What factors matter for trade at the global level? Testing five approaches to globalization,1820–2007

Network society: technology and trade globalization

Some scholars claim that technological change is the principal source of social change (e.g. Lenski, 2005). Important for the current investigation is that the advancement of technology is closely linked to the ability of human societies to discover and exploit new sources of energy (Nolan and Lenski, 2004). To this end, scholars long recognized that advancements in industrial technology decrease transportation costs and increase international trade. Specifically, economic historians find that the expansion of commodities exchange during the mid-19th century is primarily driven by the technological advancements of the industrial revolution (Harley, 1988; North, 1958; O’Rourke, 2002). As a prominent example, O’Rourke and Williamson (1999) study the growth of Atlantic sea trade and discover that the development of intercontinental trade networks during the early 1800s is attributable to the cost-cutting propensity of railroads and steamships. Recent studies observe that sea-freight costs do not decrease during the postwar period and consequently cannot be responsible for the recent explosion of international trade. However, others attribute the past half-century expansion of trade to decreases in the cost of air freight (e.g. Hummels, 2007).

Manuel Castells (1996, 1997, 1998) formulates a technology-centered approach to globalization that attempts to move beyond the traditional industrial-based view outlined above. Referred to as the network society perspective, its major premise is that the world transitioned beyond the industrial paradigm of economic production, to a unique era of human history defined by the use of electronics-based information and communications technology. That is, just as human societies were transformed by the industrial paradigm’s alteration of production and commerce during the early 19th century, the shift to the network society is similarly transforming the contemporary world economy starting in the mid-20th century. Specifically, the explosion of communications technology gives rise to a network form of economic organization, the network enterprise. Unlike previous corporate forms that stress vertically integrated structures which encompass the entire production process, under one roof and at one location, the network enterprise utilizes a horizontal structure that allows corporations to employ a strategy of changing alliances specific to the production of a given product. No longer the hierarchically organized firms of times past, the corporation now serves as subdivided nodes of production which can be roughly grouped into the categories of supplier networks, producer networks, customer networks, and technology-cooperation networks (Castells, 2000).

It is important to note that these observations are not meant to refute the historical evidence which clearly shows the existence of international trade prior to the birth of the network society. As noted by Castells (2004), small-scale global integration certainly did occur before the formation of the ‘new economy’, as even wind-powered ships could travel long distances and establish networks of trade. Nevertheless, pre-network society forms of global interaction could not expand beyond a certain threshold, given the time lag in the communications process as well as the vertical control of trade by states and their subordinates in charge of economic production. As such, what communications technology offers is an expansion in the scale of the global economy, by generating flat organizational structures that permit a wider array of actors to partake in the global trade network (see Castells, 2004: 5).

What should be clear from this discussion is the discrepancy that exists between the industrial- and communications-based views of trade. Thus, it is necessary to distinguish between these perspectives by testing the following set of hypotheses:

Political regimes: democracy and trade globalization

A number of economists and political scientists observe that the recent boom of world trade corresponds with a ‘third wave of democratization’ that began in the 1970s (e.g. Milner and Mukherjee, 2009). Over the years, many scholars studied this relationship in an attempt to scrutinize the link between a country’s political regime type and its level of international trade. Early theoretical propositions claim that autocracies are better positioned than their democratic counterparts to implement trade-conducive policies. As a prominent example, Haggard (1990) argues that autocratic governments are more insulated from the influence of interest groups that lobby in favor of trade protectionism. By extension, the unencumbered political power of autocratic states allows these regimes to increase their tax returns by implementing free trade policies (Haggard and Kaufman, 1995). However, recent works question the notion that autocracies are more conducive for trade. For example, Mansfield et al. (2000) examine the bilateral trade levels of countries paired by regime type. They find that bilateral trade within democratic and autocratic pairs is significantly higher than within mixed pairs. Furthermore, these scholars find no significant difference in the overall level of trade when comparing democratic and autocratic pairs.

