Comparative Research: Beyond Linear-causal Explanation
Theoretical conversations about comparison in the social sciences have been dominated by a relatively restrictive view as to what comparisons are good for, and what kinds of comparisons are legitimate. This view takes various forms in different research contexts, and we should not oversimplify when we engage with it. Nevertheless, it can be briefly characterised as based on the premise that comparison can be avoided; that one must compare ‘like with like’; that comparison aims at causal explanation; and, one should add, that it aims at a very specific type of causal explanation.
The practice of comparison in the social and human sciences has, of course, always been more diverse. Comparison is an aspect of any kind of practice that invokes knowledge in the sense that all information implies some version of comparison to other possible realities; comparison of a formal and relatively more self-conscious kind has always been used to various ends in academic practice. However, the diverse practices of comparison have rarely found their way into textbooks on methodology – if comparisons have made it into textbooks at all.
In this context, this paper aims to free the academic practice of comparison from its theory. Or rather, the aim is to free the theory of comparison from itself so that it can better do its job of providing reflection on existing practices of thinking and research, and perhaps improve them. Recently, there have been efforts to rediscover diverse forms of comparison in different disciplines (Fox and Gingrich 2002; Dear 2005; Pickvance 2001; Brubaker 2003; Robinson 2004; Jensen 2011; Scheffer and Niewöhner 2010; McFarlane and Robinson 2012). This work provides thoughtful reflection and exemplary pieces of research, and I am building on it here.
Nevertheless, more work can be done: building on these efforts can move broadly in three directions. Firstly, we can work towards a more differentiated understanding of the ‘old’. Both textbooks and critical accounts have tended to overstate the homogeneity of traditional practices of comparison. We can benefit from examining the assumptions that have been associated with comparison in the past in more detail, and, in particular, from examining the diverse material research practices that have been involved. Secondly, and relatedly, we can develop a better sense of how diverse practices of comparison have always been in the history of different disciplines. This can provide us with exemplars of good research and help us direct our energies beyond overstated claims of newness in ongoing work. And thirdly, we can work on reflecting on (and perhaps shedding) some of the normative baggage of the debate. Traditional discourses of comparison have tended to put the weight of ‘science’, ‘reason’, and ‘progress’ on choosing the right kinds of comparisons. In some of the newer work, this baggage is sometimes simply reversed: the aim is to avoid ‘traditional’ or ‘hegemonic’ practice at all cost. In some of the rhetoric around alternative ways of doing comparison, every research project has to bear at least the weight of innovation – if not resistance or transcendence – in a way that is not realistic or helpful.
This paper discusses comparison based on attention to research practices and on their different ends. It asks what comparisons can be useful for, and in what ways. I begin with the observation that comparison has been identified with one specific aim: that of causal explanation, and in particular that of linear-causal explanation. It is commonly said that ‘one cannot compare apples and oranges’. This sentence only becomes intelligible in the context of linear-causal explanations, and I will suggest it only becomes intelligible in the context of the hold clinical trials have had on the methodological imagination of the social sciences. I will then discuss a few other ends that comparisons can be useful for, such as description, concept development, critique, and different kinds of explanation. Lastly, I will discuss a few forms of comparison that can be recognised as being useful once we broaden the ends that comparison can legitimately be used towards.
Comparison and Linear-Causal Explanation
There are many different ways in which ‘old’ understandings of comparison can be described and named. The way one labels what one is departing from of course has implications for how one frames the limitations of previous research, and for the kind of alternatives that can be imagined. Some textbooks have equated comparative research with cross-national comparisons, and some critical reflections also associate traditional comparisons with specific units of analysis such as ‘cross-national’ or ‘cross-cultural’ comparison (see Brubaker 2003). Others take a broader view: alongside the obviously polemical labelling of common assumptions concerning comparisons as ‘hegemonic’, ‘mainstream’, or ‘orthodox’, we find seemingly more theorised notions such as ‘generalising’, ‘deductive’, ‘positivist’, or ‘universalist’.
In what follows, I explore the benefits of one specific way of analysing what may unite some traditional ways of talking about comparisons. I will argue that in the theory of comparisons, comparisons have long been associated with one specific end: they were thought to serve causal explanation, and, more specifically, the kind of causal explanation associated with general linear reality (Abbott 1988; Abbott 2005). What is characteristic of this vision is not so much a deeply held positivist (or universalist) conviction – whatever that would mean – but a set of implicit assumptions which privileges a clear separation between cause and the outcome, conceived as things and a linear model of causality. In addition, it is usually thought to be important to establish how important variables are in relation to other variables.
We can characterise the tradition in terms of a specific philosophy of science; a vision of scholarly work that draws on the model of mechanical physics that has shaped representations in the traditional history of science and philosophy of science, in other disciplines, and the public.1 But rather than address the issue on the level of ideology, I am interested in work that identifies specific practices and research problems that are at the origin of different regimes of doing comparison. Looking at practices has two advantages: 1) it allows us to investigate regimes of legitimising comparisons in their diversity, and 2) it allows us to take into account the original usefulness of specific assumptions about comparisons.
