlikely to be supported elsewhere. Secondly, by restricting our analysis to Latin America, we benefit from the stable borders in the region. Stable borders are important for satisfying the ordinary least squares’ (OLS) unit homogeneity assumption. Violations of this assumption are both very likely and problematic in historical studies of trade, when acquiring or losing territory can change a country’s mass in the gravity model.Footnote ⁶⁵ To concretize this, the Allied FIRC of Nazi Germany was coincident with significant border changes along the frontier with Poland and France as well as the undoing of the Anschluss with Austria. As a result of these border changes, US trade decreased significantly with Germany and increased with Austria, France and Poland solely due to border changes.Footnote ⁶⁶ Border changes, however, are uniquely uncommon in Latin America. The last major change occurred when Paraguay acquired two-thirds of the disputed Chaco territory from Bolivia in the Chaco War (1932–35). This territorial stability makes the unit homogeneity assumption of OLS justifiable across more than a century of data.

When regime changes coincide with border changes, it is difficult to identify the independent effect of FIRC. There are two ways to address this inferential problem: focusing solely on the post-World War II era or gathering data from regions where borders are relatively stable.Footnote ⁶⁷ The former is the empirical strategy adopted by Berger et al.,Footnote ⁶⁸ who draw evidence only from the Cold War period. While this might seem reasonable at first, there are two serious issues with this strategy. First, it assumes that the treatment effect of FIRC during the Cold War generalizes to FIRCs that occurred before and after it. For reasons discussed below, this is unlikely to be true, and uncovering the effect of FIRC on trade requires studying the entirety of a relevant sample. Secondly, without a clear theoretical justification, Berger et al.’sFootnote ⁶⁹ strategy of dropping cases from before 1945 risks introducing significant bias when estimating treatment effects. The high salience of military alliances and regime type distorted trade flows during the Cold War.Footnote ⁷⁰ Generalizing from this period to either before or after the Cold War requires scholars to assume that firms’ and states’ preferences for trading with ideological allies do not vary. Given the increase in trade between the West and autocracies since the 1990s, this assumption is likely untenable. As such, it is likely that Cold War interventions systematically vary from earlier interventions because of the heightened salience of rivalry between economic systems. As we believe that focusing solely on the Cold War introduces significant bias and makes it difficult to test an important question, we concentrate on Latin America because it experienced interventions before, during and after the Cold War.

Beyond the availability of relevant cases, we argue that Latin America is a substantively important region in which to study the impact of FIRC on trade. Rather than fade into obscurity, US-led interventions in Latin America remain highly politically salient and continue to influence US foreign policy in the region. Across the political spectrum, Latin American leaders and politicians cite American imperialism as a rationale for avoiding deeper regional integration, free trade agreements and even adopting ‘American’ social policies. In a notable example, Hugo Chávez of Venezuela used his 2006 speech before the United Nations General Assembly to accuse the United States of threatening Latin American sovereignty for over a century. In yet another example, when Bolivian President Evo Morales’ plane was forced to land in Austria because of suspicions that Edward Snowden was on board, several Latin American leaders included references to US interventions in their comments. President Cristina Fernandez de Kirchner of Argentina, for instance, described it as an act of ‘[American] colonialism that we thought was completely overcome’.Footnote ⁷¹ Given the harm that Latin American FIRCs have caused to interstate relations in the region, it is substantively interesting to consider whether US businesses – the very group cited as commercial imperialists – actually benefited from the interventions.

Focusing on Latin America imposes two important scope conditions on our analysis. First, Latin America does not represent the universe of FIRCs; many other interveners have imposed governments in different geographic regions. Restricting our analysis to US-imposed FIRCs thus prevents us from exploring whether factors about different interveners – such as regime type – affects trade with the targeted state. Secondly, limiting our focus to US interventions in Latin America obviously means that we cannot test whether US FIRCs in this region have different effects on trade than US FIRCs in other areas of the world. Analyzing only Latin America, however, captures the vast majority of US FIRCs not associated with World War II, which were idiosyncratic in that they either involved massive post-FIRC reconstruction and lengthy occupation, or reinstated to power previous democratic governments ousted by the Germans. Our analysis omits only four US FIRCs that did not occur during World War II: Iran (1953), South Vietnam (1963), Afghanistan (2001) and Iraq (2003).