More recent works tend to argue that democratic regimes are more likely to engage in trade than their autocratic counterparts (e.g. Przeworski, 1991). According to this line of research, democratization reduces the political power of those that benefit most from protectionist policies, thereby paving the way for the adoption of free trade (Stokes, 2001). Dutt and Mitra (2002) extend the literature by presenting the observation that democratization increases trade, only if a majority of voters will accrue economic gains as a result of liberalization. Still others suggest that increased trade benefits the low-skill segments of the national economy, especially in developing countries. As such, governments of developing economies tend to favor lowering their trade barriers in order to appease the mass electorate (Milner and Kubota, 2005).

However, some suggest that the effect of regime type on trade declines over time. According to this line of reasoning, the political regime of a given nation-state is less important today than the influence of global governance institutions. As such, international bodies such as the International Monetary Fund, World Bank, and WTO are able to exert pressure on all countries regardless of their domestic politics (Friedman, 2005). The growing influence of these international governance institutions – in addition to the lackluster performance of inward-oriented economies, when compared to their export-oriented counterparts – results in a situation where all political regimes gravitate toward a policy of trade liberalization (Dicken, 2003; Nelson, 1990; Remmer, 1990).

Nevertheless, the claim that neoliberalism decreases the impact of political regimes on trade is questioned by various scholars. In fact, many note that the spread of a neoliberal ideology actually enhances the link between regime type and international trade. In particular, democratic countries are generally more exposed to the outside world and are likely to adopt the norms and ideologies of the international community. By extension, democratic states are more likely to succumb to external pressures and adopt free trade policies when compared to their autocratic counterparts (Russett and Oneal, 2001). Thus, insofar as democratic nation-states are more likely to adopt an ideology free trade, regime type should remain a robust predictor of international trade.

World-system: hegemony and trade globalization

World-system theorists argue that national development can only be understood through a comprehensive examination of how local economies interact with the capitalist world-system (Wallerstein, 1974, 1980, 1989). A unique characteristic of world-system theory is its contention that globalization is not a novel development of the 20th century but a historical process underway for at least 500 years. Viewing globalization as being synonymous with the birth of capitalism in Europe during the 16th century, these scholars argue that the capitalist world economy continued to expand outward to gradually absorb regions of the world that previously had little contact with the European world-system (Frank, 1998). ¹ This process of incorporation involved the establishment of a global market and production networks that forcibly entrenched all nations into the Eurocentric world-system, which operates under the single logic of capitalist accumulation (Arrighi, 1994).

For world-system scholars, a comprehensive understanding of the interaction between economic and political–military global social processes is the main foundation upon which globalization should be examined. The most crucial political–military trend in the world-system is the rise and fall process of hegemonic nation-states (Hopkins and Wallerstein, 1979). Specifically, scholars observe that the world economy fluctuates between periods of unicentric and multicentric hegemony (Chase-Dunn and Rubinson, 1979). Unicentric hegemony is characterized by a low level of interstate rivalry as a single polity is powerful enough to provide the resources and rules necessary to produce global peace and stability. In contrast, periods of multicentric hegemony are defined by high levels of interstate rivalry as a multitude of core states compete for power in the world-system (Boswell and Chase-Dunn, 2000). ²

Important to note at this point is that there is a heated debate, especially among political scientists, as to whether the international system is more stable during periods of unicentric or multicentric hegemony. For balance-of-power scholars, geopolitical stability is largely the product of a bipolar/multipolar balance of power (Kaplan, 1957; Waltz, 1979). ‘Balance’ in this context ‘expresses the idea of a counterweight, specifically, the ability to generate sufficient material capabilities to match – or offset – those of a would-be, or actual hegemon’ (Layne, 2004: 106). In other words, given the tendency of powerful states to expand their influence via military conquest, it is argued that the domination of the world-system by a single nation-state is counterproductive for peace in the interstate system (Walt, 1987). In contrast, hegemonic stability theory turns the balance-of-power argument on its head and claims that it is precisely the imbalance of power that is the defining characteristic of a peaceful international system (Gilpin, 1975). Specifically, relative stability exists only when a single hegemon possesses a preponderant level of power in order to supply security and order (Gilpin, 1981; Krasner, 1976). Interestingly, it is clear that military capability is the central determinant of power for these scholars. As argued by Gilpin (1981: 133), it is ultimately stable military relations that produce an orderly international system.

