Counterfactual and consilience

Introduction

At the 2006 Summer Institute in Economic Geography Workshop, Jamie Peck invoked the donut metaphor to describe the field’s open and fragmentary pluralism; a pluralism that seems to trend toward polyphonic disunity. The hole metaphor implies a field without a center, a composition of disparate explanatory parts that pull in different directions, and that could benefit from some degree of integration, if not unification, of those parts.

One response to Peck’s donut hole has been engaged pluralism which Barnes and Sheppard (2010) suggest is reasonably exemplified in the GIS model of engagement. Engaged pluralism has usefully begun the conversation for addressing the hole. Nonetheless, the lure of a strong center that shrinks the space of explanations has also been strong. Goodchild (2004) favored positioning GIS as science, more specifically, as GIScience, to unify the field. Reitsma (2013: 218) argues that GIScience’s focus on developing unified descriptions of space and time parallels science’s unification goal by aiming for a “common language.” A center that focuses on unified explanations is likely to be rejected by economic geographers who remain agnostic to transcendent explanations achieved through a common scientific language. Yet, unificatory voices, motivated by goals of gaining wider academic legitimacy, have been persistent (Yeung, 2024). A unificatory goal need not eschew explanatory autonomy. ¹ It acknowledges a variety of unificatory possibilities, from minimalist unification that recognizes mutual illumination of different explanations, engaged pluralism, to one that works toward integration and synthesis of explanations. I explore the above in the context of consilience and counterfactual explanations, and in dialogue with two major approaches in the field (evolutionary economic geography (EEG) and global production networks (GPN)).

In the broader academic community, unificatory tendencies have been predominantly expressed in two ways. First, they assume the form of consilience explanation. Consilience puts forward the idea that radically different epistemologies and methodologies that underscore disciplinary fragmentation may be integrated into a unity of knowledge (Scheff, 2015). In this sense, consilience explanations would favor that the hole in economic geography be plugged through integration of different approaches. Consilience advocates for interdisciplinary explanation through a model of vertical integration despite radically different approaches and methodologies. For instance, the integration of biological concepts in the humanities and social sciences (Slingerland, 2008). In this vertical integration, the humanities and social sciences are higher up the hierarchy but synthesize explanations and concepts drawn from the sciences at the base. Consilience is not without controversy. Wilson’s (1998) seminal work envisioning exchanges between the humanities and sciences implies subsumption of the humanities under science through his desire to secure certainty of the human condition; one that is suggestive of a strong center inhabited by science. Despite a strong integrationist bias, his view has been endorsed in some literary quarters, for example, Carroll’s (2011) claim that serious literary work needs to be reductionist. ²

Second, within economic geography, causal explanation has been touted as the superior approach to explanation (Martin and Sunley, 2003, 2015; Yeung, 2024). One reason lies in its explanatory power in addressing adequacy. Causal explanation as adequate explanation of facts and phenomena compared to rival explanations may be traced to Kitcher’s (1981, 1989) seminal essays advancing explanatory adequacy as the principal lens to evaluate unification. For Kitcher, causal explanation suffices as a unified explanation. Clearly, causal explanation is not the only approach to explanatory unification as others have shown, yet its potential as a unification theory continues to attract (Sansom and Shields, 2018). In economic geography, rumblings of inadequate attention to causal explanation have surfaced persistently. An example being Martin and Sunley’s (2003) complaint of Michael Porter’s lack of causal explanation in his cluster concept. Here, a theory or explanation is thought to be inadequate and chaotic, if it does not identify and empirically validate causal mechanisms (see also Martin and Sunley, 2015; Yeung, 2024).