Dependent Variable

We identify three means of measuring post-intervention trade flows: total trade, US market share and sectoral trade flows. While the latter two measures provide a finer-grained view of commerce, the historical nature of this article poses significant challenges. Measuring market share requires firm-level data about all of a country’s trading partners. Despite an exhaustive search, much of these data are permanently unavailable for two reasons: natural disasters destroyed the country’s archives,Footnote ⁷² and multiple countries had not yet begun collecting trade data prior to experiencing intervention. For instance, the Dominican Republic published its first Annual Statistical Report in 1936 – twenty years after undergoing FIRC. As the predictions of our argument are similar for total trade and market share, we use trade flows as our primary dependent variable. To consider whether US firms outperform their foreign rivals, we conduct a case study below using archival data from Panama.

Sectoral trade flows provide leverage on a variant of the economic intervention theories. Scholars commonly argue that US foreign policy is deeply influenced by particular interest groups that push leaders in Washington to support their interests.Footnote ⁷³ US interventions in the ‘banana republics’ of Central America – as amply demonstrated above – are held up as iconic examples of such interventions. To address whether particular sectors benefited from these interventions, we test the impact of intervention on US banana imports using archival data gathered from port records held at the library of the University of Texas, Austin.

The most commonly cited dataset that includes pre-World War II trade flows is the Correlates of War Bilateral Trade Data Set (COWBT). Unfortunately, it is inappropriate for our analysis for two reasons. First, and most importantly, it excludes many country-years when the state was under foreign occupation.Footnote ⁷⁴ Data from these years are essential for determining the impact of FIRC. Secondly, COWBT relies on inaccurate secondary sources, such as consular reports and almanacs, for its historical trade data. As an example of the errors contained in these documents, COWBT states that the United States imported $1.75 million worth of goods from Haiti in 1886. The value in the official statistical record of the United States, $2.6 million, is nearly 50 per cent larger.Footnote ⁷⁵

To overcome this limitation, we conducted archival research at the Department of Commerce library in Washington, DC, to produce a novel dataset on bilateral trade between the United States and Latin America from 1873 to 2007. We report figures from the Statistical Abstracts of the United States, which contain abundant data about all aspects of commercial activities in the United States. Notably, the abstracts also have values for all missing country-years in the COWBT. Not every country enters the dataset in 1873, as early versions of the abstract grouped smaller countries into regional units. Each country, therefore, enters the dataset either in the year it became politically independent or the year the abstract reports non-conglomerated figures for it. We drop colonies from our analysis until their formal independence because the metropolitan government could replace leaders that angered it without a FIRC. Once in the dataset, we convert all values to 1996 US dollars to match GDP data from BoixFootnote ⁷⁶ using the Federal Reserve Bank consumer price index (CPI).Footnote ⁷⁷

Independent Variable

Our data on FIRC are taken from Downes and Monten, who define it as ‘the forcible or coerced removal of the effective leader of one state — which remains formally sovereign afterward — by the government of another state’.Footnote ⁷⁸ In a handful of cases, external actors use covert means to overthrow the targeted regime. In such cases, the foreign government must make removing the target regime its official (if not public) objective, and external actors or aid play a decisive role in toppling the leader. The removal of Guatemala’s Jacobo Árbenz in 1954 is an example. An additional requirement to qualify as a case of FIRC is that target states must be independent before intervention and remain (at least nominally) so afterwards. This requirement disqualifies governments imposed during the process of decolonization as well as instances of territorial annexation. FIRCs followed by military occupations are included so long as the occupier does not annex the target. In their dataset of all FIRCs from 1816 to 2008, they report fifteen US-implemented regime changes in Latin America, all of which fall within our time period.Footnote ⁷⁹

