HOW CREDIT INSTITUTIONS LOOK AT SOCIETY: Economics, sociology, and the problem of social reflexivity reconsidered

Present times are challenging for sociology, history, and anthropology, whose impacts outside academia appear restricted and confined to very particular matters, at least when compared with the magnitude of economics. Economics is, of course, also a social science, but one that has tended to distance itself from its neighbouring disciplines and, through the use of econometric and statistical models, to become apparently closer to the scientific paradigms of physics and mathematics. Not that classic sociological discourse has simply been dismissed as irrelevant by economic agents: the notions of ‘social capital’ or, less frequently, ‘embeddedness’ have undergone recent conceptual and political interpretations and used to better identify certain aspects of social relations that can be explored so as to maximize the profitability and efficiency of economic organizations (Marques 2003: 25–6; Cunha 2006: 226–7; for an interesting case study, see Narotzky 2006). The same could be said of Georges Friedmann's views on work as a process that should be, for the most part, controlled by the worker, which somehow prefigured the neoclassical economic orientation towards the idea of homo economicus as an autonomous agent (cf. Callon 2007b: 157). Notwithstanding Michel Callon's observation that the contribution of the social sciences in the shaping of homo economicus should not be neglected, the truth is, when adapted or translated into economic terms, many classic sociological concepts become a caricature of what their mentors and followers intended them to be and easily comply with the hegemonic principle of accumulation. On the other hand, the claims of some economists about the need for incorporating a stronger historicist and sociological persuasion into the discourse of economics – particularly through an attention to longue durée cycles or path dependence processes – have been generally disregarded, even in the face of the generalised predictive incapacity (see David 1985, 1997; Stoffaës 1991 [1987]).

In this article, we discuss how patterns of behaviour, collective aspirations, regional economic trends, and other matters of a certain sociological nature are used by credit institutions. We rely mainly on fieldwork done in the marketing, financial, and risk departments of two Portuguese retail banks (henceforth Bank A and Bank B). Our aim is not to demonstrate the existence of some lay sociology or folk psychology, since it is most probable that this sort of narrative will be present wherever there is a group of people doing things together – that is, wherever there is a ‘society’ functioning. Rather, by focusing our attention on these local varieties of sociological discourse, we seek to reconsider the Giddens predictions about the growing permeability of society and institutions to sociological insights and narratives of the self in late modernity (see Giddens 1990), and to compare this general version of reflexivity with the stricter perspective on the performativity of economics (see Callon 1998, 2007a; MacKenzie 2006; MacKenzie et al.2007). By showing how the images of society put forward by financial institutions bear more resemblance with the world view of economists than with current sociological descriptions stemming from the academy, we are achieving more than a statement of the obvious. Indeed, we may be taking one step forwards in the comprehension of economics as the actual pan-discourse of advanced societies.

It is worth noting that we are talking here of different kinds of images of society. For the marketing and risk departments of retail banks – as well as for many operational areas – ‘society’ may be considered as a ‘population’ or ‘universe’ of thousands of customers, be they effective or potential. Sometimes it is also a geopolitical entity equivalent to a state (in this case, the Portuguese nation state), traversed by global or regional socioeconomic trends. On other occasions, it is the ensemble of agents that compose ‘the market’ in its global, European, or national dimension. Finally, ‘society’ may also be akin to an institutional entity corresponding to the bank itself, with its hundreds of members of staff allocated across departments and branches, all working in articulation with one another. These different conceptions of society are purportedly vague, intuitive, and more grounded in situated practice and local common sense than in academic sociological formulations. They also correspond to different levels of perception and action, thus testifying that credit institutions, like any other economic institution, are far from being homogeneous worlds operating on the basis of uniform principles of coordination as simplistic digressions on hierarchy and the maximization of profit would lead us to think. Indeed, they are interwoven with tensions deriving from different coexisting modes of perception and organization, as well as from internal power struggles.

In this paper, we concentrate mainly on the perspectives of retail bank marketing, risk and financial departments in Portugal. Largely controlled by the state in the 1980s, with two or three prominent players dictating the rules and distributing goals among smaller parties, the Portuguese banking system underwent important changes as from the beginning of the 1990s, marked essentially by growing privatization and the deregulation of financial markets, along with a mass commercialization of credit products (especially credit cards and home loans). Retail banks played an important role in this social process that entailed a profound transformation in collective attitudes towards money, from a general private savings orientation to high levels of depositing and indebtedness, while the role of pawn shops and other fringe banks has remained almost residual. As a matter of fact, retail banks constitute the core of the Portuguese banking system, and, even if their dimension cannot be compared with the biggest players in Europe and the USA, there is some equivalence in terms of performance indicators.¹ Notwithstanding these national particularities, the truth is that similar movements of privatization, deregulation, and credit mass commercialization also occurred in many countries of Western Europe, following in the leading footsteps of the USA. Thus, most of what we seek to convey in this paper appeals to organizational processes and environments that have a strong transnational (if not global) character. When necessary, other specifications related to the Portuguese banking system will be set out, but the general argument and empirical evidence clearly both point to a transnational reality.