Isabelle Stengers (2011) does a brilliant job of analysing the preconditions of comparability in experimental physics, but another practice which has perhaps been even more important for the social sciences is the clinical trial. The aim of clinical trials has been to establish the efficacy of different kinds of treatment. For example, Iain Chalmers quotes a seventeenth-century physician who issues the following challenge to his ‘competitors’:
If ye speak truth, Oh ye Schools, that ye can cure any kind of Fevers […] come down to the contest ye Humorists: Let us take out of the hospitals, out of the Camps, or from elsewhere, 200 or 500 poor People that have Fevers, Pleurisies, etc. Let us divide them in halves, let us cast lots, that one half of them may fall to my share and the other to yours; I will cure them without blood-letting […] but you do as ye know […] we shall see how many Funerals both of us shall have: But let the reward of the contention or wager be 300 Florens deposited, on both sides (2001: 1157).
In this vision, a comparison is like a race between two horses. Or rather, if the treatments are the horses, the comparison is the racetrack. The comparison is thus the staging ground of a competition, and the concern is for the competition to be fair. For this reason, the two groups (which are compared) need to be as similar as possible, except in regard to the treatment.
It would be interesting and important to sketch in detail the history of different research practices associated with comparison, and their impact across the social sciences. However, this is beyond the scope of this paper. I would hypothesise a process whereby specific research problems serve as ‘model systems’ (Creager, Lunbeck, and Wise 2007; Howlett and Morgan 2011) that have an impact on how research is imagined beyond their original application. In sociology (like in other disciplines) certain cases have served as privileged reference points for certain objects or areas of research, and have shaped perceptions in unacknowledged ways (Guggenheim and Krause 2012). Similar processes may be at work with regard to research methods.
The clinical trial seems to be a prime example for such a privileged reference point for social science methods: it is on the one hand a very specific practice responding to a very specific research problem – the testing of medicines – with very specific units of analysis and elements such as individuals, diseases, and medical interventions. On the other hand, it has become central to how causality, and, by implication, explanation, is understood much more broadly.
The hold of the clinical trial on the social scientific imagination has, at times, been very explicit. Glenn Firebaugh, for example, writes in The Seven Rules of Social Research: ‘[t]he trick in causal analysis is to create those conditions by comparing like with like – through investigator-produced random assignment (as in controlled experiments), through naturally occurring random assignment or through some other means of matching’ (2008: 163). Regression analyses, but also fixed-effect models and sibling models, are all attempts to simulate clinical trials for observed – as opposed to more self-consciously created – data (see Firebaugh 2008).
These types of comparison inherit the notion of causation as a form of competition among treatments. They imagine all types of causes as a form of ‘treatment’ and seek to establish their impact independently of all other factors. Regression analysis is used with individuals as units of analysis, but it is also used in quantitative forms of cross-national research in political science, and in cross-cultural comparisons such as those published in Cross-cultural Research based on the Human Area Relations File (see, for example, Deaner and Smith 2012; Martínez and Khalil 2012).
Echoes of the clinical trial are also audible in more qualitative work, if only in the form of a kind of defensiveness with regard to the trial-inspired notion of causality. More specifically, one hears echoes of the clinical trial whenever causality is conceived of as an analogy to the impact a ‘treatment’ might have, and whenever the causal effect of some phenomenon needs to be established in competition with all other factors.
These tendencies have been influential in American comparative-historical sociology, for example. This has shaped, for instance, the way the state has been conceptualised in American sociology (Skocpol 1979; 1985). In its attempts to prove the causal impact of its structures, the state had to be conceptualised as a separate ‘thing’ in a way that excluded cultural aspects; it also had to be constructed as entirely separate from economic forces and class interests. Attempts to prove the causal impact of culture in this particular way have similarly led to a definition of culture that makes it hard to understand how culture and other aspects of social practice intertwine (see, for example, Alexander and Smith 2001).
Comparison and Better Description
I have argued that there is a widespread assumption that comparisons aim to establish linear-causal relationships. This assumption has obscured other ends of social analysis in general, and of comparison in particular. To the extent that linear-causal analysis is the aim, description and category-construction are only considered as a means towards an end. In clinical trials, description of the disease itself is only a preliminary step – though other medical research may explore symptoms in more detail and the definition of a disease may of course be problematic, especially in hindsight (see Bowker and Leigh Starr 1999). In regression analysis too, description or measurement is a preliminary step, important only as a means towards establishing correlation among observed values (Marradi 1990). By subordinating description and category construction to linear-causal explanation, the concern with linear-causal explanation has proliferated trivial forms of description and category construction, while precluding some of the yields that might result were these aspects of social science research considered as ends in themselves.
For pragmatic reasons (and because only explanation is really considered valuable) research has often focused on descriptions that can be easily collected about a large number of cases, and it has focused on descriptions that can be quantified. But description is important, and this is not just about ‘getting the facts right’, as is sometimes acknowledged in the quantitative tradition. Description also situates a case with regard to sensitising concepts. Questions such as: ‘What is it like? What are its properties? How does it work? What kind of pattern can we observe? How is something changing?’ are descriptive and important, and comparisons are useful for answering them.
Some in the interpretative tradition emphasise the value of understanding the case in its particularity in opposition to the practice of comparison.2 The trope that comparison hinders understanding keeps recurring: Estrid Sorensen, for example, notes that ‘[s]pecificities are likely to become invisible through comparison’ (2008: 312). However, it has also repeatedly been pointed out from within interpretative traditions that even understanding a particular case is a comparative task (see Barnes 1973: 190; Kuper 2002: 146; see also Skocpol and Somers 1980).