This coding rule differs from other datasets that identify cases of the external removal of leaders, such as Archigos and Polity III. Archigos, for instance, ‘includes only cases where a foreign state directly removes a leader, for example through invasion or kidnapping’.Footnote ⁸⁰ The Polity III data also include a variable for ‘externally caused regime change’, coded as occurring ‘when a polity is “terminated in circumstances of international war, threat, or intervention [and] if the nation or its component parts maintain their autonomy”’.Footnote ⁸¹ These collections, however, exclude most cases of US FIRCs in Latin America, few of which were accomplished with outright invasions.Footnote ⁸² Most US FIRCs in the region resulted from compellent threats or ultimata backed by the deployment of forces, as when President Woodrow Wilson demanded that Costa Rican President Federico Tinoco resign in 1919 and sent US warships to back up his threat. Other cases resulted from US support for coups or rebels, as in Chile and Guatemala. We agree with Downes and Monten that the attitude and intervention of the United States is crucial to explaining why these leaders left power, and thus employ their data.Footnote ⁸³

Gravity Model

First, following previous research on the determinants of bilateral trade, we run a linear gravity model using country fixed effects and cubic restricted time splines. The gravity model of bilateral trade is commonly used in economics and political science to generate predicted values of bilateral trade within a dyad. We compare outcomes for countries that undergo intervention to those that did not. The basic covariates for a bilateral gravity model are the distance between two states’ capital cities and the size of their respective gross domestic products (GDPs).Footnote ⁸⁴ Other scholars further specify these basic assumptions by including additional independent variables to account for variation in the relationship of the trading dyad, such as joint democracy and trading organization membership.Footnote ⁸⁵

We estimate the predicted values of total bilateral trade with Equation 1:

(1) $$\eqalignno{ & log\,Trade_{{ijt}} =\beta _{1} {\plus}\beta _{2} \,log\,GDP_{{USA_{t} }} {\plus}\beta _{3} \,log\,GDP_{{jt}} {\plus}\beta _{4} \,FIRC_{{jt}} \cr & {\plus}\beta _{5} \,Democracy_{{jt}} \,{\plus}\beta _{6} \,TradeOrg_{{ij}} {\plus}\beta _{7} \,CivWar_{{jt}} {\plus}\mathop{\sum}\limits_{j=1}^n {\alpha _{j} {\plus}f(\gamma j){\plus}\epsilon } . $$

While it is standard practice for gravity models in the literature to pool data into common intercept linear regressions, this could bias results.Footnote ⁸⁶ In this instance, each country likely has a unique baseline of trade that depends on its geography, urbanization, industrialization or other factors. To address this concern, we include country fixed effects, defined as α _j , which is a set of unobserved fixed parameters for each of the N units.Footnote ⁸⁷ To account for the possibility of temporal dependence in our dependent variables, we follow Beck, Katz and TuckerFootnote ⁸⁸ and include restricted cubic time splines with knots at each decade. A spline function is a ‘smoothly joined piecewise polynomial of degree n’.Footnote ⁸⁹ Splines are included in our model to control for nonlinear time effects, such as wars or new technology, which affect all countries in the panel differently. Following Dupont and Plummer,Footnote ⁹⁰ restricted time splines are defined as:

(2) $$\eqalignno{ \gamma _{1} = \,& \gamma \cr \gamma _{j} =\, & (\gamma {\minus}t_{{j{\minus}1}} )_{{\plus}}^{{\rm 3}} {\minus}\left( {{{(\gamma {\minus}t_{{k{\minus}1}} )_{{\rm {\plus}}}^{{\rm 3}} (t_{k} {\minus}t_{{j{\minus}1}} )} \over {{\rm (}t_{k} {\minus}t_{{k{\minus}{\rm 1}}} {\rm )}}}{\plus}{{(\gamma {\minus}t_{k} )_{{\rm {\plus}}}^{{\rm 3}} (t_{{k{\minus}1}} {\minus}\mathop{{t_{{j{\minus}1}} }}\limits)} \over {{\rm (}t_{k} {\minus}t_{{k{\minus}{\rm 1}}} {\rm )}}}} \right)^{{\gamma _{{\rm 1}} =\gamma }} \cr & \quad \quad \quad \quad \quad \quad \quad \quad \quad \quad \quad \quad \quad \quad {\rm for}\,j={\rm 2},\,\ldots\,,k{\minus}{\rm 1} \cr & \quad \quad \quad \quad \quad \quad \quad \quad \quad \quad \quad \quad \quad \quad {\rm (}u{\rm )}_{{\plus}} =\left\{ {\matrix{ {u:} & {u\gt 0} \cr 0 & {u\leq 0} \cr } } \right., $$

where t _i , i=1,…,n are the spline knot values, k is the number of knots and each decade from 1890 to 2000 is assigned a unique γ _j.Footnote ⁹¹

The dependent variable in Equation 1, log Trade _ijt , is the natural logarithm of the value of total bilateral trade between country i and country j in year t.Footnote ⁹² Since our dataset covers only trade between the United States and Latin America, country i is always the United States while country j is the country with which it trades.

The independent variables in the model include standard gravity model covariates such as GDP as well as membership in a trading organization, joint democracy, ongoing civil war and FIRC. First, we include the natural logarithm of US GDP (logGDP _USAt ) and country j (logGDP _jt ) in year t given in 1996 USD. These figures were generated by multiplying the GDP per capita figures from BoixFootnote ⁹³ by annual population figures from Maddison.Footnote ⁹⁴ As Boix’s dataset does not extend past 2000, we perform multiple imputations with Amelia in order to avoid listwise deletion of later years.Footnote ⁹⁵ Secondly, Goldstein, Rivers and TomzFootnote ⁹⁶ have shown that common membership in trade-promoting organizations such as GATT or the WTO can increase bilateral trade. After 1945, Tradeorg takes a value of 1 when both the United States and country j belong to the WTO or GATT and 0 when country j did not belong to either institution. Thirdly, we include a dummy variable Democracy denoting whether country j was a democracy to capture the effects of joint democracy on trade.Footnote ⁹⁷ To capture whether internal conflict is driving these results, CivWar takes a value of 1 when either the Correlates of War or PRIO report more than 1,000 battle deaths in year t.Footnote ⁹⁸ Summary statistics for all variables included in the analysis are displayed in Table 1.

Table 1 Summary Statistics for Key Variables

Our independent variable of interest, FIRC, is a dichotomous variable that takes a value of 1 to indicate that country j experienced a FIRC by the United States. To account for variation in the resulting regime type and how beholden their leaders are to American interests, we include indicator variables for whether the resulting regime was a democracy or autocracy. FIRCNonDem takes a value of 1 when a country underwent a regime change and was subsequently ruled by a non-democratic regime. FIRCDem takes a value of 1 when the resulting regime is democratic.Footnote ⁹⁹ As shown in Table 2, ten states in Latin America and the Caribbean experienced a total of fifteen FIRCs. In two cases,Footnote ¹⁰⁰ the United States removed multiple leaders in quick succession. These episodes are coded as single FIRCs in our data.