The general function of banking institution marketing departments is the creation and management of products in order to explore market opportunities and to expand business activity, taking into consideration the perceived needs of consumers and customers. The first thing to note about these needs is that they are not neutral facts or client behavioural properties, but rather the object of a collective elaboration involving specialized agents (working in banks, firms, consumer rights organizations, or supervisory institutions) as well as a heterogeneous multitude of consumers who – sometimes methodically, other times more randomly – contribute towards framing and reifying what are generally conceived of as ‘needs’ or ‘desires’ (cf. Knights et al.1994; Sturdy and Knights 1996). Therefore, it is hardly surprising to find that our Bank A and B marketing professionals do not rely greatly on regular and extensive consumer behaviour surveys. Some quantitative research may take place along with some interpretation based on inputs received by commercial branches. Nevertheless, the creation and management of credit products are highly conditioned by the legislation in effect as well as by similar products provided by banking competitors – following the coercive, mimetic, or normative premises of institutional isomorphism delineated by DiMaggio and Powell (1983; see also Meyer and Rowan 1977: 346). In other words, when developing credit cards, personal loans, car financing solutions, home mortgages, and even non-financial products (with the bank actually selling watches, computers, or travel services), retail bank activities are effectively encapsulated by circumstances that ultimately escape its own control – given they originate beyond its formal borders and represent that usually acknowledged as ‘the market’ as an organizational environment or a social entity, and not only the credit market but also the real estate market, car market, etc.

Accordingly, the shaping of credit products has to follow trends generated on the supply side, sometimes rather irrespectively of consumer proclivities. For instance, during the summer of 2007, Bank A decided to raise interest rates applied to mortgages (or what in Portugal is known as the ‘spread’, i.e. a small percentage added to the Euribor rate, usually described as the bank profit margin) and to adopt a more careful posture regarding collateral valuations as the first signs of the subprime credit crisis began to emerge. This happened at a time when Portuguese consumers – reacting to some aggressive marketing campaigns promoting the idea of a 0 percent spread – were still eager for more credit products, and consequently, the decision had a slight negative impact on Bank A's business.²

Therefore, what is known as ‘analysis of competitor products’ (that is, the methodical scrutinizing of what other banks offer in order to develop a still more appealing range) plays an important mediating role when it comes to purportedly adapting credit solutions to presupposed consumer needs. In fact, most of the tasks involved in the creation and management of credit products may be said to consist of bare imitation and conformity to the rules, leaving only a narrow margin for innovation. Furthermore, due to the absence of patents in this sector, credit services appear to be easy to imitate – a fact which, as Knights et al. (1994: 45) suggest, may explain the absence of deep research work beyond regular financial monitoring. In other words, all retail banks provide the same general credit products (credit cards, personal loans, car financing, home mortgages), and among these, almost all offer the same general solutions; for example, fixed or variable interest rates for home mortgages, interest-free payments for credit cards, spread reductions for all other products, associated insurance products, and so forth. As is otherwise common among all sorts of market products, which have to distinguish themselves against a background of sameness – see the reasoning of Callon et al. (2002: 202–3), inspired by E. H. Chamberlin's theory of monopolistic competition – the differences from bank to bank in standardized credit products are usually to be found in details such as the minimum spread offered, the maximum loan-to-value ratio allowed for mortgages, or the cost of fees and commissions.

Competition analysis developed in the marketing departments of banks A and B involved a set of specific strategies. First of all, there is constant and ongoing monitoring of other bank websites, outdoor campaigns, and mass media advertising. When encountering any novelty, there followed an activity described by marketing members of staff as ‘peeling the product’ (descascar o produto), which consisted of deconstructing the credit service into its fundamental parts in order to understand how it really worked (the lowest spread allowed, eligibility conditions, the additional costs involved, etc.). Evidently, consulting available information may be not enough to dissipate all doubts on the products of other banks, so it is common practice to establish contact with employees at rival institutions to obtain more information. These contacts are often by telephone and target specific interlocutors in other banks based upon a tacit rule of reciprocal cooperation that began in the early 1990s when Portuguese retail banks entered an expansion phase in their commercial activities. At the time of fieldwork (2008–2009), each marketing department held a list of contacts in other banks, a clear sign that these cooperative relationships were functioning on regular terms and viewed as perfectly natural. In certain cases, particularly when the interlocutors remain the same for a long time, this reciprocal collaboration mediated by phone may evolve into a more familiar relationship between peers, even if at different institutions. At this point, a particular marketer will often take the initiative of freely informing third-party and external interlocutors whenever a change in credit products eligible for publication occurs. Data gathered from other banks can also be compared to check whether there are any discrepancies.

