The Time-Space Mapping of World Labor Unrest[1]




Beverly J. Silver

Sociology Department

The Johns Hopkins University

Baltimore, MD 21218



I.                    Theses


In Forces of Labor[2], I put forward a set of theses about the time-space dynamics of world labor unrest from the late-nineteenth century to the present. They can be summed up as follows:


1)      The main location of working-class formation and protest has shifted within global industries along with shifts in the geographical location of production (spatial fixes). Major waves of labor unrest are both a significant cause and a significant effect of this process.


2)      The main sites of working-class formation and protest has shifted from industry to industry together with the rise/decline of leading sectors of capitalist development (product fixes).


3)      Intra-industry spatial shifts (thesis #1) tend to be from core (high wage) to more peripheral (low wage) locations (consonant with the expectations of product cycle theory).


4)      Technological fixes (the reorganization of the labor process and the introduction of new technologies) have tended to re-establish the competitive advantage of core locales, leading to a reconsolidation of production in the core, and a concomitant reversal of the core-periphery shift in working-class formation/protest.


5)      With each spatial fix (within a product life cycle), new working-class formation and protest takes place in an increasingly competitive environment, making it more difficult to secure the resources needed to establish stable labor-capital accords and bring labor militancy under control. This thesis is consonant with the thesis that sees the semiperiphery (and increasingly the periphery) as a “zone of turbulence”.


6)      Variations from the above dynamics are to be expected as a result of contingent (although not random) outcomes of political struggles that shape relations among labor, capital and states (see for example the discussion of the “Japanese anomaly” in Forces of Labor, chapter 2). 


The foregoing theses focus on world-economic dynamics; however, the time-space patterning of world labor unrest is also shaped by (and shapes) world-political dynamics. As such:


7)      World wars have had a strong effect on the overall pattern of labor unrest: world labor unrest rose on the eve of the world wars, declined during the initial years of the wars, and exploded in their aftermath. This pattern is characteristic not only of the belligerent countries, but also of countries not directly involved in the fighting. The above relationship is less strong in the case of wars that are not world wars.


8)      Periods of world hegemonic crisis/breakdown have been periods of relatively high levels of “dysfunctional” social conflict (including dysfunctional labor-capital conflict). Periods of world hegemony have been periods of relatively stable social compacts and low levels of “dysfunctional” social conflict. The “dysfunctional” social conflict that exists tends to be localized outside the core in periods of world hegemony (consonant with thesis #5); it tends to become more spatially widespread in periods of world hegemonic crisis/breakdown.


9)      World labor unrest in periods of hegemonic crisis/breakdown has shaped the institutional structures of subsequent hegemonic world orders in significant ways, transforming the social-political terrain on which world labor unrest unfolds.


10)  Industrialized warfare in the twentieth century increased labor’s bargaining power. Post-industrial warfare in the early twenty-first century has weakened labor’s bargaining power.


One underlying assumption of the foregoing theses is that the outcome of waves of labor unrest depends in important ways on the nature and extent of workers’ bargaining power. Spatiality, in turn, is an important component of the conceptualization and measurement of the main forms of workers’ bargaining power. Thus: Workplace bargaining power is defined as the power that results from the ability of strategically located workers to disrupt production in an entire workplace, firm, industry, national, regional, and/or global economy (or an entire network of distribution, as can be the case with transport workers). Associational bargaining power is defined as the power that comes from the collective organization of workers, which in turn, is shaped by the location of workers within such non-workplace networks as those of kinship, neighborhood and community.




II.                 Methods of Analysis



Relational Processes


A central premise underlying the above theses is that workers and workers’ movements located in different states/regions (and different time periods) are linked to each other by the world-scale division of labor and by global political processes. Thus, an understanding of relational processes among “cases” on a world scale across both time and space is fundamental to understanding the dynamics of labor movements.


It is helpful to distinguish between two different types of relational processes:  (1) “direct” and (2) “indirect”. In the case of direct relational processes, the actors are aware of and consciously promoting the links among the cases. These direct relational processes can be further subdivided into two different forms: diffusion and solidarity. In the case of diffusion, actors located in “cases” that are separated in time and space are influenced by the spread of information about the behavior of others and its consequences. “Social contagion” is a common image used in the methodological literature on diffusion. Historically, labor migration has been an important process promoting the diffusion of political ideologies and forms of labor protest. Diffusion can take place without active cooperation between the source site and recipient site of the “social disease” (e.g., as with the spread of information through the print or electronic media). In contrast, the second form of direct relational processes—solidarity—involves personal contact and the development of social networks and organizations—transnational social networks and organizations in the case of labor internationalism.