It is, of course, this emphasis on military relations which distinguishes political science from world-system theory in sociology. ³ While world-system scholars largely agree with some of the major tenants of hegemonic stability theory, they differ in their claim that it is ultimately economic power which allows the hegemon to control global affairs (Wallerstein, 1974). Furthermore, although hegemonic stability theorists point toward a normative element in hegemonic rule, it is alleged that non-hegemonic states do not ‘consider certain courses of action because it is obvious they are likely to incur the displeasure of [the hegemon]’ (Knorr, 1975: 10). In contrast, world-system theory does not see the consensual nature of the hegemonic state’s reign as being solely determined by the real or perceived threat of force. Rather, consensus is produced by the fact that the hegemon’s rules-of-the-game generates economic benefits for core nation-states (Wallerstein, 1984).

Thus, largely in line with hegemonic stability theory, world-system scholars argue that economic globalization can only take place when a hegemonic nation-state is powerful enough to provide the stability and rules that are necessary to sustain international trade structures. In other words, periods of unicentric hegemony generate a ‘relatively peaceful international system of states … so merchants trade with one another more freely and more often across international boundaries than they can when the system is splitting into warring factions’ (Chase-Dunn et al., 2000: 80). Furthermore, given the superiority of the hegemon’s economy and the comparative advantages enjoyed by its national industries, hegemonic states traditionally advocate for a global ideology of free trade and commerce (Wallerstein, 1984). This push of the hegemon for free trade relations can be observed at the height of British power during the 1800s and again with US hegemony after World War II (Bairoch, 1993; Kindleberger, 1975).

There are several empirical works that find evidence of a strong positive relationship between hegemony and trade. Some scholars utilize simple descriptive statistics and find that the trade openness of individual nation-states is highly correlated with measures of economic power concentration (Krasner, 1976). Still others discover that bilateral trade relations are more likely to evolve into free trade agreements (FTA) during periods of unicentric hegemony (Gowa and Mansfield, 1993). More recently, Rasler and Thompson (2005) employ vector autoregression (AR) techniques and find that sea-power concentration displays powerful associations with trade globalization. Kwon (2013) further expands the literature by examining the link between hegemony and trade globalization net of multiple control variables, finding that an index of economic and military power concentration is one of the more consistent predictors of trade during the postwar period. This provides the following hypotheses:

World polity: the UN and trade globalization

With theoretical foundations in the literature on the new institutionalism, world polity theorists stress the role of organizational isomorphism, at the global level, and argue that the structure and content of nation-states converge over time (Meyer et al., 1997). The core argument is that the world polity is founded on a unique set of cultural principles that defines the nature and activity of actors in the global social system. Often referred to as ‘cultural scripts’, these principles represent the legitimate values of the global community as they gradually ‘become embedded in social organization, especially in organizations operating at the global level’ (Boli and Thomas, 1997: 172). Thus, as ‘cultural scripts’ become increasingly globalized and begin to operate with a significant amount of independence from nation-states, the more likely they are to create a worldwide convergence in the structures of countries and the values of actors in the global social system (Lechner, 1989; McNeely, 1995).

Consistent with the organizations as institutions literature (Meyer and Rowan, 1977; Stinchcombe, 1965), world polity theorists argue that the institutional structure of the world polity is defined by the total field of international organizations. According to this view, the totality of all international organizations are themselves the carriers of world culture and are responsible for the dissemination of ‘cultural scripts’. As such, both IGOs and international nongovernmental organizations (INGOs) receive the attention of scholars as being representative of the world polity’s institutional structure. INGOs are voluntary not-for-profit international organizations such as Amnesty International, with a membership base composed of individuals. IGOs, which are of particular interest for this study, are international governing bodies created and run by states.