Table 2 Foreign-Imposed Regime Changes by the United States in Latin America

A final issue is how to deal with Cuba. When Fidel Castro came to power in 1959, Cuba was the United States’ third-largest trading partner in the region after Venezuela and Mexico. In response to the Cuban Missile Crisis of 1962, President John F. Kennedy slapped an embargo on trade, dropping trade flows to 0. Because US trade with Cuba was artificially constrained for much of its history, as well as the status of this case as an extreme outlier, we exclude Cuba from the analysis.Footnote ¹⁰¹

Synthetic Control Analysis

In addition to the traditional gravity model, we use synthetic controls to estimate the effect of FIRC on bilateral trade. The fundamental problem of causal inference is that analysts can never observe the same unit both receiving the treatment and not receiving the treatment, and thus cannot know what would have happened had that unit not received the treatment (and vice versa). In this application, we cannot observe the level of trade a country would have had with the United States if FIRC had not occurred. To partially sidestep this problem we use synthetic controls. Synthetic control analysis is a statistical technique that uses a weighted average of nearby countries that did not receive the treatment in question (in this case, FIRC) to create the previously unobserved counterfactuals. As a result, we are able to create equal pre-treatment time trends and satisfy the difference-in-difference identifying assumption. This allows us to compare the evolution of US trade with the real countries to the counterfactual that did not undergo intervention.

One benefit of this technique is that we can ignore economic and time trends since, on average, these factors will impact all countries. As Figure 1 shows, trade between Latin America the United States has increased since 1873. Synthetic control analysis is especially useful in the situation we observe here with an upward-trending variable of interest. In a regression analysis, it is difficult to distinguish an increase in trade owing to the passage of time from an increase in trade caused by FIRC because the period after FIRC should, on average, already have higher trade owing to time trends. For instance, trade may increase after FIRC, but it may not increase as fast as it would have absent FIRC. The synthetic controls analysis allows us to determine whether a country enjoys trade that is higher or lower than it would have been without FIRC.

Fig. 1 Trends in trade between the United States and Latin America, 1873–2007

Synthetic control analysis requires data from countries that experience FIRC and those that do not, over a time period that extends from before the first to after the last FIRC. Only Latin American countries that do not experience FIRC in our time frame are used to create counterfactuals, which are supposed to represent outcomes a country would have experienced had it never experienced FIRC.

For each outcome, synthetic controls are created using the following equation:

(3) $$Interve\widehat{{ned\,Out}}come_{t} =\mathop{\sum}\limits_{{\rm Countries}\,{\rm in}\,{\rm Region}} {\beta (Non{\minus}Intervened\,Outcome)_{t} {\plus}{\varepsilon}_{t} .} $$

This formula shows that in each year prior to the FIRC, the trade between a country that experiences FIRC and the United States is recast as a weighted multiple of the trade that other countries in Latin America that never experienced FIRC had with the United States. For example, Guatemala experienced an intervention in 1954. In order to create a counterfactual Guatemala, the level of total bilateral trade in Guatemala was regressed against the total bilateral trade in Brazil, Colombia, Venezuela, Ecuador and Bolivia.Footnote ¹⁰² We performed the same calculation for imports and exports. For each country that experienced FIRC, we created a counterfactual country that was statistically identical to that country before FIRC occurred. We then test whether the average trade flows in the twenty years following a FIRC are statistically distinct.Footnote ¹⁰³

Market Share

Our quantitative analysis shows that FIRC causes total trade, imports, and exports to decrease between the United States and countries that experienced a US FIRC in Latin America. Yet a FIRC that resulted in lower trade might still be considered a ‘success’ if it increased the market share held by US firms. Estimating the effect of FIRC on market share requires data from the targeted state about all of its trading partners. These data are generally unavailable, but fortunately the National Archives of Panama located that country’s trade data before and after the US-led FIRC in 1990. We use these data to test the robustness of our findings by considering an alternative dependent variable: US firm market share.

