There are two crucial questions. What are the ‘chief’ quantitative approaches in comparative international sociology? And to what extent do they build upon Charles Tilly’s (1984, 14) endeavor to establish ‘concrete descriptions’ of ‘big (macro-social) structures and large (macro-social) processes within the global frame of reference, in which, for example, prodigious abstractions like nation-states, multinational firms, international social networks (e.g. interlocking corporate directorates, etc. form tangible macro-social units of analysis, such that large-scale macro-social processes, like proletarianization, urbanization, capital accumulation, bureaucratization, etc., lend themselves to empirical generalizations? (1984, 61) specifically with respect perceived complex interdependencies between certain ‘structural’ features of macro-social units (Ragin 1987, 55)?
According to Charles Ragin (1987, 55), quantitative methods in comparative international sociology recognize that macro-social units have features “which interact in the sense that changes in some features produce changes in other features, which in turn may produce changes in others”. ‘Features’ of social structure are characteristics viewed as more or less permanent social attributes, since they are thought to be slow in changing. As such, social phenomena that capture their interconnections are considered ‘permanent causes’ since they concern social processes that are fundamentally enduring properties of macro-social units. (Ragin 1985, 55, cf. Mill 1843). The intension is to probe data on macro-social structural attributes in terms of ‘variables’, in order to verify inferences regarding social causation by demonstrating the degree to which statistical uniformities (or irregularities) of one set of macro-social units are accompanied with uniformities (or irregularities) in others.
The dissection of ‘permanent causes’ entails concomitant statistical ‘control’ for extraneous variation, i.e. circumstantial, historically-specific social contingencies, since these factors prevent functional equivalence, which is imperative for drawing ‘intelligible comparisons’ between macro-social units. Diametrically opposed to this is the self-contradiction of circumstantial holism, the historicist conviction that only ‘incomparables are comparable’, i.e. when all that a comparative study does is illustrate a theoretically formed concept by describing an empirical manifestation that is in some sense interpreted as sociologically ‘peculiar’ (Zeldith 1971, 271 ; this particularism is contrarian in that it intrinsically assumes comparative social research for producing ‘general knowledge of human behavior is not possible” (Zelditch 1971, 278, 283; Ragin 1987, 56).
What follows is that the drawing of empirical generalizations does not arise unless they are ‘explanatory’, i.e. they elucidate causal associations macro-social units with respect to certain macro-social characteristics. In the sense, the empirical research design is infused with theoretical formulation" (Zelditch 1971, 270, 307), that is, it pivots on the objective of empirically ‘testing’ a priori assumptions drawn from interrelated substantive ideas about the social world (social theories) in conceivable, sociologically imaginable frameworks. These frameworks, in turn, provide the ontological mappings, and thus research questions concerning the nature of social interactions, the coherent sense of meaning for purposes of comparability (Zelditch 1971, 282) of the macro-social units in question.
The logical corollary is ontological hypostatization of their perceived interdependent and interrelated social characteristics and social processes, such that heuristic devices are constructed for formulating working hypotheses in order for quantitative analysis to derive plausible specifications of particular social phenomena, of interest, involved in such interactions. For example, “the state is the unit in investigations of economic development because it is the largest permanent source of the kinds of decisions that determine the conditions of economic growth. One of the benefits of viewing the matter this way is that it makes it possible to ask whether particular states satisfy the theoretical requirement: is Monaco a state for this purpose?” (Zelditch 1971, 282) To this end, what are the ‘chief’ empirical methods employed, and what
One approach is multiple linear regression, in which the social scientist regresses a single dependent variables, such as growth in gdp per capita (Kentor & Boswell, 2003) is regressed on a vector of several independent variables—such as levels of economic development (GDP), the extent to which an economy is liberalized, export commodity concentration, etc. In this case, when the independent variables result in a lack of correlation with each other and only moderately correlated with the dependent variable, the coefficients in the regression analysis can be interpreted with ease as directly a statistical process that measures the effect of the independent variables on the dependent variable. Nevertheless, since the international social world is our oyster for social research, independent variables are rarely not highly correlated with each other, and moreover, lagged versions of dependent variables are often themselves used as independent variables. Moreover the behavior of variables in multiple regression models is the subject of controversy, as exposed by the now infamous so-called ‘denominator effect’ debate on the extent to which foreign capital (FDI) in less developed countries either contributes or does not contribute to economic growth (Firebaugh, 1996; Dixon & Boswell 1996, Kentor & Boswell 2003).