However, some third-party interlocutors may pretend to be very busy, lacking the time to give out all the information requested, while others do so with great caution and reservation, thereby stressing the limits imposed by commercial rivalries. Hence, in this sense, informal inter-bank networks assisting in analysis of the competition may serve to promote cooperation as well as competition, with each marketer having their own perception of ‘warm’ (i.e. friendly and understanding) and ‘cold’ (impersonal, formal) contacts. Furthermore, as banks endorse a periodical rotation of functions among employees, certain ‘warm’ channels suddenly get shut down and replaced by more formal exchanges, since, while it is true that contact lists are handed on from one marketer to another, the same cannot be said of the friendly tone, cooperative will, and other situated qualities.³

The truth is that information provided by these contacts never breaks the rule of professional discretion. In practice, marketing staff are not authorized to pass on everything their colleagues in rival banks wish to know. Usually, confidential territory begins with what is referred to as the ‘spread table’ listing the profit margins of the bank discriminated by differing amounts and risk score categories. However, marketing analysts responsible for scrutinizing the products of other banks strive to obtain such information through the use of other strategies, not at all distinct from sociological inquiries, interviews, and even fieldwork – thus confirming that sociological instruments and techniques are much more readily accepted by organizations at large than social theory itself. In this case, marketing analysts disguise themselves as normal customers interested in certain credit products and call up other banks in search of information or even visit branches where they can talk personally with local clerks and pose specific questions. This widespread technique, known as ‘mystery shopping’, may also help in detecting other important business characteristics, such as the ‘quality’ of the assistance provided to the potential customer.

These anonymity-based techniques of analysis are made more viable due to a perceived inclination of certain Portuguese customers to pose several questions about credit services, sometimes to the extent of comparing one bank's products with those of rival institutions and confronting her local bank with the latter's presupposed advantages. Correspondingly, competition analysis also has to deal with client decision-making processes, setting out ‘arguments’ to be used in operational areas that highlight the virtues of the bank's own products and reveal the weaknesses of all others. Regardless, all banking institutions aim to protect a certain degree of information, saving them right to the contractual moment when a customer is actually buying a specific credit service and thus imposing limits on mystery shopping and other similar means of inquiry.

Notwithstanding the importance of competitor product analysis within the overall framework of marketing activities, some attention must also be paid to customer behaviour, to the urges and needs of credit consumers as well as to the performance of the bank's commercial areas responsible for the allocation of credit products. The quantity of contracts that need to be sold by each branch in order to meet business goals and objectives is generally set by marketing professionals. Once again, banker sociological proclivities come to the fore but in some ways less methodically than perhaps expected and sometimes clearly improvised. There is, of course, regular research in order to perceive how well certain products fare. The evolution of older credit contracts as well as the amount of new contracts is subject to monthly analysis that is then integrated into the set of reflexive activities that measure the prevailing state of the business. On this basis, it becomes possible to detect the regularity of certain tendencies, such as peaks in applications and contractual negotiation during the month or year, or the growing preference for laptops instead of desktops (in both cases, taking the universe of customers as reference), or the regional commercial departments that sell more and those that sell less (taking the institutional universe of the bank as reference).

This kind of control is based exclusively on numerical and financial data. (Credit contracts in effects? How much revenue is the bank due to receive in principal and in interest? How many contracts each regional department has to sign in order to meet its commercial objectives?) Once in a while, other types of studies are developed that aim at customer preferences, but which still draw on quantitative analysis. For instance, the bank may have an interest in knowing which personal loan types generate more success or why certain credit cards seem moribund. In any case, the data underpinning this internal research is generated at branches during the signing of credit contracts and subsequently stored in the bank's data warehouses to be finally processed and deployed by marketing professionals in terms of percentages and global amounts.

Indubitably, some enquiries appear to be clearly directed to more qualitative aspects of customer behaviour. Regarding home mortgages, for example, Bank B developed a study of bonus conditions designed to detect customers who were awarded a spread reduction but never regularized their situation (they had agreed to complete certain bureaucratic procedures but did not or could not act accordingly). The objective of this study was to understand and typify the reasons behind these irregular situations⁴ , so that, in the future, decisions concerning these cases could be automatic. Another study focused on specific modalities of home mortgages concerning building a dwelling, moving from one place to another or the carrying out of home improvements: each modality had a maximum duration of two to three years, but members of the marketing department suspected this period might be too short, and the study did show that this was in fact the case in certain contracts. In both examples, it proved possible to take note of concrete, specific, even atypical situations, mainly through collaboration between the marketing department and local branches: in the first study, a sample of customers with arrears was contacted via mail and invited to come to the nearest branch to regularize the situation.

Sometimes, however, banks may not have important data about their customers available. When responding to an inquiry by a Portuguese newspaper interested in overall home mortgage customer trends, one marketing professional at Bank A could not answer questions regarding the number of clients extending the terms of their loan or requesting deferment on the principal capital payment since the beginning of the year. The journalist was told that the bank was not in the habit of gathering such data, and while it would be possible to calculate a number, doing so would take an unaffordable amount of time, and thus the question went unanswered.

The truth is that, besides the financial information made up of customer account movements, banks know little more about their clients. The general information gathered during contracts is not updated and sometimes not even standardized, given it is dispersed across different computerized systems. For example, when guarantors were required for a credit contract – that is, people designated by the client to pay her instalments whenever she was unable to do so – the computerised application for car financing contracts in one of the banks included only two classification options (parents or not applicable), whereas the applications for other products considered a wider range of options (not applicable, spouse, parents, children, etc.). Furthermore, since the standardization of the computer systems gathering this information is rather expensive, the option is instead to create matching rules between pieces of information that do not coincide in order to insert them into large-scale calculation processes. It is exactly at this point that improvisation, experience, intuition, and tacit knowledge make their entrance in marketing activities, in large part as a consequence of a shortage of information and techniques.