The specific contribution of the world-systems perspective to social analysis is its focus on what I have called indirect relational processes. Indeed, the theses outlined above mostly refer to indirect relational processes. In the case of indirect relational processes, the affected actors are often not fully conscious of the relational links. Rather, actors are linked behind their back by systemic processes including the unintended consequences of a series of actions and reactions to the contradictions characteristic of historical capitalism. Thus, for example, if capitalists respond to a strong labor movement by relocating production to a new site (thus weakening labor in the de-industrializing site but strengthening labor in the industrializing site), then we can say that the fates of these two labor movements are linked by indirect relational processes.


Taking the example of late-nineteenth-century migration, we can detect both indirect and direct relational processes linking labor movements across time and space. The spread of labor movement ideologies and practices as workers moved across the globe is an example of diffusion. But, we can also detect critical indirect relational processes. The U.S. labor movement’s success in having open immigration outlawed in the 1920s set the stage for the stabilization of the U.S. working class and contributed to the subsequent CIO (Congress of Industrial Organizations) victories in the 1930s. At the same time, however, this U.S. labor movement “success” shut off what had been an essential safety valve for Europe in the nineteenth century (mass emigration). It thereby transformed the terrain on which workers’ movements operated in Europe, and according to E.H. Carr[3], helped set the stage for the defeat of European labor movements and the rise of fascism.




Limits of the Comparative-Historical Approach


If we are to make relations among the cases across time and space a central part of the explanatory framework, then we cannot rely on the comparative-historical strategy of research. The comparative-historical perspective, like the approach taken here, criticizes the strategy of making generalizations from one or a limited number of cases, and thus calls for widening the geographical scope of analyses. Much of the comparative-historical literature follows the strategy of “splitting” in search of distinctiveness in contrast to a strategy of “lumping” cases in search of commonalities and generalizations (Hexter 1979, 241-2; Collier and Collier 1991, 13-15). The problem arises, however, in the next research step in comparative-historical analyses—that is, tracing the differences in outcome to preexisting and independently produced differences in the internal characteristics of the various cases.


While comparative-historical sociologists have produced some of the most interesting recent scholarship in labor studies, the research strategy nonetheless impedes full access to what we take to be a key explanatory variable of labor movement behavior and outcome (i.e., the relationships among the cases themselves). As Charles Tilly among others has pointed out, the results of a strict cross-national analysis may be misleading. A social unit’s connection to the whole system of social relationships in which it is embedded “frequently produces effects [that] seem to be autonomous properties of the social unit itself.”[4] As a result, the patterned diversity among the social units appears to be consistent with cross-national variation-finding explanations. This has been referred to as “Galton’s problem” in the anthropology literature: that is, in a situation in which cases are presumed to be independent—but are actually linked relationally—the relations among the cases become a lurking (unexamined) variable. Indeed, in the theses outlined above, similarity/variation among labor movements is not merely the outcome of the cases’ similar/different independent and pre-existing characteristics. Rather, the relationships among the cases, and relationships between the cases and the totality, are key parts of the explanation of similar/different outcomes.[5]


In sum, we require an analytical strategy that is sensitive to the relational processes among key actors (labor, capital, states) in the world capitalist system as a whole, as well as the systemic constraints affecting those actors. Needless to say, such an approach presents enormous problems of complexity, and a strategy for reducing complexity and making research feasible is needed.



Encompassing Versus Incorporating Comparisons


The most well known strategy for reducing the complexity of world-historical analysis is what Tilly labeled “encompassing comparison” and is best illustrated by Immanuel Wallerstein’s approach to the study of the “modern world system” and John Meyer’s approach to the study of “world society”. Encompassing comparisons reduce complexity by starting “with a mental map of the whole system and a theory of its operation.” Similarities/differences in the attributes and behavior of the units are then traced to their similar/different position within the overarching totality.[6]  Meyer’s “mental map” of the system leads him to emphasize a growing convergence among national cases as a result of a world-scale process of “rationalization”. Wallerstein’s mental map, in contrast, leads him to emphasize a process of recurrent geographical differentiation among core, semiperiphery and periphery, resulting from the unequal distribution of rewards in a capitalist world economy. Yet for both, local attributes and behavior are seen as the product of a unit’s location in the system. The larger system has a steamroller-like quality, transforming social relations at the local level along a theoretically expected path. This approach has led to complaints from otherwise sympathetic scholars that “world systems theory” in “assuming the systematicity and functionality of the capitalist world system”, has produced a “mechanical picture of different labor forms in different parts of the world.”[7]