Moreover, Tilly (1984, 14) mentions the specificity of time. This is where longitudinal analysis with panel data is appropriate. One model is a static panel analysis with the lagged dependent variable; static because change in the independent variable is not analyzed, rather, the attempt is to assess the effect an independent variable that only occurred once in time. This model is useful if the dependent variable measured at two time points in time, and, thus, the focus is on time-invariant predictors, such as the social locations of gender, class, or geographic allocation of a nation-state in the capitalst world economy. The strength of this model is that by including a lagged dependent variable, one accounts for prior level of the dependent variable, in order to determine to what extent the effect independent variable a time-invariant measure, on the change in the dependent variable. Despite such insight one can gain from the methodological approach, the drawback is that is “vulnerable to omitted variable bias and [thus] cannot capture stable, unobserved differences among countries” (Brady, Kaya, and Beckfield 2007)
Another longitudinal model with panel data is the fixed effects model, which is used assessing dependent and independent variables measured at three or more time points in time. On the one hand if data is data available at even time intervals, the panel data set is ‘balanced’. On the other hand, if there are uneven intervals with missing data, panel data set is ‘unbalanced’. For the empirical analysis Brady, Kaya, and Beckfield (2007), in order “to be included in [the] sample, a country had to have a real per capita GDP of less than $5,000 and a population of 500,000 in 1980, and data had to be available for at least two time points […] because of missing data, [the panel set] is unbalanced. Nevertheless, with the conduction of a sensitivity analysis in the appendix, the empirical generalization of the insignificance of GDP growth on infant and child mortality is unchanged.
In general, fixed effects models are appropriate if a researcher is interested in ‘within case’ variation over time. As such, a fixed effects model would not be useful if key variables of interest are time invariant. In addition, the fixed effects model is not suggestible if the research design calls for ‘between’ case variation e.g. an FE model would not be sufficient to examine factors that contribute to variation in income inequality because most of this variation is between country rather than within country variation.
Another prominent quantitative method in comparative international sociology is social network analysis. Alderson & Beckfield (2008) use social network analysis to explore the structure of the ‘world cities’. Sassen (2001) presupposes that, depending on scale, pertinent cities contain an agglomeration of central functions for ensuring the fluidity global capitalism; “[the] focus [is on the] production of those inputs that constitute the capability to control [the capitalist world economy].” Such cities are geographical scales of specialized ‘producer’ services of transnational corporate activity, which range from the concentration of managerial consulting, to financial innovation, to advertising, to knowledge creation (cf. Matthiessen et. al, 2010). In this sense, the extent to which steer global capital accumulation is considered in relational terms, as the product of interconnected networking activities embodied in logistical flows of international trade.
Capitalist globalization ensues dynamic connective configurations between commercial metropolises in relationships of complex interdependency. As such, ‘world cities’ are measured on whether they house high versus low quality value-added producer services for the world economy to function accordingly. The salient point is that transnational corporations are vertically disintegrated and horizontally integrated globally; multiple interlocking corporate offices are required in major cities around the world to underpin the accretion of surplus value-the more important the office, in terms of capital accumulation, the more important the ‘world city’ (Pereira & Derruder, 2010).
It is important to note, however, that ‘world cities’ do not have power in and of themselves; they have power to the extent that they function as command points and centers of planning and thus establish the framework in which other cities operate in the capitalist world economy. Alderson & Beckfield (2004) assess the degree of power world cities and the positions that they hold in the capitalist world economy by focusing, with the use of social network analysis on key relations linking cities, that is, between multinational enterprises (MNEs) and their subsidiaries. Referencing Hymer (1972), Alderson & Beckfield (2004) note that capitalist globalization represents the concentration of administrative functions of transnational corporations in major of the traditional center, while major cities of the periphery are consigned to mere ‘branch plant’ subordinate statuses—an extrapolation of ‘unequal exchange theory’ from dependency/world-systems analysis. Using regular equivalence blockmodeling, the empirical generalization that is advanced is that ‘world cities’ are broadly shaped in a substantively significant fashion by traditional world-systemic center-periphery interactions. ‘World cities’ located in core countries have grown more powerful and prestigious in the recent decades, while non-core ‘global cities’ are more likely to play passive supporting roles.
The empirical research of Derruder et. Al (2010), on the other hand, based on estimated size and worldwide distributions of corporate offices, suggest that the world city network does not necessarily map onto the classic world-system hierarchy. The spread of management consultancy offices to skilled labor forces in cities like Shanghai and Beijing confirm the conjecture that globalization cuts across strict inequalities between rich and poor countries.
Quantitative social network analysis allows researchers to patterns of flows, exchanges, or linkages between articulated nodes, e.g. major cities, in order to reveal the patterning of intricate social connections that, in the final instance, constitute and entire systemic network of positional social entities, which in the case of world-cities, as analyzed above, allows for an ideal matching between theory and empirical research methodology. In addition to examining similarity in relations between nodes network analysis allows for examining centrality, although methodological issues with this process still exist. As Aldrson & Beckfield (2004) ague, current empirical research designs using social network analyses with respect to world cities are based on current geographical mappings, not to the degree of slippage betweentwo relative to some point in the past. As such, future empirical research could test the longitudinal hypothesis that globalization is leading to a reconfiguration of traditional core-periphery relations, and thus a remaping world-systemic structure of the hierarchy of world cities. “Are the two actually tightly linked? Or are the processes that generate the global urban hierarchy largely independent of those that generate a core/periphery structure in the [world] system (Alderson & Beckfield 2004).
Halaby (2004) writes, “panel studies are fast displacing their cross-sectional counterpart at the heart of sociological research”. Longitudinal data analysis is the future for quantitative comparative international sociological research. As such, my plan is to get more acquainted with the empirical techniques involved in this line of inquiry.