Large-scale calculations based on customer data are the competence of risk analysts, and we shall return to them in the next section. For the meantime, and to conclude this excursion through marketing, let us consider other aspects of improvisation that sustain the approach to customer needs and behaviours. Marketing professionals gain hints about customer opinions on certain products through information provided by branches (or, indirectly, by regional commercial departments). The contact is normally established by branches when facing a concrete situation: one customer wants to contract a credit service under certain specific conditions, another wants to renegotiate her home mortgage claiming to have found a better offer with another bank, still another wishes to file a complaint, and so on. Since credit products have some degree of malleability, it becomes possible, in many situations, to create a specific solution for a specific case. It should be noted that marketers do not interact directly with clients but only with bank colleagues working in operational areas to whom they issue instructions by phone, concerning the functioning of credit products. Furthermore, these transversal contacts are not carried out on a structured basis, that is, they happen almost every day but are still dependent on the hesitations of branch staff when faced by a particular customer's questions. In other words, marketing professionals are thus prompted by phone calls from their colleagues at the commercial branches who, in turn, are prompted by direct contact with peculiar clients – and no systematic record is made of these frequent and rather contingent situations. However, through this chain, a certain image of customer preferences and behaviours is forged in the mind of marketers that forms part of their situated knowledge and may later assist in the reconfiguration of credit products.

Switching from the marketing to the risk department results in significant changes occurring in the way ‘society’ is viewed. Risk analysts concentrate more on the universe of customers, leaving aside that happening in rival organizations or in other areas of the bank (except as regards what is termed ‘operational risk’, which tries to measure eventual losses deriving from human resources, internal systems, and processes). The purpose of risk analysts may be characterized as the design of large-scale calculating models in order to classify clients (individuals or firms) on the grounds of a probabilistic conception of risk (cf. Marron 2007: 103).

The probability of default by a client – usually referred to by the acronym ‘PD’ – is the central issue at stake, or the dependent variable that must be explained. The data supporting this calculation are essentially quantitative and of a very large scale, signalling the adoption, by economic institutions, of sample methods originally tested in social statistics – a process which began in the 1970s in the USA (see Desrosières 2001: 344) and has now spread worldwide. These data are usually collected by branches during credit applications. In the case of individual customers, socio-demographic information such as marital status, employment status (precarious or stable), education level, home income, etc, is never updated – except if the customer intends to renegotiate the contract. On the contrary, firms and enterprises frequently provide the bank with their latest balance sheets and other accountancy reports in order to prove their creditworthiness. The fact that credit products for businesses and firms have shorter durations, with the possibility of being renovated thereafter, facilitates this free exchange of information. In any event, the content recorded in bank databases is treated as ‘reality’, independently of how accurately it may represent real people and institutions. We term this perspective ‘numerical realism’ in order to stress what forms its core (that is, figures and binary variables), but we could as well follow Desrosières (2001): 346) and term this a modality of ‘proof realism’ – where the reality that matters starts with the database, from which the analyst proceeds to obtain results that have to be plausible and consistent with the recorded information, irrespective of how this information was treated before it entered the database.

We have stated that bank customers are classified according to their probability of default. This means that the socio-demographic and financial data collected during each application are statistically processed through logistic regression calculation models so that, in the end, a specific score is obtained. Different scores are then organized by classes of risk. The number of these classes may vary according to the dimension and quality of the database: sometimes there are 10 classes, other times only six or seven (the formation of these classes is an automatic procedure carried out by SPSS software). In general, customers belonging to the first class are those whose probability of default is considered to be null. Correspondingly, customers whose score puts them among the last two classes are tainted with the highest default probabilities, and the bank should avoid negotiating any credit contract with them. If the socio-demographic information of a particular customer shares many features with customers belonging to the first class (who never default), then her score is likely to be high. Inversely, if she has too many factors in common with defaulting customers, the result will be a low score and possible rejection.

Logistic regression models establish this kind of relational correspondence by evaluating the explicative weight of a series of socio-demographic variables for the probability of default. There are several aspects at stake depending on the respective credit products. Considering credit cards, the variables most used are age, gender, marital status and personal property agreement, number of persons in the household, address code, degree of formal education, occupation and number of years in the profession, residence status (rented, owned, family home, etc.), payoff of the current credit card, possession of other credit cards, possession of long-term or short-term loans, ownership of a car (y/n), payment through bank account debit, and time of her relationship with the bank. Each customer forms a particular combination of these aspects (variables and co-variables), and the explicative weighting of each aspect depends on the previous default records existing in the bank's database.

Supposing that married and graduate applicants receive the highest scores, it may well be inferred that this is a consequence of the fact that marriage usually implies a higher household (providing both spouses are employed) income, and graduation favours access to better paid employment and subsequent career improvement. However, these are mere suppositions that might be uttered in daily chatter, informal meetings, circumstantial comments, and other casual situations that make up lay sociology in this context and without ever being set out in periodical reports. The sounder justifications for the scoring of clients are mathematical and statistical in nature. In fact, the explicative weighting of marriage status or degree of formal education is never fixed in advance: it may be more significant or more neutral depending on the cases registered in the database, itself always evolving and expanding.