The strength of Wallerstein’s perspective is that it emphasizes the very real constraints that the totality imposes on the range of possible action open to local actors. But its weakness is that it excludes a priori a situation in which local action (agency) significantly impacts local outcomes, much less a situation in which local agency impacts the operation of the system as a whole (as are foreseen, for example, in Theses #6 and #9 above). Thus, while keeping in focus the real systemic constraints that the totality imposes on local actors, the “encompassing comparison” approach was not adopted in Forces of Labor as a strategy for reducing complexity. Instead, the research strategy adopted is closest to what Philip McMichael has called “incorporating comparisons”—a strategy in which the interactions among the multiplicity of subunits of the system are seen as creating the system itself over time. The resultant conceptualization is one in which relational processes in space unfold in and through time.[8]


Constructing Narratives as a Mode of Causal Analysis


The approach outlined above necessarily leads to a narrative mode of causal analysis—in some ways similar to that advocated by historical sociologists. The narrative strategy, Larry Griffin has argued, allows us to understand social phenomena “as temporally ordered, sequential, unfolding and open-ended ‘stories’ fraught with conjunctures and contingency”.[9] As a strategy for explanation, “descriptively accurate narratives, which depict a sequence in chronological order… do more than tell a story,” according to Jill Quadagno and Stan Knapp. Such narratives can “serve among other purposes, to identify causal mechanisms” because “when things happen…affects how they happen.”[10]  But while historical sociologists have stressed the importance of treating time as dynamic, they have generally continued to treat space as static (e.g., conceptualizing national cases as fixed, independent units). In contrast, Forces of Labor sought to create a narrative of working-class formation in which events unfold in dynamic time-space. 


Terrence Hopkins’ contention about the purposes and limits of statistical elaboration in the historical social sciences is relevant here. Hopkins argued that the purpose of statistical elaborations should not be “explanation” but rather should be the identification of patterns across time and space that then become the explicandum of a multidimensional causal “story”.  Capturing the causal processes themselves, Hopkins argued, requires a narrative rather than a statistical mode of analysis.[11] 


Question for the Conference Specialists: An obvious question to pose now, in the context of this conference, is whether GIS in combination with network analysis and/or hierarchical linear modeling can offer more sophisticated tools for the clear identification of patterns across time and space as the first step in setting out the “what is to be explained,” with the second step being the construction of a compelling causal narrative? Is this a widely accepted way of using GIS (with or without HLM and/or network analysis), in contrast, say, to the almost universal understanding that regression analysis is a tool for causal modeling rather than pattern identification?  (Indeed, when I have used regression analysis as a pattern identification tool, readers have almost invariably misunderstood what I was doing; they assumed I was attempting to build a causal “model”, and critiqued it on those grounds—despite explicit statements made about the methodological approach being used.)



Mapping Patterns with the World Labor Group Database


In order to pursue the research strategy described above, we need empirical maps of the time-space patterning of the phenomena of interest—in the case of Forces of Labor, an empirical map of the time-space patterning of labor unrest. The map is needed to allow us to navigate a path through the bewildering totality of potentially relevant episodes of the phenomenon of interest in the time-space span of interest. In other words, it allows us to identify patterns across time/space and thus make informed decisions about what/where/when to study more closely. The empirical map allows us to “lump” and “split” cases as a tactic for uncovering the patterns to be explained through the construction of relational narratives.


The first problem is that for most phenomena of interest, information of sufficient geographical and historical scope to construct the type of empirical map needed is not readily available. Forces of Labor relied on the World Labor Group (WLG) Database, which was constructed specifically for the purpose of this and related research. Building on a well-established tradition within the social sciences, the database was constructed by using information from newspaper reports of labor unrest (strikes, demonstrations, factory occupations, food riots, etc.) throughout the world beginning in 1870. The result is a database with over 91,947 “mentions” of labor unrest for 168 “countries” covering the 1870-1996 period, with information on the article date, action-type, industry (if specified) and sub-national location (if specified) for each mention.


The database was compiled by reading through the Indexes of The Times (London) and the New York Times from 1870 to 1996 and recording each incident of labor unrest identified onto a standard data collection sheet. This was a process that required an enormous input of labor, first in the design and testing of the coding procedures, and then in the very labor-intensive process of reading through the Indexes and recording the information, and finally entering and numerically coding the data in computer files in database and time series form, and subjecting the resulting time series for individual countries to reliability studies. I often joke that if we had known what we were getting into (that is, accurately estimated the total amount of work involved), we never would have done it.