In economics, consilience explanation has seen the rise of narrative-based explanations of causality. Shiller (2019) encourages economists to integrate narrative approaches from other social sciences and the humanities. Best known for his work on the stock market, he correlates “irrational exuberance” in the stock market to the power of emotion from narratives. Specifically, he argues that individuals’ actions and intentionality are best understood in terms of the stories that they tell and connect with. Notwithstanding his call for consilience, Shiller has not abandoned neoclassical economics’ proclivity toward a methodologically individualistic (MI) mode of explanation where socioeconomic outcomes are understood in the context of downward causation to the individual. Under MI, causation can only be adequately explained by individual behavior. ³ For Shiller, emotion (e.g. fear) gives meaning to individuals and their motivation as opposed to individual rationality that dominates neoclassical economics. Despite his repudiation of the rational individual, Shiller’s approach to consilience does not stray too far from the discipline’s MI understanding of causation which ascribes causal power to the individual. ⁴ Emotions may be accessed through storytelling that relies on texts than the “tyranny of metrics” (p. 75). Moreover, deductivism that favors econometric-derived causality highlighting exogeneity has not escaped his narrative methodology. Personal narratives, for example, are said to be exogenous and could generate shocks to the market. Exogeneity is central in econometric causal modeling by minimizing problems of endogeneity and reverse causality. Nonetheless, vertical integration does not seem to be the goal in his reading of consilience. A significant degree of autonomy is retained in both the MI and exogenous causation approach that resonate in significant quarters of the discipline. Shiller’s proposed consilience explanation does not purport to emulate literary Darwinism where “terms of art used in biology continue to hold further up the pyramid of explanation” (Kramnick, 2012: 434). Rather, it is a model of collaboration that draws on epidemiological models and humanities’ text analysis, but primacy of the economics discipline is retained. Cognitive structures between economics and epidemiology for example remain distinct. There is little or limited growth of either disciplines toward each other.

Exogenous thinking has influenced econometric analysis in economic geography. It is popular among quantitative economic geographers (e.g. Lee and Rodríguez-Pose, 2021; Rodríguez-Pose and Muštra, 2022). Referees in this journal routinely seek proof of exogeneity in econometric approaches to causation. Often this involves introducing instrumental variables and two-stage least squares regressions (see e.g. Yang and Bathelt, 2023). There are other ways to approach causality quantitatively, for example using certain mathematics to examine emergent causality (Bergmann et al., 2009), and incorporating adaptation in agent-based modeling under complexity theory (Byrne and Callaghan, 2023). ⁵ But these are not common methodologies judging by the handful of publications in the field’s major journals. The larger point is that a unificatory approach cannot be centered on an econometric approach to causality. This would deny possibilities of explanatory autonomy. A benefit of pluralism is that tension serves as a check to the field’s honesty by giving a space to those who press for unification and those who press for the particular. Peck (2023) aptly describes the tension in terms of lumpers and splitters.

Rather than an econometric mode of causal reasoning, an argument may be made for causal inference through counterfactual reasoning. Counterfactuals are counter claims to what actually happened. If a certain X initial condition is found to causally lead to Y outcome, a counterfactual attempts to answer the “what if” question. For example, would Taiwan have become a chip center if its entrepreneurs had gone to the Dominican Republic instead of the United States to study and work? ⁶ While such a question is difficult to verify empirically since the what-if question is speculative, even fictitious, knowing what has not happened, or what might happen, can provide insights. It potentially helps clear away distorting variables. Qualitative researchers tend to consider counterfactuals as necessary conditions relying on methods such as process tracing to infer causality. For Goetz and Mahoney (2012), counterfactuals come more naturally to qualitative researchers because of their predisposition to the single case and the particular. Processing tracing, a within-case qualitative method of causal inference, compares cases with different constellations of causal factors to determine necessary causes (Harding and Seefeldt, 2013). But counterfactuals are also implicit in econometric analysis that presumes a conjunction of causation such that X is necessary for Y. This implies that if X did not occur, but Z had occurred instead (counterfactual reasoning), then Y is unlikely to occur. In this sense, counterfactual causal reasoning may be applied to qualitative and quantitative research, and potentially offers a bridge for mutual illumination through proximate language. In the rest of the paper, I explore possibilities of consilience and counterfactuals in the context of two popular explanations in the field, namely EEG and GPNs.

Coe and Yeung (2015) distinguish GPN2 from GPN1 by paying greater attention to causal factors such as market dynamics and firm capability to explain network configurations. Each factor is richly conceptualized and does not lend itself easily to quantification and econometric causation. Moreover, the ultimate goal is to understand distal causes of network configurations that extend to individual agency and behavior, specifically the entrepreneur’s intentional strategy. In the counterfactual hypothesis, plausible alternatives are important. The counterfactual of the Dominican in the success of Taiwan’s semiconductors presumes a causal mechanism, namely learning and knowledge production especially in the Silicon Valley. However, the analysis elongates ⁷ in an agentic context and also presumes a significant role for entrepreneurs like Acer’s founder, Stanley Shih, who was central in promoting learning and knowledge production in Taiwan’s Hsinchu Science Park. GPN2 insists on the purposive and intentional agent in driving firm strategic action. In this case, the counterfactual could ask: had Mr. Shih not developed firm capability in Acer and encouraged spin-offs, would the Science Park have developed the inter-firm networks for learning?