All scholarly description involves selective forms of contextualisation and decontextualisation, and there are costs and benefits to different strategies. But we need to reflect on these costs and benefits in the context of concrete research projects, and we need to do so without implying two extreme positions which are misleading. Firstly, we need to avoid implying that description without comparison is possible. Any object becomes an object in relation to other objects – even an object considered holistically becomes a whole in relation to other wholes. A good repertoire of comparative cases also allows us to precisely characterise a phenomenon in its uniqueness and particularity. Secondly, we need to avoid implying that the only available form of comparison in the scholarly tradition is cross-national comparison, large-n linear-causal comparison, or cross-cultural comparison in the service of linear-causal explanation. The trade-offs do not seem to be in choosing between description and comparative practice, but between different kinds of comparative practices, and between different kinds of comparative practices maximised for different ends.
Comparison and Concept Development
Like description, concepts are a means to an end in the linear-causal tradition. In clinical trials, the unit of ‘patient’ and the categories ‘ill’ and ‘healthy’ are the starting point of the analysis: the real interest lies in exactly how to randomise, and which treatment has better results. In the social sciences, this has led to a lot of work with received categories, and for pragmatic reasons to work with categories, in which the data is already coded.
The use of concepts as categories for linear-causal analysis also has follow-on costs – in some ways regardless of the quality of the category used: once a unit of analysis or a category is ‘baked’ into a linear-causal analysis, the practices involved in its construction slip under the surface. It can thus become harder to ask questions about the historical conditions of the unit of analysis, and about the specific ways in which description was achieved.
Linear-causal analysis also casts its shadow backwards into the way non-causal use of categories and comparative description is read. That is, if linear-causal assumptions are widespread, even a comparison that is purely descriptive can invite readings that treat the unit of comparison as an explanatory variable. Cross-national comparisons in particular seem to almost automatically be read in some corners in a way that turns the category of description into a causal hypothesis. If someone studies a social movement in Finland and in Norway, it is assumed that something about these countries can explain the observed difference. This assumption shapes expectations for EU-funded projects that combine field sites in different countries (Akrich and Rabeharisoa, this volume; see also Sorensen 2008).
But comparative description using concepts as categories does not have to be read as preparing linear-causal explanation (where the category is the explanans). It can be a provocation for further research and conceptual development. It could raise questions about further variation within the unit of analysis, particularly in view of partial data collection: an ethnography, for example, has rarely covered the whole unit of analysis; it usually studies a culture through a particular village (see Kuper 2002: 143). It could also raise questions about further differentiation within the descriptions pertaining to the unit of analysis.
What I am suggesting here is that some of the problems associated with categories have to do with their specific uses in linear-causal analysis. I want to suggest that categories are useful for description, and comparison is useful for improving categories. Indeed, I would argue that developing categories is the proper aim of theoretical practice, and in that sense, comparison is essential to theoretical practice (Krause 2010).
Some kinds of comparisons are especially useful for developing concepts. Comparisons that take their departure from existing units of analysis have played an important and useful role: that of comparing common units of analysis to their opposite. Rather than just compare individuals, social theorists have compared discourses that construct individuals with those that do not, thus asking how the individual has become the dominant unit of analysis (Weber 1964 (1904); Foucault 1978, 1979). Rather than just compare nation-states, we can compare them to other social forms such as hordes, the mafia, or city states. This has enabled Saskia Sassen (2006), for example, to break the nation-state down into some of its constituent elements and ask new questions about different combinations of territory, authority, and rights. Rather than just do a comparison between cities, we can compare the city to its supposed opposite in terms of density or size, and ask broader questions about forms of settlement (Gans 2009).
Comparison and Critique
The most fundamental operation of critique is not to say ‘this is bad’, but rather to say: ‘It is not necessary. It could be otherwise’ (Calhoun 2001). Critique has been practised in different intellectual traditions and I would suggest that comparison is central to all of them. At its most modest, critiquing means to say ‘this could be looked at in another way’. Unsettling established views of the world requires comparison, at least among ways of seeing the world. Critique can also use comparison between different cases to show that any given social phenomenon is not natural or necessary.
Though comparison is often associated with universalism in contexts where that is seen as a bad thing, comparison of course has an important role to play in challenging ethnocentric, or otherwise provincial, assumptions. Foucauldian (and Nietzschean) genealogy, for example, is a comparative practice: discourse analysis is holistic and imminent, but it also depends on juxtaposition. It asks the analyst to go back in time until he or she finds assumptions and practices that are significantly different from the present (Foucault 1977, 1978; Baert 2010).
This denaturalisation can also be an effect of comparisons in more conventional social science. Max Weber’s comparison of the Protestant and Catholic work ethic, for example, can be read as an attempt to establish culture as a causal variable. But it has also shown that it is not natural for people to want to work as much as possible, or even to respond to incentives (Weber 1963 (1904)).
