In defence of sociological description: A ‘world-making’ perspective

I am pleased to contribute to the long-standing debate about the relationship between descriptive and causal strategies in sociology. This familiar question goes to the heart of understanding the purpose of social science itself and forces us to think through, at a fundamental level, what we are trying to achieve. My aim here is not criticise causal analysis as such, which undoubtedly has a vital role to play, but to defend descriptive sociology for two linked reasons. Firstly, strategically, in the early 21^st century, descriptive social science has great public as well as academic resonance. If we exclude descriptive social science from our baggage, we lose vital, critical, contributions to contemporary debate. Secondly, this capacity of descriptive social science comes from its capacity to be ‘world-making’—to open up vistas of wonder, concern, empathy and horror which are vital for renewing the sociological imagination—and for engaging wider publics. Descriptive assemblages open up new worlds to academic and non-academic audiences, shatter older assumptions shattered and disclose new possibilities. Causal analysis, by contrast, is forced to manipulate different pre-defined conditions in order to infer relative causal relations and lacks this world making capacity.

My unease with the mobilisation of ‘causality’ as superior to ‘description’, is in some ways a gut feeling, tied to Pierre Bourdieu's (2000) critique of the ‘scholastic point of view’. One of my worries when social scientists invoke the primacy of ‘causality’ is that research becomes locked—mostly inadvertently—into an academic politics of closure, in which group of experts winnow better (causal) from worse (descriptive) ways of addressing any given topic. The term ‘descriptive’ is routinely deployed as secondary to the prized ‘causal’, and being able to adjudicate these boundaries ultimately becomes bound up with claims to scholarly excellence—whether this is staged through statistical sophistication, theoretical acumen, political proclivities, or some other way. In this game of academic closure, those who can claim to conduct ‘causal’ analysis become better able to command the high ground of the ‘scholastic point of view’ itself. But following not only Bourdieu but a host of writers who insist on the need to position ourselves from the subaltern point of view, we cannot take this claim at face value—it needs to be exposed as a strategy of empowerment.

This line of argument means that I do not need to address directly the philosophy of social science, where the analysis of causation has a huge and venerable literature which I can't do justice to here. In fact, for what it is worth, I have always been inspired by critical realism, which to my mind offers a convincing defence of the value of establishing causal relations in a deep and rigorous way. Therefore, I have no interest in challenging causal analysis as such. Rather, my reflections are rooted in the specific challenges facing 21^st century sociology, and the vital need in these dark and gloomy days to renew the sociological imagination which has lost ground amidst the welter of data generating and manipulating agencies. These digital infrastructures routinely mobilise descriptive assemblages through mundane scanning, surveying, and profiling activities. Like it or not, ‘the mobilisation of description’ has become a key turf on which disputes about knowledge, expertise and legitimacy are now fought out. We need to champion rich and powerful ways of doing descriptive work so that our research both disrupts these modes of thinking, in the name of offering more productive perspectives. If we don't do this, then the field is left open for other parties—commercial, governmental, and corporate—to colonise this informational terrain, leaving sociologists increasingly talking to each other in academic silos. We need to be inspired by the ‘golden age of sociology’ from the later 1950s to the early 1970s, when sociological research commanded public attention for its revelatory power (see Savage, 2010), and seek to position our thinking on this crucial public and academic stage.

I have written a considerable amount over the years on what I have called the ‘descriptive turn’ (Savage, 2009, 2020, 2021) and I don't want to rehearse my previous reflections here. Suffice to say, I was inspired by Andrew Abbott's (1988, 1990) brilliant, provocative critique of linear statistical models and by his championing of ‘descriptive’ methods, notably sequence analysis, but this can be extended this to include social network analysis, multiple correspondence analysis and indeed even standard multivariate methods such as regression which can be seen to be establishing relationships. Abbott insisted that social scientists should abandon their insular self-absorption into their own internal protocols and recognise that natural scientists and commercial interests were making huge inroads by taking up descriptive projects, for instance through gene sequencing which was revolutionising biological sciences. In a similar way—like it or not—market research uses profiling methods to claim authority over studies of consumer behaviour. Much of this kind of descriptive profiling is politically fraught, being implicated in myriad forms of surveillance and control, notably around the policing of borders (Amoore, 2006). But there is no way of dodging these political minefields. We can't just bury our heads in the sand and pretend these powerful descriptive forms of knowledge don't happen, and that we can just carry on with our preferred approaches (which we can label ‘causal’, ‘analytical’ or whatever). This will just leave the path clear for other less scrupulous players to pile in. We need to get out of our comfort zones if we are to fully redeem the critical power of sociology, speak to the ‘powers that be’ and command wider public interest.

Let me be clear, my call is therefore not in opposition to causal analysis as such, much of which depends on analysing rich descriptive material (see e.g. Goldthorpe, 2001). But privileging causal analysis itself as some kind of ‘holy grail' can distract attention from the turbulent world where data is mobilised, manipulated, and assembled by all sorts of powerful agents. Sociology needs to be adept in working across numerous sources of data and information. Mobilising effective and appealing descriptions are crucial if we are to communicate effectively in this turbulent environment. It is central to championing the ‘world-making’ qualities of sociology—to allow us to see things differently, and with hope.