In an influential work, Beckfield (2003) shows that a nation’s total number of IGO and INGO memberships are significantly influenced by state differences in economic development, military power, and culture. This finding combined with his subsequent analysis of the world polity’s network structure leads him to assert that ‘the world polity shows no evidence of flattening. Nor is it becoming a small world. Instead, the world polity more closely resembles “a world of regions”’ (Beckfield, 2010: 1052). Thus, although it is important to note that IGOs are not devoid of power and are influenced by powerful countries, they can also take on a life of their own to constrain and even shape the actions of authoritative member states. Most critically, IGOs like the UN and its predecessor, the League of Nations (LON), produce a relatively high degree of participation and conformity through its resolutions and declarations, given that they ‘are institutional arrangements created and used by state actors … [that] embody cultural assumptions about the world … to set global policies, provide incentive structures for states and other actors, and carry world cultural principles’ (Thomas, 2007: 91).

But how does the creation of a homogenous world culture amplify levels of trade globalization? According to Williamson (1975, 1981), trade is often complicated by opportunism, that is, self-interest seeking with guile, which results in the malfeasance and uncertainty that is inherent in economic exchange (see Beckert, 1996; Erikson and Bearman, 2006). These risks inherent in economic transactions are further amplified when it comes to global commodities exchange, where differences in legal rules, geographical distance, colonial histories, and language all impede bilateral trade (Frankel, 2000; Zhou, 2010). However, Ingram et al. (2005) find that IGO networks are capable of generating the trust, empathy, and sympathy required to overcome these problems of opportunism. In particular, they discover that ‘a doubling of the level of connection between two countries across all IGOs is associated with a 58% increase in trade’ (Ingram et al., 2005: 850). Interestingly, sociocultural IGOs return a more significant impact on trade than economic IGOs, which provides relatively conclusive evidence that ‘the economic impact of relationships depends, to an important extent, on social mechanisms’ (Ingram et al., 2005: 850). This produces the following hypotheses:

Global capitalism: the transnational state and trade globalization

Beginning with the work of Leslie Sklair (2000, 2002b), global capitalism scholars argue that an increasing number of non-state actors play a critical role in global affairs. This argument starts with the premise that capitalist dynamics since the 1980s cannot be explained within the boundary and logic of the nation-state. Instead, capitalist globalization entails the emergence of a transnational capitalist class, which successfully brought together the global elite of several social domains to assist in the creation of an expanding capitalist system: executives of transnational corporations, globalizing bureaucrats and politicians, globalizing professionals, and merchants and the media in the commercial sector (Sklair, 2002a: 145).

Robinson (2004) builds on Sklair’s work to advance the theory of global capitalism by seeing its major components as transnational production, transnational capitalists, and the transnational state. Unlike previous phases of capitalism that saw each state develop a national economy that was only partially connected through trade and capital linkages, the ‘epochal shift’ during global capitalism involves the globalization of the production process itself. Specifically, each element of production is no longer performed nationally but in a new global circuit of production. Central to this transnationalization of production is the rise of the transnational capitalist class, who are the managers of these global circuits of production and exert a great deal of influence on global affairs through their control of ‘a majority of national state apparatuses, advancing their project of capitalist globalization as [they] attempt to achieve a transnational hegemony around the “Washington Consensus”’ (Robinson, 2005: 318).

For these and other reasons, scholars of the global capitalism perspective argue that there is an emergent transnational state composed of supranational political and economic institutions that are infiltrated and transformed by the transnational capitalist class (Robinson, 2007: 131). Particularly important in this regard is their penetration of preexisting neoliberal organizations, such as the WTO. In the words of Robinson (2001),

The global elite set out to convert the world into a single unified field for global capitalism … [and it] pushed for greater uniformity and standardization in the codes and rules of the global market … [by creating] new sets of institutions and forums, such as the WTO. (p. 178–179)

By extension, trade should increase as more nation-states join the WTO, especially when considering that this international body provides a uniform set of rules for the governance of commodities exchange.