Would building another such world-historical database (on say, student unrest or peasant unrest) require another decade-long commitment?  The WLG project began in the mid-1980s, before newspaper content was digitized and available online. Does the availability of digitized sources open up possibilities for the construction of a wide array of similar databases with a fraction of the time input?  The question is worth exploring, but two cautions are already in order. The first caution is based on my experience with digitized searches. Hoping to update the WLG from 1997-2003 without having to hand search the paper indexes, I carried out a test on the digitized full text version of the New York Times using search strings designed to “catch” the same labor unrest articles that would have been caught by a human coder reading the paper Index. Significant problems in the comparability of results from the two methods arose. A computerized search of the full text both missed articles found by human readers and found articles missed by human readers. The latter were relying on the less extensive content of the Index compared to the digitized articles’ full text. (Digitized versions of the Index—as opposed to the full text of the articles—are not available.)


Of course, this would not be a problem for a database created from scratch, as long as each method produced internally consistent results over time. Nevertheless, the surprising finding was that the time involved was not substantially less for the computerized search. This is because the computerized search produced a large number of false positives that had to be eliminated by hand. (The computer had a hard time distinguishing, for example, between oil workers striking and workers striking oil). Nevertheless, with care, a sufficiently valid and reliable approach to data collection using digitized newspaper sources might be developed (albeit, most likely not one that is helpful for updating the WLG database in a comparable fashion). Bruce Podobnik (at Lewis and Clark), for one, has been constructing a database on anti-globalization movements relying on digitized newspaper reports from around the world.


The second caution is the usual data-theory caution—for one can go from dying of thirst for data to drowning in a sea of data. The massive amount of data that can be produced (in the case of the WLG, 91,947 mentions of labor unrest) can only be turned into interesting/observable patterns and compelling explanations by bringing the lenses of theory to bear on what would otherwise be an overwhelming jumble of empirical observations.

[1] Paper to be presented at the “Specialist Meeting on Globalization in the World-System: Mapping Change Over Time”, University of California, Riverside, February 7-8, 2004.

[2] Beverly J. Silver, Forces of Labor: Workers’ Movements and Globalization Since 1870, Cambridge University Press, 2003.

[3] E.H. Carr, Nationalism and After, Macmillan, London, 1945.

[4] Charles Tilly, Big Structures, Large Processes, Huge Comparisons, Russell Sage Foundation, 1984, page 146.

[5] For a methodological critique of the comparative-historical approach from a world-systems perspective, see Terence K. Hopkins, “The Study of the Capitalist World-Economy”, in T.K. Hopkins, I. Wallerstein  and Associates, World-Systems Analysis: Theory and Methodology, 1982, pages 9-38. On Galton’s problem, see Raoul Naroll, “Galton’s Problem”, in Raoul Naroll and Ronald Cohen, editors, A Handbook of Method in Cultural Anthropology, The Natural History Press, 1970, pages 974-89; and E. A. Hammel, “The Comparative Method in Anthropological Perspective”, in Comparative Studies in Society and History, 22, 2, April, pages 145-55.

[6] Tilly, op cit, page 124.

[7] Frederick Cooper, “Farewell to the Category-Producing Class?” In International Labor and Working-Class History, 57, Spring, 2000, pages 60-68 (quote from page 62).

[8] Philip McMichael, Íncorporating Comparison within A World-Historical Perspective: An Alternative Comparative Method”. American Sociological Review, 55, 1990, pages 385-97.

[9] Larry Griffin, “Temporality, Events, and Explanation in Historical Sociology: An Introduction”, in Sociological Methods and Research, 20, 4, May, 1992, pages 403-427 (quote from page 405).

[10] Jill Quadagno and Stan Knapp, “Have Historical Sociologists Forsaken Theory: Thoughts on the History/Theory Relationship”, in Sociological Methods and Research, 20, 4, May, 1992, pages 481-507 (quotes on pages 486, 502.)

[11] Terence K. Hopkins, “World-Systems Analysis: Methodological Issues,” in T.K. Hopkins, I. Wallerstein and Associates, World-Systems Analysis: Theory and Methodology, 1982,  pages 145-58 (quote from page 32); see also Arthur C. Danto, Analytical Philosophy of History, Cambridge University Press, 1965,  page 237.