A lot of my interest in descriptive methods came from an aesthetic pleasure in the wonderful visualisations that they can produce. They offered a remedy to what Guggenheim (2015) discusses as their marginalisation within sociology. The austere ‘high road’ of causal analysis has normally been trodden through statistical brilliance or theoretical finesse elaborated through textual narrative strategies. But visualisations open up such beautiful terrain! I still remember my sheer pleasure at the colourful visuals produced by Modesto Gayo-Cal and Gindo Tampubolon when we first used multiple correspondence analysis to map the organisation of cultural capital in Britain nearly 20 years ago (Savage et al., 2005 and see the more mature use of these methods in Bennett et al., 2009). More broadly, much good descriptive work in social science deploys visuals to evoke an aesthetic response, what Susan Halford and Savage (2017) characterise as ‘symphonic’ quality. This visual aesthetic allows a way of communicating across boundaries which can mobilise wider audiences and championing powerful and compelling narratives—and counter-narratives. They convey social science as ‘world making’, where new arrays of data and analysis open up vistas which have been obscured in previous research and thus disclose horizons of possibility. In Deleuze's terms, they follow the ‘plane of immanence’, the ground of ‘becoming’ itself, revealing new domains of possibility.

Let me give two examples which make this point very clearly.

The economist Thomas Piketty, and numerous collaborators has made the analysis of granular economic distributions, breaking these down in minute detail, central to a renewed interest in the study of inequality. Using detailed taxation data, this project first came to attention with studies of the increasingly warped American income distribution in which the top 1% of earners could be shown to be taking a higher proportion of American national income over several decades (Piketty & Saez, 2003). Later work used these methods to extrapolate across an increasing number of nations (Piketty, 2014, 2020) to demonstrate how inequality was in general a rising trend, both with respect to income but also wealth.

This—descriptive!—work had huge take up. The use of percentile breakdowns of income distributions became a key motto for Occupy Wall Street, with its famous call that ‘We are the 99%’. Piketty's Capital in the 21^st Century has been the best-selling social science research monograph of the 21^st century and spawned extensive discussions about the broader discussions of extreme inequality. The—descriptive!—project of modifying national accounts so that they do not just fixate on growth rates, but include metrics for inequality, is of huge political importance in seeking to force governments, businesses and the wider public to report data on inequality as a public responsibility (Piketty et al., 2018). Part of the appeal lay in the adept visualisations it produced. In contrast to the austere abstraction of the gini-coefficient, the mainstream tool of economics which reduced inequality to a singular number between 0 and 1, percentile income breakdowns could be arrayed graphically with deft sparklines. The World Inequality Laboratory's website allows non-experts to easily obtain descriptive information on inequality trends across most of the globe, using a variety of metrics, rendered in attractive visualisations, which can be downloaded on an open access basis. I include an example below (see Figure 1), which demonstrates, for instance, that South Africa has seen dramatic income inequality shifts in recent years, whereas France has remained relatively stable.

Why has this descriptive work been so effective? It certainly cannot be that it was yoked to an effective causal analysis. Here Piketty's various proposals have had a very mixed reception. In Capital and the 21^st Century he proposed that r > g, that the net rate of return to capital exceeds the growth rate was the ‘central contradiction of capitalism’, since the implication was that economic growth will intensify, and not modulate, economic inequality, leading ultimately to unsustainable inequality. This is a lovely, parsimonious, explanatory model. However it has met with near universal critique, on the basis that it reworks an untenable form of neo-classical theory (Soskice, 2014), and also that it does not recognise that inequality trends are much more variable than can be captured by this deterministic theory.

It is striking that in his more recent Capital and Ideology, Piketty (2020) rows back from the apparent determinism of r > g by emphasising that politics makes a difference and that there are conjunctural factors which can shift inequality, for better or worse. This opens the door to a political institutionalist interpretation where inequality trends are not the product of an underlying ‘contradiction of capitalism’ but depend on the effectiveness of differing kinds of political mobilisation (see the discussion in Savage & Waitkus, 2021). But in an even more recent book, Brief History of Equality Piketty (2021) pivots again, arguing that the long-term historical trend is towards greater equality—on the face of it, a complete volte face from Capital in the 21^st Century. His theory now seems to be a version of sociological theories of reflexivity. To make ‘real progress….. requires us to accept deliberation, the confrontation of different points of view, compromises, and experimentation’ (p 11–12). Although he does not refer to Durkheim, Weber, Giddens or Beck, his emphasis on ‘learning and collective engagement’ (p13) has some remarkable parallels to sociological theories of modernity. He now downplays political conflict (which he had highlighted in Capital and Ideology).