Most empirical studies on the effect of neoliberal institutions for international trade are based on examinations of bilateral trade. Most prominently, Rose (2004, 2005) examines the effect of membership in the WTO and its predecessor, the General Agreement on Tariffs and Trade (GATT), on bilateral trade using the well-known gravity model. Contrary to popular belief, Rose finds that the impact of GATT/WTO membership on trade is quite small and inconsistent across different estimation techniques. Specifically, Rose (2005) discovers that although ‘joining the GATT/WTO is associated with a trade-creating effect … simply belonging to it is not’ (p. 692). A somewhat related topic is the literature on FTA. Similar to the research on the impact of GATT/WTO membership, scholars find that the connection between FTA and trade is mixed at best, although some claim that the inconsistent results may be partially attributable to model misspecification (see Baier and Bergstrand, 2007). While acknowledging this disagreement, given the numerous claims that an emergent transnational state generates a unified field for the operation of global capitalism, this study proposes the following hypothesis:

Dependent variable

The dependent variable is trade globalization, measured as the world’s total imports divided by the world’s gross domestic product (GDP). This information comes from Chase-Dunn et al. (2000) who draw on Mitchell’s (1992, 1993, 1995) national estimates of imports to create a measure of trade globalization that extends from 1795 to 1995.

Recent sociological scholarship indicates that the calculation of national GDPs are unduly affected by differences in currency conversion techniques. The currency conversion issue is a serious concern, given the nature of the dependant variable. There are two primary measures used by social scientists to estimate the relative distribution of GDP across nations, exchange rates (FX) that peg national incomes to the US dollar, and purchasing power parities (PPP) which entail the estimation of incomes based on a ‘basket of goods’ estimate. However, Firebaugh (2003) notes that FX measures tend to overvalue goods that are traded internationally, resulting in the undervaluation of the currencies of poorer countries. Furthermore, Korzeniewicz and Moran (2009: 60–63) also show that PPP conversions are unrealistic for research that examines long periods of time, unless PPP weights are recalculated for earlier time periods.

To avoid the currency conversion issue, Chase-Dunn et al. (2000) estimate trade globalization by computing each nation’s yearly trade ratio, separately. This trade ratio is calculated by dividing a nation’s total trade by its total GDP using local currencies in both the numerator and denominator. This approach eliminates the need to convert local currencies into comparable units. Each nation’s yearly trade ratio is then respectively weighted by multiplying these figures by each country’s yearly population ratio, which is a nation’s yearly population expressed as a proportion of the world’s total population. These weighted import figures are then summed for all countries, taken together, and for every year, taken separately, to obtain an accurate measure of international trade (for a more detail, see Chase-Dunn et al., 2000: 84–86).

The Chase-Dunn data ends in 1995. To expand the temporal scope of the estimates and include recent years, the trade globalization time-series is extended to 2009 using the World Development Indicators (World Bank, 2011). Given the previous discussion, it is important to note that the World Development Indicators report international trade figures in PPP. However, since ‘basket of goods’ estimates are more readily available, the use of PPP-converted data for recent time points are less subject to the problems encountered when applying these conversions to historical information. ⁵

Independent variables

Multiple independent variables of interest are utilized to test the hypotheses. The industrial-based view of trade is examined via estimates of energy consumption. This variable represents the per capita converted thousands of coal-ton equivalents consumed by all nation-states from 1816 to 2007. ⁶ The communications-based view is tested by employing a set of three variables. First, the world’s total mileage of telegraph lines comes from Wejnert (2007). Second, the total number of telephones also comes from the same Wejnert dataset until 1970 and is subsequently updated thereafter using information from the World Bank (2011). And finally, statistics for the world’s total number of Internet users is also from the World Bank (2011). ⁷

The second variable is a measure of world democracy that comes from the Polity IV project available at http://www.systemicpeace.org/polity/polity4.htm (Marshall and Jaggers, 2008). The Polity IV project estimates each nation’s democracy level through a set of national political characteristics, including (1) the presence of mechanisms that allow citizens to express their policy and leadership preferences, (2) the existence of institutionalized constraints on the powers of the executive, and (3) the guarantee of civil liberties to all citizens in their daily life and political participation. Marshall and Jaggers use these criteria to create a polity index that ranges from −10 (strongly autocratic) to +10 (strongly democratic) for each country. These state-level democracy scores across are summed for all countries and for every year. These figures are then divided by the number of countries included in the summed total to generate an average annual democracy score. ⁸