Panama is a particularly good case with which to test the economic effects of FIRC. For the United States and American firms, the incentives to have a pliable, pro-business leader in charge of Panama are clear. Domestic turmoil might disrupt shipping through the world’s busiest seaway, the Panama Canal. The Panamanian government is also able to increase usage fees, hold cargo ransom or prevent ships from particular countries from using the Canal, which could radically affect the bottom lines of US businesses. This fear influenced how Presidents Ronald Reagan and George H. W. Bush dealt with the country. Owing to his increasingly open support for the communist governments in Nicaragua and Cuba, President Manuel Noriega of Panama worried American political and business leaders.Footnote ¹⁰⁹ Large street protests calling for Noriega to step down broke out in 1987.Footnote ¹¹⁰ Despite repeated attempts to encourage Noriega to leave of his own accord, he refused and declared war on the United States in December 1989. The United States invaded a week later, and deposed him in January 1990. When asked why he intervened in Panama, George H. W. Bush echoed the business community and justified his actions as a means of ‘protect[ing] the integrity of the Panama Canal Treaty’.Footnote ¹¹¹

Table 6 presents data from archival trade reports from the Panamanian government. As the United States imposed a new government on Panama in 1990, we use 1989 as our pre-FIRC year and compare it to 1992, which is the last year data are available.

Table 6 Panamanian Trade, 1989 and 1992

Note: dollar amounts reported in thousands. Source: Comisión de Publicaciones (1994).

The United States’ market share declined following the FIRC. This decrease, moreover, is statistically significant.Footnote ¹¹² It also cannot be explained by economic contraction; total trade increased between 1989 and 1992. As these are data from only one country, this result is not conclusive proof that FIRC causes a decrease in market share for firms from imposing countries.Footnote ¹¹³ It does suggest, however, that using total trade as a dependent variable does not drive the results presented in the previous section.

Sectoral Trade

It is also possible that the United States intervenes abroad to support particular sectors. This claim is prevalent in the literature on the United States and its role in Latin America. In particular, the ‘banana republics’ of Central America are held up as prime examples of Washington’s responsiveness to business interests. In this view, the banana industry lobbied Washington to install friendly leaders who would protect banana plantations, decrease labor costs, and ultimately increase production and imports. Hitherto, the data necessary to test this argument were unavailable. We use archival records on imports from US ports held at the library of the University of Texas, Austin to assemble a dataset of US imports of bananas by country from 1905 to 1946. Were the sectoral argument correct, we would expect interventions to be driven by pre-intervention trends in import data and to increase imports relative to untreated country-years. To test whether this is the case, we estimate the following equation:

(4) $$\log BANANA_{{ijt}} =\beta _{0} {\plus}\beta _{{\rm 1}} PREFIRC_{{jt}} {\plus}\beta _{{\rm 2}} FIRC{\plus}\mathop{\sum}\limits_{j=1}^n {\alpha _{j} } {\plus}\mathop{\sum}\limits_{j=1}^n {\lambda _{t} {\plus}\epsilon _{{ijt}} ,} $$

where logBANANA _ijt is the value of imports of bananas from country j to the United States in year t in inflation-adjusted 1996 USD.Footnote ¹¹⁴ PREFIRC takes a value of 1 for five years prior to a FIRC and FIRC takes a value of 1 for either five, ten or twenty years following intervention. $\mathop{\sum}\nolimits_{j={\rm 1}}^n {\alpha _{j} } $ represents the sum of country fixed effects for each country j in the sample and $\mathop{\sum}\nolimits_{t=1}^n {\lambda _{t} } $ is the sum of year fixed effects for each year t in the sample. Year fixed effects account for yearly events that might influence banana production and imports, such as drought and hurricanes. While banana-producing states are geospatially clustered around the Caribbean basin, it is still possible that adverse climate events do not affect all countries in our sample equally. Regrettably, we are unable to locate consistent climate data from the appropriate time period. To address this issue, one suggestion from Carneiro et al. might be to include country-year fixed effects.Footnote ¹¹⁵ As our data have only one observation per country-year, we have insufficient degrees of freedom to incorporate this suggestion. With that caveat in mind, Table 7 presents the results from Equation 4.

Table 7 Effect of Foreign-Imposed Regime Change on US Banana Imports from Latin America, 1905–1946

Note: robust standard errors (clustered on country) in parentheses. ***p<0.01, **p<0.05, *p<0.1.Source: Bureau of the Census (1905–1946).