Critique in a stronger Marxian sense is at its best also not so much a denunciation, but a comparison. In other words, Marxian analysis is a comparison of contemporary society to a different social reality that is thought to be better and that is analytically specified. The analysis of capitalism as a mode of production based on private ownership of the means of production becomes possible only in conjunction with its imagined counterpart: a society based on the shared ownership of the means of production. This comparison allows Marxism to produce unique perspectives on the experience of work in capitalism, as well as on housing markets in capitalist cities, and on the state.
Comparison and Different Kinds of Explanations
To the extent that comparison is associated with explanation, it is generally associated with linear-causal explanation. But there are different approaches to explanation, and comparison is also useful for explanations of different kinds. Chris Pickvance (2001), for example, has argued that while comparison is often used to explain different outcomes, it is also useful for explaining similar outcomes with different causes and for functionalist explanation.
Comparison can also be used for explaining a phenomenon by listing its conditions of possibility. In this mode of investigation, a particular phenomenon is examined in comparison to other phenomena in order to specify its form.3 In dialogue with other observations, we can ask: ‘What would have to be different for this to be different from what it is?’ In this version of explanation, analytical description and explanation – and critique in the sense discussed above – fall together.
One way to make sense of scholarship on the Holocaust, for example, is to list a number of conditions that were necessary to make this particular atrocity possible. Comparing the genocide to spontaneous, short-lived outbursts of violence, we can underscore the modernity of the event and highlight the role of bureaucratic organisations (Browning 1993), national boundaries, the nature of the modern state (Bauman 1990), indirect domination (Postone 1980, 1993), profit-driven organisations (Hayes 2001), and the logic of specialised professional practice (Benedict 2003). In making comparisons with other nation-states, we can mention the German heritage of anti-Semitism (Goldhagen 1996) and the complicated relationship with violence in German forms of selfhood (Elias 1996). Comparing hypotheses about social dynamics on different scales, we can single out situational factors such as the tendency to obey authority focused on by the Milgram (1974) experiments.
In another example based on research on humanitarian-relief NGOs, I argue the form that humanitarian relief takes today can be described as a market for projects (Krause 2014). Humanitarian relief agencies provide assistance to populations in need, and they do so in the form of projects. Agencies produce projects, and they strive to produce good projects . The pursuit of the good project develops a logic of its own that shapes the allocation of resources and the kind of activities we see independently of external interests, but also relatively independently of beneficiaries’ needs and preferences. Agencies produce projects for a quasi-market in which donors are consumers. The project is a commodity, and with that, those helped become part of a commodity. The pursuit of the good project encourages agencies to focus on short-term results for selected beneficiaries. The market also puts beneficiaries in a position where they are in competition with each other to become part of a project.
We can identify this social form and what makes it possible by comparing it to other ways in which needs are produced, and in which populations are, or are not, provided with what they need. Historicising attention to distant suffering and comparing it to earlier times, we can identify means for learning about suffering and providing relief (Haskell 1985) as a condition of possibility of the market for projects. Making comparisons to responses from within the local community (which form part of existing political and religious structures), we can add the emergence of a specific form of humanitarian authority since the late nineteenth century, and especially since the 1970s. Comparisons to a situation where no one responds to distant suffering can identify Western governments’ and Western publics’ interest in relief of suffering, influence, and stability as a contributing factor. Comparisons to traditional welfare states allow us to identify competition between providers of different national origins and the fact that no organisation has a commitment to a specific population as an element of the market for projects. Making comparisons to a situation where all human communities would be able to provide for their needs themselves, we can identify different phases of primitive accumulation as contributing to the current form of humanitarian relief. For instance, the IMF-led restructuring of the 1980s and 1990s has contributed to the vulnerability of large populations, and is also a precondition for disaster relief.
Explanatory comparisons in the linear-causal tradition can tend to normalise what is common to the cases considered. If we are just asking, for example, why among ‘modern’ nations it was the Germans who organised the killing of Jews and other minorities on a large scale, the nation-state and other aspects of modernity do not come into view as factors contributing to the Holocaust.
Cross-national research can tend to focus on variation among policies that, on the world historical scale, are relatively similar. The mode of explanation presented here allows us to retain the significance of the phenomenon as a whole, and explicate it by listing and naming all conditions of possibility, including features common across capitalist nation-states, but not universal across social forms more broadly.
Naming conditions of possibility does not allow us to assess the relative impact of competing factors. The Holocaust emerges both as a particularly modern, and a particularly German, phenomenon. However, it gives us an overview of differences to other phenomena and possible analytical, as well as political, leverage points. This allows us to situate practical proposals within the full range of conditions which it might be possible to affect.
Kinds of Comparison: Like with Like/Like with Unlike or Different Types of Serial Comparison
If all things can be described in terms of both similarities and differences (and if a shared category can always be found), the distinction between comparing ‘like and like’, and comparing ‘like and unlike’, collapses. The admonition to compare ‘like with like’ does not, as such, have meaning. It follows that the term ‘comparable’ also does not, as such, have meaning – nor does the term ‘incomparable’ (be it used as a prohibition or in rebellious embrace). Rather, we need to investigate how, and in which practical context, the rule to compare ‘like with like’ has assumed meaning, and what kind of similarity and difference have been meant?