In short, Piketty's causal analysis is a complete mess. His arguments have veered inconsistently in less than a decade from a version of economic determinism to a political institutionalism, and now to a half-baked sociological theory of evolutionary reflexivity.

What makes this descriptive body of work so powerful is therefore not its success in conveying a clear causal analysis, but rather its capacity for ‘world making’—offering new visions and perspectives that can be revelatory. By bringing into view very small social groups—privileged elites comprising only 1%, or at times only 0.1%, 0.01% or even 0.001% of the population, this scholarship disclosed a different kind of world, one which drew the veil from the inordinate wealth of a tiny number of people. Previously dominant social scientific framings, geared to the central tendencies of the distribution, (as articulated in the gini coefficient, for instance), were jolted from their previous hegemonic place and alternative universes were revealed.

The lesson is clear. In the absence of convincing causal analysis, should we throw the remarkable data assemblages of this kind into the bin? Thereby eradicating one of the most powerful political mobilisations of social science we have seen in the 21^st century? I doubt that many social scientists—however committed they might be to the principles of causal analysis—would go this far.

Let me take a second example, Wilkinson and Pickett's The Spirit Level. This was published in 2009 and became famous for its claim that unequal societies were also characterised by more systemic social problems. This argument was buttressed by correlations visually arrayed so that it could readily be seen that those nations with the highest income inequality—such as the US—also scored worse on numerous indicators of lifechances and well-being. By contrast, nations which had lower income inequality—such as Japan, as well as a variety of Scandinavian nations—scored much better. It was this descriptive aspect of The Spirit Level which commanded interest, because it suggested links between a wide range of variables that had not previously been widely considered together.

However, following the banal mantra that ‘correlation does not entail causation’, it is clear that The Spirit Level was not able to establish the specific causal link mechanisms at work. The authors clearly had causal claims in mind, and were attracted to seeing psychological mechanisms relating to shame and stigma as playing an important part in driving these associations (see Pickett & Wilkinson, 2015). Yet, while their interesting, recent work, The Inner Level offers some interesting insights along these themes, it hardly establishes their causal basis.

Given this causal uncertainty, why did the Spirit Level have such huge influence? For there is no doubt that it served as a huge animating force in provoking public discussion and in re-energising debates about health and inequality. This went so far as see the formation of a powerful campaigning group, The Equality Trust, which has done much to highlight the systemic nature of health inequality (see the wider discussion in Savage & Vaughan, 2024). Once again it is the ‘world making’ qualities of The Spirit Level that stand out. It's clever use of visual assemblages offered anomalies to dominant modernising paradigms that saw economic growth as leading to better health and wellbeing. In arraying an alternative data assemblage, a new vista was opened up, and new associations, possibilities, and ideas could be revealed.

Let me be clear. These examples show the ‘world making’ power of descriptive social science but they are not empiricist, in the sense that there is any naïve belief that ‘facts speak for themselves’. Rather, both projects are deeply aware of the politics of data construction, the need to develop alternative metrics, and to embed their findings as critical interventions which point to discrepancies from the expectations that orthodox social scientific framings would invoke. In no way do they constitute data mining exercises. Descriptive sociology requires care and rigour. It is this careful assemblage that then poses anomalies to conventional perspectives. In both cases above, the array of data was inconsistent with modernising theories which prioritise economic growth. In short, good descriptive work needs to be theoretically situated and purposeful. We might see this kind of descriptive project as akin to Kuhn's argument about how paradigms are eclipsed—not by contestation between opposing views, but by elaborating how descriptive findings cannot be properly understood within conventional models. This process of paradigm shattering allows new worlds to be made visible.

Here again, the analogy with the aesthetic is again helpful. Artistic interventions rarely rely on a simplistic ‘realist’ rendering of a phenomenon, and do not seek didactically to insist on specific causal associations, as if there is some underlying message that audiences need to grasp, or they are missing something. They offer new ways of seeing, reading, listening and feeling which unsettle, arouse, attract, provoke and engage.

The contrast with sophisticated causal models, including those which have become fashionable with causal inference models, and randomised control tests, is very clear. Such methods depend on being able to isolate a range of factors so that their causal effects can be clearly identified. In medical trials, specific ‘treatments’ need to be identified, for instance by comparing the effects of a trial drug from a placebo and then clearly measuring the differential effects using standardised procedures on comparable samples of people. This is all well and good. But these analyses can only take place on the basis of ‘already existing’ factors for which measurement protocols have been established: one cannot conduct a randomised control test for a drug which does not yet exist. This is not a ‘world making’ kind of inquiry, even though the results may have real world impact.

In conclusion, I have defended a vision of descriptive repertoires in sociology as a transformative and powerful discipline. Descriptive strategies evoking new aesthetics, new imaginations and new possibilities can allow sociology to be ‘world making’. By posing new associations, patterns and displays, older paradigms and assumptions can be shattered. This is not to disparage causal analysis which also has a vital role. But it is an argument to reassert the aesthetic and the sense of wonder to the wider sociological palette. Now, more than ever, this is desperately needed.