The results shown in Table 7 contradict the sectoral argument about FIRC. The control variable for pre-intervention trends is not significant at any level. Our results indicate that there is no statistical difference between country-years before intervention and other country-years, which suggests that there were no observable trends to cause American firms to lobby for intervention abroad. Secondly, FIRC has no effect on banana imports. These results suggest that specific sectors – in this case, banana firms – had no observable data to cause them to lobby for intervention and did not benefit from it.

Firm-Specific Benefits

A third potential explanation maintains that US government officials act on behalf of particular US firms, and that these firms – rather than entire sectors – benefit from regime change.Footnote ¹¹⁶ After it took power in 1951, the Árbenz government in Guatemala expropriated 70 per cent of the United Fruit Company’s land. Company officials, who had extensive ties to the Eisenhower administration, lobbied the US government to overthrow Árbenz. As we have documented, scholars disagree over the extent to which US officials acted at UFCO’s behest (although most scholars argue that security, reputational and anti-communist motives dominated), but one fact is not in dispute: UFCO did not benefit from US intervention. The Justice Department, which had determined that UFCO’s ‘monopoly on banana exports from countries like Guatemala was a violation of American antitrust laws’,Footnote ¹¹⁷ delayed filing suit against the company during the Guatemala crisis, but did so within a few days of Árbenz’s ouster. Although Castillo Armas returned all of UFCO’s confiscated land and drove off all but 0.5 per cent of the peasants who had settled on it,Footnote ¹¹⁸ UFCO could not escape US courts. Despite handing over 100,000 acres of its Guatemalan land holdings, in 1958 the company agreed to divest itself of much of the remainder.Footnote ¹¹⁹ According to Schlesinger and Kinzer, two of the biggest advocates of the economic argument for intervention, ‘the suit had a major impact on breaking up the firm’s banana business and ending its role in Guatemala’.Footnote ¹²⁰

A similar story can be told about Chile. As a candidate for president in 1970, Salvador Allende pledged to nationalize the nation’s copper mines without compensation, a promise that led the Anaconda Company to lobby the Nixon administration to prevent him from coming to power. After being elected, Allende made good on his pledge. The loss of its Chuquicamata mine cost Anaconda between two-thirds and three-quarters of its profits. After Pinochet seized power in the US-backed coup that toppled Allende, the new dictator paid Anaconda $253 million in compensation for the expropriation, but did not return the mine, which remained state property. Anaconda retained some copper services contracts, but according to one history, ‘Anaconda never fully recovered from the loss of its Chilean riches.’Footnote ¹²¹ The company was eventually sold to Atlantic Richfield, which ceased all copper mining operations in 1983.

It is possible that UFCO and Anaconda may have done better under Árbenz’s and Allende’s successors than they would have done if those two leaders had remained in power. Yet the differences, if any, are not large. UFCO, for example, lost 70 per cent of its land to expropriation under Árbenz, but after his overthrow eventually lost all its business in Guatemala despite getting its land back in the short term. Anaconda, which lost everything – its mines and service contracts – under Allende, regained some contracts and monetary compensation from Pinochet, but never got its mines back and eventually got out of the copper mining business entirely. Although the actual outcomes might be slightly preferable to the counterfactual outcomes, neither case provides much evidence for the argument that FIRC provides sizable benefits to US firms. Obviously we are limited in what we can conclude from two unrepresentative cases, but both firms were major economic powerhouses. That they did not gain much – if anything – from these supposedly exemplary cases of economic intervention is cause for some skepticism of economic arguments.

Physical Destruction

A fourth potential explanation of the negative effect of regime change on trade is that the process of imposing a new leader destroys a country’s physical infrastructure and economic resources. This should result in an immediate decrease in trade as the targeted country can no longer produce and ship the same quantity of goods as before the FIRC.