There have always been both similarities and differences involved in comparing ‘like with like’. The question under discussion has been about how many differences a given research project can handle at the same time (see Macfarlane and Robinson 2012). We have seen that in a clinical trial, the comparison is a competition and the concern is for it to be fair. The two groups included in the comparison need to be as similar as possible, except in regards to the treatment. Comparing ‘like for like’, then, means we need to hold all other sources of variation constant.
For description, concept development, and critique, it can also be useful to compare objects that are expected to be the same in some or most respects. It might allow us to better focus in on specific differences that we want to describe or conceptualise. If we are comparing men’s and women’s experiences of unemployment and we have limited resources, for example, we might want to construct sample groups of men and women that have a similar socio-economic status or cultural location. It can also be useful to compare cases with only one similarity and as many other differences as possible. If we are studying cities or professions or a certain type of organisation, we might construct a sample that contains the largest number of differences within it. One may also compare one case to as many other cases that are different in as many different ways as possible.
One important difference between comparisons that are maximised for linear-causal explanation and comparisons that are maximised for description, concept formation, or analytical explanation is that they envision a different kind of seriality in their ambitions. Linear-causal comparison aims at repeating the same comparison many times to achieve statistical significance. In other words, the comparison between treatment group and control group is designed to be the same comparison multiplied by the number of people in the group. Whenever we count the number of comparisons involved in a study by the term ‘n’, we are implying that any additional comparison does not add a qualitatively new dimension.
Other aims are better served by many separate comparisons in succession with objects that are the same, and different, in various ways. We may want to compare a nation-state to an empire, to the Mafia, to an NGO, and so on. Serial comparison of this kind can eventually erode some of the undesirable consequences of category construction. For example, with the kind of seriality that repeats the same comparison multiple times, every additional comparison naturalises the unit of comparison and obscures its condition of possibility. In contrast, with the kind of seriality that strings together many different kinds of comparisons, every additional comparison questions, but also clarifies, the unit of analysis.
The Asymmetrical Comparison
To me, the assumption that seems to most limit current practices of comparison in the social sciences is the assumption that all compared cases need to be given equal amounts of attention. In writing and assessing explicitly comparative research designs, we spend a lot of time thinking about which case(s) should be considered. Meanwhile, most qualitative designs still consider only one case, and both single-case and explicitly comparative studies spend too little time thinking about other relevant cases and literatures (or about the larger diversity in the world), which could be considered by drawing from information that is widely available, or on research that is less in-depth.
The expectation of symmetry is related to certain assumptions about the purpose of comparison. If the concern is for the comparison to be a staging ground for a fair competition, then it seems natural to treat all competitors in the same way. The assumption of symmetry also ties in with underlying assumptions about the nature of measurement or description. First, symmetry of attention appears as the default if measurement has no costs because one is working with existing data sets. Second, if measurement is unproblematically assumed to be either true or false, it makes sense to exclude data that is very poor.
Most social science research encounters more complicated questions of measurement, however, and it encounters trade-offs between different aims of the research – among them quality of information and various strategic aims of the research. Though we tend to not make them explicit, a lot of different kinds of comparisons go into conceptualisation and description – and most of them are asymmetrical. Of course, good research has to contribute new in-depth knowledge on some case or cases, but such research can be enriched with information about other cases from secondary sources (and even by pure thinking exercises). Asymmetrical comparison enhances the interpretation of findings in any study – whether that study has formally been comparative or not.
The assumption of symmetry leads to the defensiveness of single case designs that do not look left or right, and it leaves the potential insights of comparative studies as underexploited. This leads to duplication of efforts among researchers. The seriality that I outlined above potentially undermines the distinction between comparative and non-comparative designs, and between included and excluded cases.
In order to take the asymmetrical comparison to one of its conclusions, comparisons can also of course be made with hypothetical cases. Firstly, cases that are considered in relation to a specific question need not be real. Secondly, information about cases considered for comparison need not necessarily be true. Moreover, it is not always essential to know if a case is real, or if our information about it is real, if we present our analysis in measured ways as the result of a hypothetical comparison.
Counterfactual imagination has a long tradition in social science work. That is, we can ask: ‘What are the kinds of things that could be happening but are not? Did they ever happen? What would have to be different for them to be happening? How might this phenomenon be different under various analytically specified circumstances? How might this be different in the best, or in the worst, of worlds?’ Making comparisons with hypothetical cases is an element of all theoretical work. We use hypotheses produced by theories, and then test them against the data. This has also long been known in quantitative work. Regression analysis is based on comparisons of patterns in actual data to different kinds of models, based on equations.
One distinction I want to draw here is between hypothetical comparisons that aim to find a close match to reality, and those that aim to engage with the imagination in order to identify new forms of possible similarities and differences. Borrowing from Tilly (1984), we might call the latter ‘the variation-finding hypothetical comparison’.
In network analysis, for example, comparison to hypothetical cases is used to better describe patterns, and to explore the consequences of assumed, observed, and other possible patterns. Bearman and colleagues, for instance, analyse ‘reports of relationships among adolescents in one town […] and compare the structural characteristics of the observed network to simulated networks’ (2004: 44). They thus compare network structures, but the simulations also allow them to map the implications of different network-structures for the spread of sexually transmitted diseases.