World-system views on trade are assessed by way of two variables, hegemony, and great power war. Starting with the former, while sociologists traditionally focus on the role of economic power in their conceptualization of hegemony, political scientists Modelski and Thompson (1988, 1996) persuasively argue for the use of sea power. ⁹ As such, both economic and military indicators are used to formulate an index designed to measure the power of the hegemonic nation-state. The economic indicators are GDP and GDP per capita from Maddison’s (2007) estimates available at http://www.ggdc.net/maddison/, while the military indicator is a measure of naval power taken from Modelski and Thompson (1988). ¹⁰

In order to measure the relative distribution of power in the world-system, the economic and military indicators are measured as a relative proportion of the world’s total/average and combined to create a country-specific score. In particular, the GDP of each nation-state is converted to a proportion of the world’s total GDP, GDP per capita is scaled by the world’s average GDP per capita, and sea power is estimated by dividing the total number of warships in the possession of a given nation-state by the total number of world warships. The hegemony index is then created by multiplying these estimates to obtain a numerical indicator of economic and military power. ¹¹ As a final step, these national-level indices are then compared to determine which country possesses the highest hegemony index and would subsequently serve as this study’s measure of hegemony. ¹² Thus, England serves as the hegemon from 1820 to 1918 and the United States from 1919 to 2008. This, of course, is consistent with theoretical accounts (Hopkins and Wallerstein, 1979; Modelski, 1987).

As for great war intensity, this variable measures for those years when trade may be negatively affected by the impact of war between great powers. These data are acquired from Levy (1983), who defines a great war as those military conflicts involving at least one great power fighting on opposing sides. Although Levy calculates great power war as battle casualties divided by the total population of Europe, this study converts the raw casualty figures to a proportion of world population. Levy’s list of great wars since 1820 includes the conflicts surrounding the Crimean and Franco-Prussian Wars during the 1850s and 1870s, World Wars I and II in the 1910s and 1940s, and the Korean War during the 1950s. This variable logged as a test of power transformations indicates a strong positive skew.

World polity contentions are captured with multiple variables. First, IGO proliferation is measured via a set of three covariates: first, the total count of IGOs; second, the average number of member states per IGO; and finally, IGO network density or the total IGO memberships for all states divided by all possible memberships. The second world polity measure is the total number of member states in the LON and its successor the UN. This variable extends from the establishment of the LON in 1920 ¹³ to the last year UN membership data are available in 2007. ¹⁴ Estimates for the LON/UN variable are only available in 5-year intervals prior to 1965 and is, thus, supplemented with information from http://www.unog.ch/.

Global capitalism contentions are analyzed by a measure that represents the total raw count of member states in the GATT and its successor the WTO. ¹⁵ This variable extends from the inception of the GATT in 1947 through the last year WTO membership information is available in 2007. The GATT/WTO information is from the aforementioned Correlates of War Project, which only report membership data in 5-year intervals prior to 1965. As such, this information is supplemented with the more complete annual GATT/WTO membership data available at http://www.wto.org/

Finally, the regression models also include a measure for world GDP per capita growth from Maddison (2007). Most importantly, this covariate controls for the effect of world development on trade globalization. ¹⁶ Interesting to note is that this variable may also serve as a relatively accurate alternative estimate of transportation cost, as a separate analysis indicates that world GDP per capita shares a −.858 correlation with air freight and a −.815 correlation with sea freight. Additionally noteworthy is that energy consumption, which serves as this study’s proxy for industrial technology, shares a correlation of −.877 with air freight and −.802 with sea freight. ¹⁷

Statistical techniques

The time-series dataset analyzed in the current investigation contains high levels of volatility that varies across time. As such, the usage of the more conventional autoregressive integrated moving average (ARIMA) regressions may be problematic as this technique is best suited for data that display a relatively consistent level of variance over the life of the time-series. In contrast, the ARCH regression technique is a popular statistical approach employed by researchers to find parsimonious and unbiased parameters for data in which volatility rates vary across time. The usefulness of this approach is the ARCH technique’s ability to group together periods of low and high volatility, separately, in order to fit a conditional maximum likelihood specification, whereby variation in the dependent variable is expressed as a time-dependent function of prior observed volatility in the independent variable (Wooldridge, 2009). ¹⁸

Most importantly, the ARCH technique allows for the introduction of structural components, by way of regressors, that controls for multiplicative heteroskedasticity. The determination of the proper ARCH structural component requires a number of preliminary tests to ensure the proper ARCH specification for each individual model. These ARCH models are expressed in a (a, g) format, whereby a represents the ARCH specification, and g denotes the GARCH, or generalized, specification. Specifically, the benefit of the a and g structural regression components is that they allow the researcher to identify and control for any time-specific heteroskedasticity patterns that may be present in the data. For example, an ARCH structural model of (1, 0) controls for a first-order ARCH, only, while an ARCH of (1/2, 1) introduces a first- and second-order ARCH together with a first-order GARCH.