One simple way of dismissing all of this is by simply ranking it as ‘bad social science’ justified by widespread methodological problems such as computing errors, limited sampling, and the utilisation of median values or artificial correspondence criteria to fill in gaps (for a critical review of this position, anchored in the prevalence of a reality prior to the database, see Leyshon and Thrift 1999: 448). In general, risk-analysis models have been part of what Leyshon and Thrift call a ‘quantitative revolution’ (ibid.: 436) in the business of retail banking, supported by technological devices such as computers, software, and other information systems that ensure the creation and processing of massive databases.⁵ As one might expect, more qualitative customer evaluation procedures such as interviews or personal acquaintance through long-term interaction, traditionally a prerogative of local branches, have been almost completely abandoned and substituted by quantitative approaches – an inevitable but also somewhat radical change, since qualitative methods based on personal contact were deemed to be trustworthy by those deploying them.

Another controversial dimension to probabilistic models of risk analysis is that they seek to predict what will happen in the future – default or regularity – on the basis of what has happened in the past within the restricted universe of those bank customers recorded in its databases. Implicit is a linear and deterministic conception of time enhanced by the early stages of statistics and probability calculation (see Reith 2004: 389–90) and, once again, dependent on data registration systems. It is true that the same could be said of any attempt at long-term social reproduction: having no way of knowing exactly what will happen, in many instances we assume it will be approximately like the immediate past and act accordingly. However, risk-analysis models and conceptions seem to have pushed this general presupposition to extremes, introducing a false sense of security by giving forth the idea that the future was somehow predictable and all major risks covered (for a similar reasoning applied to Value-at-Risk calculations, see De Goede 2004: 210; and Tett 2009: 39).

Of course, risk analysts ignore neither these nor other problematic issues, including the difficulty of determining the statistical relevance of a specific variable, that is, the exact weighting this variable holds in the analysis so as to enable the construction of general rather than idiosyncratic assumptions. As a result, risk-analysis models are indeed quite malleable. The process of data collection is permanent and cumulative: the data of new customers is continuously added to the mass of information already gathered. The payment behaviour of credit clients is also regularly monitored by looking at the evolution of the corresponding accounts, thereby enabling a calibration of the probabilities of default calculated at the beginning of the contract by comparing them with what really has been happening as recorded in the database. Every year or so, the models are subject to an extended revision, and certain variables may lose weight and be subsequently abandoned, while others gain greater explicative power or are newly introduced. For example, opening and building up a savings account purpose dedicated to later purchasing a home was considered an important indicator of the creditworthiness of home-loan applicants. However, when the Portuguese government decided to end the tax benefits formerly associated to this particular product, customers began to lose interest and started searching for other savings solutions. In Bank A, during fieldwork, this variable was being removed from home-loan risk models to be replaced by a new input, corresponding to the holding of any savings account. Once a new set of variables is defined, the model is fine-tuned through a series of statistical tests.

However, an impression of sociological unrealism persists perhaps as a consequence of the handling of huge quantities of data that turn empirical validation beyond the database into a futile exercise. As one corporate risk analyst at Bank B told one of the authors, ‘the particular reasons firm A or firm B defaulted do not matter that much to us; what counts is the weight of certain variables, like loan-to-value…’ (for a similar reasoning applied to American consumer credit, see Marron 2009: 163–4). It also seems clear that sociological approaches to risk analysis – which have developed enormously over the last 30 years, drawing on the charismatic influence of authors such as Giddens, Beck or Luhmann (for a review of this perspective, see Reith 2004: 384, 391–3; Rothstein et al.2006: 93–9) – have reduced impact on the numerical realism that drives these risk calculations. Risk is treated here as an objective (in the sense of accurately measurable) property of bank clients that serves to assist internal decision-making processes more than depicting a social reality evolving beyond institutional borders.

On the other hand, quantification may be dominant, but it is never absolute, even at the scale of each bank: risk analysts work mainly with samples, not with real quantities, since a lot of data are not correctly uploaded or are simply absent having been discarded at some stage. While the samples are very large in number, sometimes the lack of data may prove a significant obstacle preventing definite conclusions on certain subjects. In Bank B, for example, while developing a historical analysis of covered bonds, one analyst separated the home mortgages whose loan-to-value ratio (LTV) was inferior to 80 percent to observe their evolution in contrast with the evolution of the total number of loans included in the operation. He then detected some intriguing and unexpected discrepancies in the subgroup with the lowest LTVs that could perhaps be due either to variations in the effort rate⁶ of customers or simply to a shortage of information. His director urged him to explore the issue further, but after a series of inconclusive tests, the analyst was forced to choose the latter, unsatisfactory ‘explanation’.

Curiously enough, qualitative approaches have not been totally expurgated from the work of risk analysts, especially concerning corporate credit agreements. As stated, information regarding firms is more regularly updated. Contracted corporate credits are usually of shorter durations, but the amounts involved are significantly higher, so businesses need to be supervised more closely. Operational areas dealing with corporate credit usually include specialists on different business sectors that evaluate a specific firm in terms of regional impact, potential growth, and other aspects. Corporate risk analysts may participate in this team or work in a separate department, but they specialize and are more sensitive to socioeconomic indexes. Furthermore, the Bank A director of corporate credit (who coordinated two teams of nearly 20 people each, one in Lisbon, the other in Oporto) used to visit firms regularly: the visits were short, but he paid close attention to what might be indicators of clumsiness or disorganization in the manager's office or even in his oral speech.