This type of work stands out in the degree to which the language for capturing variation between and among different social worlds (real and simulated) is formalised. However, more qualitatively-oriented work can also draw more broadly on fiction, history, and anthropology for cases of comparison. It may be hard to be fully knowledgeable about historical and geographically distant cases. Getting the facts right is important in social science research, but so is challenging parochial taken-for-granted assumptions through comparison. According the status ‘hypothetical’ to comparisons with cases outside the researchers’ first-hand expertise may encourage the social scientific imagination of diversity. Qualitative social scientists have been reluctant to create their own fictions, though these could be created with analytical goals in mind (see Guggenheim 2009).
The Undigested Comparison
Lastly, I would like to mention the undigested comparison. A digested comparison, with either a different case or a competing analysis, is one where a comparison has been thought through in a way that fits the goals of the project: the relevant similarities and differences have been identified; the observed similarities and differences have been used to at least discover the shared or non-shared category; and the conceptual framework has been thought through in response to the comparison (see Lazarsfeld and Barton 1951). If explanation is an aim, perhaps the implications of the comparison for relevant relationships among phenomena under investigation have been thought through.
A different form of comparison is one usually introduced by ‘but see’ in footnotes. This can either point to an empirically different case, an empirical parallel, or a difference in interpretation. The ‘but see’ can indicate that the comparison has not been fully assimilated into the author’s conceptual framework; perhaps if it were fully assimilated, a major point might not result. While in general it might be good advice to try to improve academic writing by trying to fully incorporate the material from the footnotes into the main text, it is up to discussion whether any given work contains the right amount of digestion to warrant publication or sharing.
Celebrating the undigested comparison that owns up to being such also means looking at the undigested comparisons hiding in all types of other analytical operations. It is certainly preferable to have an openness towards undigested comparison, rather than practise the ‘bitchy footnote’ – if we follow Grafton (1999), this is the very origin of the footnote – where footnotes are used to dismiss competing interpretations in a way that suggests more digestion than has actually been accomplished.
Another example of an undigested comparison that is not acknowledged as such is metaphorical re-description, where one case serves as a metaphor for others without actual comparison. This is often what happens when cases become theories and are applied to other cases, instead of being compared. Foucauldian scholars, for example, can tend to apply Foucault’s account of the prison, the school, and the hospital to all other types of setting that they then identify as forms of ‘governmentality’. Perhaps no school-in-formation is exempt from this tendency. In recent years, for example, it has become popular with the growth of science and technology studies to call all kinds of practices and settings ‘laboratories’ without much attempt to consider differences, as well as similarities, between the different cases (see Guggenheim 2012).
The project to free the theory of comparison from itself has become the project to free it from the theory of clinical trials. There are, of course, epistemic and political stakes to debates about comparison in particular – and to debates about methods in general; but we should pay just as much attention to the mundane constraints of research practice.
These constraints matter as much, and they matter in ways that are not always articulated. Methods seem to draw on specific research problems as their models. Assumptions that make sense in one context get transferred from one research context to another. What makes sense as a practice for comparing treatments may bring some strange assumptions to the examination of national histories.
Reflection on the ends of comparison can have several advantages: it brings us back to specific people and specific practices. It can also remind us of the kinds of things an academic paper can and cannot achieve. That is, it suggests a sort of balancing act among different ends and limited resources, rather than a clear path with rivers full of crocodiles on both sides. The theory of comparison in the service of linear-causal explanation has clearly instilled a fear of ‘getting it wrong’, which still shapes the debate on different sides.
It can be uncomfortable to discuss the quality of research beyond ‘right’ or ‘wrong’, and beyond the content of intellectual positions. Nonetheless, it might be that on top of epistemological and practical research differences, there is a joint enemy to the different sides in the debate about comparison – which we might simply call ‘intellectual laziness’. We all encounter different kinds of laziness and our frustrations may be shaping interventions in implicit ways, and they may also produce misunderstandings. There is, on the one hand, the laziness of work that uses un-reflected units or categories of comparison (often shaped by available data in unacknowledged ways). There is, on the other hand, the laziness of works that are descriptive in a way that merely illustrates existing ‘theory’ – in a context where this sometimes just means ‘well-known arguments by authors who happened to have become famous’. There is also the laziness of comparisons that are not suited for linear-causal analysis, and do not clearly articulate how they add value on any other dimension.
I do not think we are looking for one new ‘right’ way of doing comparison, but we might want to have a conversation about desirable directions as we individually and collectively try to do the best work possible. I would suggest that every piece of information has to be, in principle, ‘accountable’ to all possible comparisons that could be made. We can use comparisons to constantly question and improve our categories of analysis and description. With the kind of serial comparison I am proposing, this can include asymmetrical and hypothetical comparisons, while considering one or more cases through original research.
We can, of course, not make the demand that every piece of information should be accountable to any comparison to any given talk or paper. All papers, talks, and books are preliminary results and legitimately works in progress. Some contributions add a lot to the conversation among scholars in other ways by showing material connections between cases: for example, those that can form the basis of constructing new units for comparative analysis. However, we can make this demand of intellectual ambitions and of shared standards of conversation. ‘Accountable’ here means ‘in principle answerable to’. The answer to ‘what do you make of a possible comparison to x’, should never be ‘I did not include x in my research’. It has to be, at least: ‘that might be interesting because…’
We have historically disproportionally demanded this type of accountability of ‘non-typical’ cases, which has meant in the current environment that non-US and non-UK cases have had to be justified in English-speaking journals, whereas UK and US cases have at times been analysed as though their interest was self-evident and as though other countries and research on other countries do not exist (see Stöckelová, this volume). Nonetheless, I think the appropriate answer is not to refuse to engage in comparisons, but to insist that everyone makes the particularities of their assumptions and cases explicit as much as possible.