The identification of the proper a and g value is an involved multistep process as summarized in Table 2. First, ordinary least squares regressions are estimated for each model as shown in Model A. These ordinary least squares regressions are then analyzed for ARCH effects by way of the Engle multiplier test. This allows the researcher to determine whether the ARCH approach is better suited than the more traditional ARIMA. According to these diagnostics, all models reported are strong candidates for the ARCH technique. Second, an array of ARCH regressions is specified for each model to determine the proper value for a and g. During this stage, every potential time-specific structural component up to 15 years is scrutinized to uncover the most significant a and g value. ¹⁹ As shown by a comparison of the ARCH regressors and the Bayesian information criterion (BIC) for Models B and C, all reported models in this investigation reveal that ARCH (1, 0) returns the more robust and parsimonious results compared to ARCH (1/2, 0). Finally, ARCH (1, 0) is compared to the GARCH-included (1, 1). This step reveals that GARCH-included models produce the best fit as shown by a comparison of the BIC for Models B and D. When taken together, these diagnostics show that GARCH-included models, that is, Model D, outperforms all potential alternatives.

OPEN IN VIEWER

Table 2.

Preliminary diagnostic models.

	Model A – OLS	Model B – ARCH (1)	Model C – ARCH (1/2)	Model D – ARCH (1, 1)	Model E – AR (1)	Model F – MA (1)
GDP per capita growth	0.218	1.118**	1.039**	1.047**	1.047**	1.047**
	(0.55)	(3.22)	(2.81)	(3.24)	(3.25)	(3.26)
Energy consumption	0.731**	0.549**	0.566**	0.606**	0.603**	0.603**
	(4.22)	(5.32)	(5.51)	(5.81)	(5.73)	(5.73)
World democracy	0.470 +	0.412**	0.414**	0.388**	0.391**	0.391**
	(1.84)	(2.89)	(2.61)	(2.85)	(2.86)	(2.87)
Countries	0.197*	0.095	0.061	0.036	0.034	0.033
	(2.25)	(1.08)	(0.58)	(0.34)	(0.32)	(0.32)
WDI control	0.017	−0.001	0.004	0.004	0.004	0.004
	(0.53)	(−0.02)	(0.09)	(0.14)	(0.15)	(0.15)
Constant	−0.019 +	−0.023*	−0.024*	−0.025*	−0.025*	−0.025*
	(−1.81)	(−2.15)	(−2.19)	(−2.49)	(−2.50)	(−2.51)
ARCH specification [a, g]		[1, 0]	[1/2, 0]	[1, 1]	[1, 1]	[1, 1]
ARCH (1)		0.411**	0.264**	0.313**	0.313**	0.313**
		(3.24)	(2.65)	(4.15)	(4.13)	(4.13)
ARCH (2)			0.162
			(1.60)
GARCH (1)				0.642**	0.641**	0.640**
				(9.47)	(9.36)	(9.35)
ARIMA specification [p, d, q]					[1, 0, 0]	[0, 0, 1]
AR (1)					−0.024
					(−0.25)
MA (1)						−0.027
						(−0.28)
R 2	0.13
χ 2		51.36**	55.20**	53.49**	53.49**	53.55**
Box–Ljung (15 years)	20.95	12.76	13.38	13.41	12.59	12.48
BIC	−280.08	−298.96	−296.63	−303.71	−298.56	−298.57
N	186	186	186	186	186	186

OLS: ordinary least squares; GDP: gross domestic product; WDI: World Development Indicator; ARCH: autoregressive conditional heteroskedasticity; ARIMA: autoregressive integrated moving average; AR: autoregression; MA: moving average; BIC: Bayesian information criterion.