We find little support for this argument. Decreases in trade caused by physical destruction should appear immediately in the years following a regime change. In the absence of repairs and investment, such destruction might cause a permanent decrease in trade. Repairs, however, should cause trade flows to return to or exceed their previous levels. Instead, we find that it takes several years to statistically detect any effect for FIRC: our variable for FIRC becomes negative and significant only after five years. While we do not have direct measures of the physical destruction caused by FIRC, the timing of our negative finding suggests that it does not result solely from damage to infrastructure.

Another possibility is that economic destruction becomes quasi-permanent in the form of a civil war. Ongoing civil war might depress trade flows by threatening property rights, creating uncertainty about physical safety and continuing destruction of physical infrastructure. Our control variable for civil war, however, is not significant in any model specification. This could be read as contradicting our argument that FIRC causes political instability – one form of which is civil war – which in turn reduces trade. We do not interpret the result in this way. Our conception of political instability is far broader than civil war alone. It includes demonstrations, protests, strikes, riots, assassinations, low-level insurgency and various forms of state repression employed in response to these disturbances. For example, of the countries that experienced US FIRCs in our study, only one of them (Chile in 1973) was followed by a civil war, according to the Correlates of War (COW) Intrastate War Data (v.4.0).Footnote ¹²² This conflict was also very brief, lasting only four days. Violence, repression and instability below the level of formal civil war, however, continued in Chile for years. This is also true of the Guatemalan case in 1954, where a burst of violence accompanied the coup but was followed by longer-term instability. By contrast, several of the countries identified by the synthetic controls analysis as experiencing a reduction in trade with the United States – compared to the counterfactual versions of those countries that did not suffer FIRCs – also experienced political violence or insurgency that remained below the level of civil war (less than 1,000 battle deaths). In addition to Chile (which experienced continuing repression by the Pinochet regime after Allende’s ouster in 1973), countries in this category include Haiti (caco rebellions after 1915 and violence between supporters and opponents of Aristide after 1994) and Nicaragua (Sandino’s insurgency from 1927 to 1933). Civil war (at least as coded by COW) is thus not common after the FIRCs in our sample, but lower-level violence and instability is more frequent, and is present in cases that our analysis identifies as enjoying less trade with the United States than they would have had if a FIRC had not occurred. It thus does not invalidate our argument that civil war does not significantly reduce trade.

Reallocation of Resources

The fifth explanation we explore is that FIRC causes leaders to reallocate resources away from the economy. This is an especially strong possibility if, as some primary and secondary sources claim, the purpose of regime change is to enhance security. In this explanation, the government transfers expenditure away from infrastructure, education and health toward the military. Such budgetary changes might cause a decline in economic conditions and make domestic exports less competitive, resulting in a decline in trade. To explore this possibility, we use data from the COW project to consider the impact of FIRC on military expenditure. The results, presented in the online appendix, show that FIRC does not impact military expenditure. While this is not the only direction in which government expenditure may be redirected, it is the likeliest to stand in opposition to economic growth and export competitiveness. As such, we interpret this to suggest that the decrease in trade is not caused by a decline in government investment in the economy.

Uncertainty over Preferences

A final explanation for our findings is that they result from uncertainty about installed leaders. Uncertainty, according to this logic, causes a temporary dip in trade before flows return to equilibrium as businesses acquire information about the new regime’s economic policies.Footnote ¹²³ Uncertainty about leaders’ preferences with respect to trade echoes recent work on bargaining and conflict, in which the duration of leadership tenure is used to proxy for uncertainty about bargaining ranges.Footnote ¹²⁴ To address this possibility, we employ a variety of lag structures for FIRC.Footnote ¹²⁵ If the uncertainty argument were correct, we would expect a brief and sharp decline in trade flows before they eventually returned to normal. Instead, as discussed above, we find that FIRC has no short-term impact on trade. It is only 5 years after FIRC that the coefficient for FIRC becomes significant and negative. This suggests that uncertainty about political stability or policy regime is not causing our negative finding.