Other types of qualitative research undertaken by corporate risk analysts included the perception of hidden connections between firms that appear quite different and totally independent in order to prevent a possible domino effect in case the central firm went bankrupt. Nevertheless, the information available was scarce and relied mostly on enquiries by branch clerks during the application process, when firm managers were asked if their company possessed any affiliates or special business relations with other firms.

Finally, it is crucial to stress that risk-analysis decisions are not unconditional and do not impose upon decisions in more operational areas. They are more like recommendations and, in many cases, are not actually observed by operational areas where members of staff continue to rely on more intuitive (from the risk-analyst perspective) procedures. In this sense, the so-called quantitative revolution is not yet complete in Portugal – or perhaps it should be conjugated with more qualitative aspects. The interaction between risk and marketing departments is also potentially conflictual: as a rule, marketing tries to present products that can be easily taken up by customers, whereas risk seeks to moderate this expectation by suggesting more severe terms and conditions regarding the spread or client profile for specific marketing campaigns. This tension is canonical and depicted in many risk-management books and manuals (see McNab and Winn 2000: 25) and illustrates perfectly something we referred to earlier: that the bank's point of view is far from uniform, and its actions are the result of several internal compromises.

Besides the marketing and risk perspectives, other images of society worthy of mention appear in the retail banking business. There is a juridical view of the society in which the bank is integrated (in this case, a nationally confined society), based upon opinions regarding the compliance of product characteristics and contractual conditions or the right to use certain customer related information. There is also a humanitarian view connected to particular non-lucrative campaigns run by the bank where profits revert in favour of charity and social assistance institutions. In this case, market principles seem to have been largely but not totally suspended, since this beneficent role usually exempts banks from a certain amount of taxation. And there is a still more complex view of the market as a social formation, a ‘life form’ or ‘greater entity’ (cf. Cetina and Bruegger 2002c: 169), where it appears as global in scale and impulsive in its moves, animated by feelings such as expectancy, uncertainty, nervousness, fear, panic, or exuberance. This psycho-sociological view evocative of nineteenth-century accounts on the hypnotic or mad character of crowds belongs to financial departments and, more particularly, to traders and financial directors responsible for controlling the bank's money and investing it in the various markets displayed across computer screens: foreign exchange currencies, equities, bonds, commodities, financial derivatives, and so on. We now take a closer look at this side of bank activities.

Trading rooms are relatively recent additions to Portuguese retail banks. They emerged in the mid-1990s in the biggest banks, encouraged by the growing importance of financial derivatives. In a certain sense, these rooms are windows open onto the world of financial markets and global economy. Their functioning depends on the constant flow of information provided by Bloomberg, Citigroup, Reuters, or Danske electronic terminals. This information comprises permanently updating graphs and numerical indicators and news statements arriving every second. The mass of data is effectively so huge that traders tend to filter it according to geographical criteria (for instance, the division between the USA, the Eurozone, and the rest of the world), taking into consideration specific market sectors (bonds and interest rates, stocks and stock indexes, foreign exchange rates). In comparison, there is less attention to the bank's retail business and even to important aspects of the national economy. As one trader explained: ‘I really love my job but sometimes I get the feeling that I know nothing about what is going on in my own country…’

Trading rooms have been the object of recent ethnographic attention. Cetina and Bruegger (2002a: 396) tell us of total trader immersion in the action in which they are taking part with eyes glued to computer screens and their unfolding flows of market information. In other articles, both authors speak of a new form of intersubjectivity marked by face-to-screen interactions in situations of co-presence and immediate response (Cetina and Bruegger 2002b: 923–30), or of ‘post-social relationships’ centred on the bonds between human beings and objects (Cetina and Bruegger 2002c). The anthropologist Caitlin Zaloom follows this line of thinking by stressing the ‘intensity of focus’ of traders and their ‘absorption in the moment-by-moment action’ (Zaloom 2004: 378; 2006: 104–5), but she takes this a step further to mention the presence of ‘weak narratives’ in the trading room, providing interpretations of the latest market movements and the development of ‘market-chatter’ between colleagues (Zaloom 2003: 266–7; 2006: 155–7). Furthermore, having done fieldwork on more than one site, she contrasts the noisy environment of a London futures dealing firm with the quietness and discipline of a similar trading room in Chicago (Zaloom 2003: 267; 2006: 157). In turn, Beunza and Stark (2004) treat the trading room not just as a ‘setting’ but also as a locale, with a specific organization that facilitates communication among traders and promotes the emergence of ‘interpretive communities’. In the same vein, Lépinay (2007) tells us of the importance of articulated narratives for the construction of a stable market ontology favourable to the circulation of financial products.