Variables are converted to percent change scores; t-values are in parentheses; GDP per capita growth measured at t−1

+

p < .10; *p < .05; **p < .01 (two-tailed tests).

In addition, the advantage of ARCH is its ability to include in its parameters the classical ARIMA structural components (Box et al., 2008). This allows for the control of higher order AR and also includes a moving average (MA) option that controls for error-term correlation. These ARIMA controls are specified in a (p, d, q) format: p represents the AR control, q the MA control, and d controls for nonstationary variables by differencing the data. For example, a (1/2, 0, 0) specification controls for first- and second-order AR, while a (0, 1, 1) specification differences the variables and controls for first-order MA. Similar to ARCH models, the identification of ARIMA regressors requires the researcher to perform a comprehensive set of diagnostics to find the proper (p, d, q) for each model.

As a preliminary step, all variables are tested to ensure they are stationary. Dickey–Fuller tests indicate that all variables used in the analysis are indeed stationary, which is expected when converting variables into a percent change format. The stationary nature of the variables indicates that models do not necessitate a d specification. The residuals for each model are then tested for AR by evaluating their autoregressive and partial autoregressive processes, which allows for the identification of a proper value for p. According to this step, no specification in this investigation contains a significant amount of AR when controlling for their ARCH effects. Nonetheless, every model is retested by including both AR and MA controls. As shown by the AR and MA regressors included in Models E and F, there is no significant AR or MA in any of the ARCH-controlled models of this study. All-in-all, the BIC indicates that the ARCH- and GARCH-controlled Model D is the most parsimonious specification. As a concluding step, the Box–Ljung test for ‘white noise’ is employed to ensure that the residual of every model is free of autoregressive and error term–correlative processes after outfitting the proper ARCH and ARIMA model. These diagnostics confirm that there is no significant AR or error-term correlation in the first 15 years of all reported equations.

Every equation reported in this study includes a variable that controls for the changing number of countries included in the measurement of trade globalization. This is important, given that the calculation of the dependant variable includes a varying number of nation-states, with information for more countries becoming increasingly available over time. In addition, since two data sources are used to construct the dependent, models are net of a dummy control for World Development Indicators (WDIs). The correlation matrix is available in Appendix 1. ²⁰

Theoretical and empirical advantages of a global-level dataset

To reiterate from earlier sections, the global-level dataset employed in the current investigation is generated from the aggregation of time-series cross-national data. It is vital to note that time-series cross-national designs explore the relationship between the independent and dependent variables at the national level, only. As a result, these approaches do not allow for an examination of the impact of the independent variables outside of the respective cross-sections in which the effects are being assessed. This is problematic when considering the possibility that a variable may augment trade through effects that ‘spillover’ into different countries. ²¹ Indeed, there is evidence which shows that ‘institutions such as the GATT/WTO create rights and obligations for nonmembers and therefore can have surprisingly broad effects’ (Goldstein et al., 2007: 38). This is also the case with the LON/UN which produces cultural scripts that shape the actions of both member and nonmember states (Thomas, 2007). Thus, a global-level approach is best suited for this study as it allows for an assessment of a variable’s impact, not within nations, but in the entire world.

Beyond these methodological issues, maybe most important is that the strategy of utilizing a global-level dataset is more consistent with the nature of the theoretical perspectives from which the hypotheses are derived. For world-system theorists, the ‘world-system is all of the economic, political, social, and cultural relations among the people of the earth … it is the whole interactive system, where the whole is greater than the sum of the parts’ (Chase-Dunn and Grimes, 1995: 389). World polity scholars assert that the ‘operation of world society through peculiarly cultural and associational processes depends heavily on its statelessness’ (Meyer et al., 1997: 145). Global capitalism scholars claim that the global economy ‘requires a centralized authority to represent the whole of competing capitals, the major combinations of which are no longer “national” capitals’ (Robinson, 2001: 167). And finally, the network society view asserts that

The network society is a global society … everybody is affected by the process that take place in the global networks of this dominant social structure. This is because the core activities that shape and control human life in every corner of the plant are organized in these global networks. (Castells, 2004: 22)