This paper benefitted from comments received by Joe Deville, Jennifer Robinson, and the editors of this volume. I am also grateful to the participants of two workshops on comparisons at Goldsmiths College: one organised by the editors, and an earlier workshop organised by the editors with Rebecca Cassidy, Claire Loussouarn, and Andrea Pisac.
1 See Mayr (1982) for a provincialisation from a biologist’s perspective.
2 See Steinmetz (2004) for an extended discussion on the origins of these arguments.
3 See also Ragin’s (1992) notion of casing (see also Ragin and Rubinson 2009).
Abbott, A., ‘Transcending General Linear Reality’, Sociological Theory, 6 (1988), 169–186
——‘The Idea of Outcome in US Sociology’, in G. Steinmetz, ed., The Politics of Method in the Human Sciences (Chapel Hill, NC: Duke University Press, 2005), pp. 393–426
Akrich, M., and V. Rabeharisoa, ‘Pulling Oneself Out of the Traps of Comparison: An Auto-ethnography of a European Project’, this volume
Alexander, J., and P. Smith, ‘The Strong Program in Cultural Theory: Elements of a Structural Hermeneutics’, in J. Turner, ed., Handbook of Sociological Theory (New York: Springer 2001), pp.135–150
Baert, P., Social Theory in the Twentieth Century (Cambridge: Polity Press, 2010)
Barnes, B., ‘The Comparison of Belief-Systems: Anomaly Versus Falsehood’, in R. Horton, and R. Finnegan, eds., Modes of Thought: Essays on Thinking in Western and Non-Western Societies (London: Faber, 1973), pp. 182–198
Bauman, Z., Modernity and Ambivalence (Cambridge: Polity, 1990)
Bearman, P. S., J. Moody, and K. Stovel, ‘Chains of Affection: the Structure of Adolescent Romantic and Sexual Networks’, American Journal of Sociology, 110.1 (2004), 44–91
Benedict, S., ‘Caring while Killing: Nursing in the “Euthanasia” Centers’, in E. R. Baer, and M. Goldenberg, eds., Experience and Expression: Women, the Nazis, and the Holocaust (Detroit: Wayne State University Press, 2003), pp. 95–110
Bowker, G. C., and S. Leigh-Starr, Sorting Things Out (Cambridge, MA: MIT Press, 1999)
Browning, C. R., Ordinary Men: Reserve Police Battalion 101 and the Final Solution in Poland (New York: Harper Perennial, 1993)
Brubaker, R., ‘Beyond Comparativism: Theory and Research in Comparative Social Analysis’, EScholarship UCLA, (2003) <http://www.escholarship.org/uc/item/7t52j73w> [accessed 4 February 2014]
Calhoun, C., ‘Critical Theory’, in G. Ritzer, and B. Smart, eds., Handbook of Sociological Theory (London: Sage, 2001), pp. 179–200
Chalmers, I., ‘Comparing Like with Like: Some Historical Milestones in the Evolution of Methods to Create Unbiased Comparison Groups in Therapeutic Experiments’, International Journal of Epidemiology, 30.5 (2001), 1156–1164
Creager, A. N., H. E. Lunbeck, and M. Norton Wise, eds., Science without Laws: Model Systems, Cases, Exemplary Narratives (Durham, NC: Duke University Press Books, 2007)
Deaner, R., and B. Smith, ‘Sex Differences in Sports across Fifty Societies’, Cross-Cultural Research, 47.3 (2013), 268–309
Dear, M., ‘Comparative Urbanism’, Urban Geography, 26.3 (2005), 247–251
Elias, N., The Germans: Power Struggles and the Development of Habitus in the Nineteenth and Twentieth Centuries (Cambridge: Polity Press, 1996)
Firebaugh, G., The Seven Rules of Social Research (Princeton, NJ: Princeton University Press, 2008)
Foucault, M., Discipline and Punish: The Birth of the Prison (New York: Pantheon Books, 1977)
——The History of Sexuality (New York: Pantheon Books, 1978)
Fox, R. G., and A. Gingrich, eds., Anthropology, by Comparison (London: Routledge, 2002)
Gans, H. J., ‘Some Problems of and Futures for Urban Sociology: Toward a Sociology of Settlements’, City & Community, 8.3 (2009), 211–219
Goldhagen, D. J., Hitler’s Willing Executioners. Ordinary Germans and the Holocaust (New York: Knopf, 1996)
Grafton, A., The Footnote. A Curious History (Cambridge, MA: Harvard University Press, 1999)
Guggenheim, M., ‘The Science Fiction of Science Studies’, unpublished manuscript, 2009
Guggenheim, M., and M. Krause, ‘How Facts Travel: The Model Systems of Sociology’, Poetics, 40 (2012), 101–117
——‘Laboratorizing and Delaboratizing the World: Changing Sociological Concepts for Places of Knowledge-production’, History of the Human Sciences, 25.1 (2012), 99–118
Haskell, T., ‘Capitalism and the Origins of the Humanitarian Sensibility, Parts 1 and 2’, The American Historical Review, 90.2/3 (1985), 339–61