In sum, existing ethnographic accounts give the impression that the trading-room environment can vary substantially according to several factors; specifically, its own physical dimension and organizational layout, the degree of involvement in financial markets (which tends to be greater in investment banks and funds than in commercial and retail banks), the extent of prevailing market liquidity and their own ontological stability and the habitus of employees, that may favour either informal conviviality and a broad exchange of information or stern discipline and concentration on one's own tasks. However – and this remark surely hangs on Cetina and Bruegger's central assumptions – market movements also contribute towards influencing the particular social atmosphere in the trading room, by fostering discussion and co-interpretation (particularly in times of high volatility) or, alternatively, individual absorption in the flow of data unfolding on computer screens.

Our own observations in this regard are dependent on fieldwork conducted in the very small trading room of Bank A (comprising only six employees directly involved in trading activities) during the effervescent months of August and September of 2008. This was a period marked by several alarming announcements concerning banks and firms deeply involved in the business of credit derivatives, which peaked on 15 September, when the famous investment bank Lehman Brothers went bankrupt. In such conditions, characterized by high volatility and uncertainty, the above designated ‘psycho-sociological’ view of the market gained particular strength. There was a deep sense of unpredictability: the reality of markets evolving basically according to news and rumour and, as some traders commented, sometimes quite independently of theoretical models used by macroeconomists or technical analysis indicators such as the Relative Strength Index (RSI). In a nutshell, words prevailed – especially those pronounced by relevant international actors, such as the president of the US Federal Reserve, the Governor of the European Central Bank, or other important opinion makers. ‘Whenever Jim O'Neill speaks’, one trader said, referring to the senior Goldman Sachs researcher, ‘something happens!’ On another occasion, the director of the trading room alluded to a supposed emergency meeting of the Federal Reserve to discuss the situation of companies Fannie Mae and Freddie Mac, thus causing an immediate reaction from another colleague: ‘In that case, there may be news to come!’

The power of rumours grew considerably during that summer of 2008, prompting either impulsive reactions to buy and sell that generated abrupt movements in market indexes or inducing an anxious expectancy that translated into a decrease in activities roughly like the quietness preceding a terrible storm. Inside the Bank A trading room, the sharing of information was intense. Every morning, around 11 am, the group used to suspend trading activities to engage in an informal meeting to collectively discuss the main issues on the global and national economic agendas. This local practice had been introduced by the new trading room director, who wanted to instil collective reflection, quite independently of market fluctuations. Nevertheless, over those tumultuous weeks, the discussion often extended to other colleagues and financial department directors, who were constantly visiting the trading room. Contact with other brokers through chat rooms and by telephone, aimed at better understanding unfolding events, was also constant and ongoing. These contacts included staff at important banking institutions such as Citigroup, who sent in daily personal comments and even small articles depicting their view on the market situation. In this particular trading room (and possibly in many others around the world), the dynamic was thus one of storytelling and interpretation/articulation, entailing dialogue and discussion, plot weaving and thickening, and even philosophical arguing, rather than intense self-concentration combined with the use of technological instruments.

In such a context, the explanation most applicable to bankers is, of course, the theory of self-fulfilling prophecies popularized in sociology by Robert K. Merton (1948). This theory is based on the assumption that, if economic actors become somehow convinced that a certain firm is at risk of insolvency and act accordingly (by massively selling their shares), then that firm is most likely to go bankrupt, even were its financial situation in fact stable. In other words, the prophecy comes true only because people believe it may be true. From a sociological perspective, this argument is far from naïve, as it seeks to capture what seems to be a path-dependent process of rising performativity of a ‘story’ or ‘tale’.⁷ Traders and financial directors of Bank A complemented this canonical view with some of Fisher Black's insights about how rumours distorted market functioning (see Black 1986, another cherished reference for financial actors) or, less frequently, with recent achievements in ‘behavioural finance’, an approach that tries to figure out, among other aspects, how market crashes may be influenced by private investors deploying trading technologies in their own homes. This argument, not at all devoid of true sociological persuasion, holds that these ‘solitary’ investors are more vulnerable to the anxieties enacted by the market and consequently prone to thoughtless actions that intensify uncertainty, whereas institutional investors must ask, consult, and reflect before they act – in sum, the classic problem of embeddedness.

Following Callon (2007a), we could note that the psychological sociology stemming from trading rooms is still far too dependent on human motivations and beliefs, therefore seeming to disregard the importance of the materialities comprising socio-technical arrangements in the production of events that happen independently of what humans presuppose. For example, when the possibility of state intervention to rescue Fannie Mae and Freddie Mac was under discussion (in the immediate run up to Lehman Brothers’ insolvency), there suddenly appeared on computer screens an item of news stating that the company United Airlines faced serious problems. Spread worldwide, this information immediately caused an abrupt fall in the company's stock price. A short time later, an official statement by United Airlines detailed how the story corresponded to an event that had happened six years earlier and had somehow been referenced as contemporary by Reuters and Bloomberg. Reflecting on this intriguing case, one Bank A financial director ventured that whoever picked out that old report from the archives and reedited it as spanking new was perfectly aware that the powerful Reuters and Bloomberg search engines would automatically detect it and distribute it globally, therefore causing a drop in the company's stock value and launching more panic into the market (and perhaps also generating considerable profits to any opportunistic investors behind the move).