According to Zahle and Kincaid (2019), MI scholars argue that an explanation that elongates by drawing on human agency potentially offers a more satisfactory explanation because it includes distal causes. Others have gone further and insisted that subjectivist methodologies including participant observation and unstructured interviews resonate with the MI tradition since they focus on human agency including human intentionality (Williams, 2000). The point of agency here is that the individual agent lends itself to plurification beyond the rational human as Shiller acknowledges of the emotional agent. Not surprising, feminist economic geographers reject the rational strategic agent as adequate explanation, championing ethnography and participatory research to address perceived inadequate account of agency in GPN (Werner, 2016). Furthermore, Werner’s Dominican counterfactual raises questions of adequacy and robustness, a counter example that may prompt new or revision of explanatory vocabularies. But it can also be seen as compositive in that explanation is not closed to the diversity of mechanisms. Elongation that intersects with the counterfactual compositional, in this sense, implies that GPN network configuration may be articulated through a variegated prism of causation.

To the extent that EEG draws on Darwin principles of selection, inheritance and variety, it manifests all signs of consilience. Organizational routines are thought to be a repository of knowledge and information, comparable to genes in biology. Routines insure the inheritance of information, and they shape the firm’s capability and ability to weather competitive selection. Regional capability in turn is identified as a causal explanation of regional diversification (Boschma and Frenken, 2009). It is possible that consilience has contributed to EEG’s popularity as Rosenman et al.’s (2020) mapping of major themes in economic geography has shown. However, it is EEG’s propensity to intra-disciplinary consilience that is of interest here by exploring potential growth toward another explanation, institutional economic geography (MacKinnon et al., 2009) without the sort of vertical integration that sacrifices explanatory autonomy. Growth and elongation stop short of horizontal integration by acknowledging the localization of similarities but also differences. For instance, Boschma and Frenken’s (2009) recognition that institutions could help explain the rise of emerging industries but also their insistence that EEG’s primary unit of analysis lies in the firm’s organizational routine, not the institution.

Similar effort has been directed at intra-disciplinary consilience between EEG and GPN (Boschma, 2022; Mackinnon, 2012; Yeung, 2021). Both Boschma and Yeung have made attempts at explanatory growth and elongation between the two explanations despite different methodological preferences. ⁸ This may be seen in Yeung’s effort to link strategic coupling to regional diversification, and Boschma’s recognition that learning and knowledge spillovers from GPN networks may explain regional capability in EEG. These efforts share narrative economics’ limited growth into different explanations, one that is shaped by theoretical and empirical possibilities than equivalence. Finally, while EEG also pays some attention to related variety as a causal factor, Bathelt and Storper (2023) suggest that causality assigned to related variety remains uncertain. Their critique provides a good illustration of the counterfactual approach; specifically, their highlight of successful instances of regional development even when related variety is low. This raises questions of related variety as a sufficient condition for the evolution of regional pathways. Thinking through counterfactual is useful for developing more robust causal relations.

Conclusion

In this exchange, I suggest some unificatory possibilities through intra-disciplinary consilience or/and counterfactual causation. This does not deny the virtue of plurality of voices in polyphonic disunity but serves as a perch to discover points of encounter to see where they may lead. Both strategies do not presume that it is possible to eradicate explanatory commotion. The goal is for limited unification and modest collaboration that preserves multivoiced explanations. This implies an openness to opportunities of crossing the hole through comparative methods of counterfactuals, and intra-disciplinary consilience that potentially draw cognate explanations and their methodologies into proximity. In this sense, counterfactuals seem relatively promising for quantitative and qualitative research that carries unificatory causal aspirations. In the case of intra-disciplinary consilience, the locus of limited integration is not for a more complete understanding that grounds inquiry, but to articulate a salient mode of interaction that may profit from horizontal traffic between potentially interdependent voices.

Finally, Barnes and Sheppard (2010) rightly note the challenges of intensifying horizontal traffic between clusters of localized networks. Both counterfactual-in-action and consilience-in-action require a modest reconfiguration of network dynamics, particularly the reshaping of social collaborations to foster more inter-cluster voices. Social practice is likely to be slow, even uncomfortable, with individuals spending some time from their conceptual and methodological spaces. It requires tolerance of difference that allows the surfacing of creative conflicts. Yet such conflicts may prompt new network formations that potentially identify and resolve counterfactuals or that encourage intra-disciplinary consilience.