Hayes, P., Industry and Ideology: IG Farben in the Nazi Era (Cambridge: Cambridge University Press, 2001)
Howlett, P., and M. S. Morgan, eds., How Well do Facts Travel?:The Dissemination of Reliable Knowledge (Cambridge: Cambridge University Press, 2011)
Jensen, C. B., ed., ‘Comparative Relativism: Symposium on an Impossibility’, Common Knowledge, 17.1 (2001), 547–566
Krause, M., ‘Theory as the Practice of Hunting Variables’, paper presented at the 105th Meeting of the American Sociological Association, Atlanta, USA, 2010
——The Good Project: Humanitarian Relief NGOs and the Fragmentation of Reason (Chicago: Chicago University Press, 2014)
Kuper, A., ‘Comparison and Contextualization’, in R. G. Fox and A. Gingrich, eds., Anthropology by Comparison (London: Routledge, 2002), pp. 143–176
Lazarsfeld, P., and A. Barton, ‘Qualitative Measurement in the Social Sciences: Classification, Typologies, and Indices’, in D. Lerner, and H. D. Lasswell, eds., The Policy Sciences (Stanford: Stanford University Press, 1951), pp.155–192
Marradi, A., ‘Classification, Typology, Taxonomy’, Quality & Quantity, 24.2 (1990), 129–158
Martínez, P. R., and H. Khalil, ‘Battery and Development: Exploring the Link between Intimate Partner Violence and Modernization’, Cross-Cultural Research, 47.3 (2012), 231–267
Mayr, E., The Growth of Biological Thought (Cambridge, MA: Belknap Press, 1982)
McFarlane, C., and J. Robinson, ‘Introduction: Experiments in Comparative Urbanism’, Urban Geography, 33.6 (2012a), 765–773
McFarlane, C., and J. Robinson, eds., ‘Experiments in Comparative Urbanism’, Urban Geography, 33.6 (2012b)
Milgram, S., Obedience to Authority: An Experimental View (New York: Taylor & Francis, 1974)
Mill, J. S., System of Logic: Ratiocinative and Inductive: Being a Connected View of the Principles of Evidence and the Methods of Scientific Investigation (London: Longmans Green, 1898)
Pickvance, C. G., ‘Four Varieties of Comparative Analysis’, Journal of Housing and the Built Environment, 16.1 (2001), 7–28
Postone, M., Time, Labor, and Social Domination: A Reinterpretation of Marx’s Critical Theory (Cambridge: Cambridge University Press, 1993)
——‘Anti-Semitism and National Socialism: Notes on the German Reaction to "Holocaust"’, New German Critique, (1980), 97–115
Ragin, C. C., ‘“Casing”’ and the Process of Social Inquiry’, in C. C. Ragin, and H. S. Becker, eds., ‘What is a Case? Exploring the Foundations of Social Inquiry’ (Cambridge: Cambridge University Press 1992), pp. 217–226
Ragin, C. C., and C. Rubinson, ‘The Distinctiveness of Comparative Research’, in T. Landman, and N. Robinson, eds., The Sage Handbook of Comparative Politics (London: Sage, 2009), pp. 13–34
Robinson, J., ‘In the Tracks of Comparative Urbanism: Difference, Urban Modernity and the Primitive’, Urban Geography, 25.8 (2004), 709–723
Sassen, S., Territory, Authority, Rights: From Medieval to Global Assemblages (New Jersey: Princeton University Press, 2006)
Scheffer, T., and J. Niewöhner, eds., Thick Comparison: Reviving the Ethnographic Aspiration, vol 114 (Leiden: International Studies in Sociology and Social Anthropology, 2010)
Sewell, W. H. Jr., ‘Ideologies and Social Revolutions: Reflections on the French Case’, Journal of Modern History, 57.1 (1985), 57–85
Skocpol, T., ‘Cultural Idioms and Political Ideologies in the Revolutionary Reconstruction of State Power: A Rejoinder to Sewell’, Journal of Modern History, 57.1 (1985), 86–96
——States and Social Revolutions: A Comparative Analysis of France, Russia, and China (Cambridge: Cambridge University Press, 1979)
Skocpol, T., and M. Somers, ‘The Uses of Comparative History in Macrosocial Inquiry’, Comparative Studies in Society and History, 22.2 (1980), 174–197
Sorensen, E., ‘Multi-sited Comparison of Doing Regulation’, Comparative Sociology, 7.3 (2008), 311–337
Steinmetz, G., ‘Odious Comparison: Incommensurability, the Case Study and Small Ns in Sociology’, Sociological Theory, 22.3 (2004), 371–400
Stengers, I., ‘Comparison as a Matter of Concern’, Common Knowledge, 17.1 (2001), 48–63
Stöckelová, T., ‘Frame against the Grain: Asymmetries, Interference and the Politics of EU Comparison’, this volume
Tilly, C., Big Structures, Large Processes, Huge Comparisons (New York: Russell Sage Foundation Publications, 1984)
Weber, M., The Protestant Ethic and the Spirit of Capitalism (London: Unwin 1968 (1904))