The notion that there are technological devices in support of system functioning and executing automatic tasks is naturally present in this view, but they appear here to be clearly subordinate to perverse intensions of agitating the market as a greater being that can only pertain to human agencies. In this, we find a good example of a trading room debated story-line or plot – inspired, in this case, on the leitmotiv of conspiracy. Other arguments took a more philosophical line such as vivid discussion over the decline of the American capitalist system right after US state intervention to rescue Fannie Mae and Freddy Mac and the rumour involving United Airlines, which took place in the informal morning meeting. In the words of the trading room director, ‘the state had to intervene in order to save the market, instead of letting the market evolve by itself’ – which to him clearly meant the system was not working as it should. Naturally, this kind of talk is by no means confined to within the walls of trading rooms: the main verbal clues continue to emanate from the various information terminals – like the daily comment by one Citigroup broker entitled ‘Welcome to the USSR (United States Socialist Republic)’, which arrived the morning after the aforementioned discussion – and so we may assume that the same kind of reasoning and plot-weaving takes place in other specialized financial contexts.

In sum, storytelling and interpretation were the rule in terms of local sociology, even if this perspective cannot be taken as absolute or totally representative of the complex agency of traders.

Let us once again return to the Giddens idea of an eminently sociological high modernity, whose institutions would be commentators and testers of sociological concepts generated in the academy, therefore extending social reflexivity beyond expert territory. Our brief excursion through Portuguese retail banking clearly reveals that banks are sites of knowledge production according to scientific terms and processes, though mostly based on situated research, method sampling, and experimentation – and thus akin to what MacKenzie (2003) calls bricolage – rather than on direct theoretical assimilation. A certain sociological inclination is also present, even if confined to an oral background realm within a context dominated by quantitative approaches derived from statistics, accounting, and economics (and, concomitantly, gaining special prominence precisely when these predominant quantitative approaches appear insufficient). In this respect, it is important to mention that the Bank B marketing director did seem deeply interested in the potential of an ethnographic methodology for studying the home loan-related decision-making process. That was, we believe, an important rationale behind the bank's cooperation with our research project, bearing in mind that ethnography is now frequently deployed by many companies worldwide as a marketing research method (see Downey and Fisher 2006: 18–19; Holmes and Marcus 2006). Thus, even while numbers seem to prevail, the scope of organizational knowledge is sufficiently broad to still integrate new cognitive factors.

However, as hinted at above, the link between these, to a greater or lesser extent vague, sociological inclinations and discourses produced in the academy is perhaps more difficult to establish. Sociologist participation in Portuguese banking staff structures remains minimal and occurs mainly at the level of graduates without any further contact with university. Of course, this is not a sufficient explanation, since most concepts, models, and tools employed by these banking professionals are, in fact, imported from the USA, a country where there is a tradition of exchange among sociologists, economists, and political scientists of theories and modelling techniques that have been assimilated by the financial sector.⁸ In any case, we feel tempted to state that the sociological perspective of retail banking more closely resembles a sociology ‘in the wild’ (to recall an expression applied by Michel Callon to economics, and that shall be discussed below), representing an amalgam of nineteenth-century crowd psychology, 1930s social statistics, and 1950s Mertonian insights mixed in with accounting, econometric, and probabilistic estimations. And we would suggest this is certainly not exactly the type of expert sociology Giddens had in mind when discussing the importance of reflexivity in late modernity.

This point deserves further elaboration. One aspect that could be said a propos the reflexivity of Giddens is that the conceptual formulation is perhaps overly simplistic, based upon the exchange of information between expert realms forming a cornerstone of high modernity and the lay realms of everyday life where people (including experts, naturally) engage in their lifestyles and carry out their own personal projects. Giddens does not say that this exchange has to be direct, but the two poles seem to be important to his vision of a reflexive modernity. Specifically regarding economic matters, this problem has been reformulated as a problem of performativity – or the study of the conditions necessary for a theoretical discourse able to shape the world according to its own parameters (cf. Callon 1998, 2007a; MacKenzie 2006; MacKenzie et al. 2007). In effect, this perspective entails more of a particular focusing rather than an abandonment of the concept of reflexivity which, as MacKenzie et al. note (MacKenzie et al. 2007: 3; see also MacKenzie 2001: 130), still matters when it comes to accounting for the nature of modernity and the interactions between science and society. The kind of feedback loops usually investigated under the assumption of performativity mostly take place within expert territory while bearing in mind that the word ‘expert’ does not designate a reality circumscribed to the academic world but also encompasses a large variety of other sites where specific competences are put into practice and deployed for the purpose of applied knowledge production – as with the five different sites involved in the decoding of a particular financial product depicted by Lépinay (2007), to cite but one example; hence the idea of an economics ‘in the wild’ or ‘at large’, and its separation from a more confined economics that assists the performativity hypothesis (cf. Callon 2007a). This is an idea and a distinction that may be not entirely convincing, especially when it comes to distinguishing the performativity of economics from the performativity of, for example, sociology, demography, accounting, or mathematics, given that what we find ‘in the wild’ emerges as more like tangled bodies of knowledge that are elsewhere confined and kept apart from one another.

To sum up, whereas Giddens considers the fate of theory as it moves on from expert to lay on the basis of a convenient separation between sociology and society, Callon, MacKenzie, and their colleagues consider the fate of theory as it passes from expert to expert on the basis of a no-less-convenient separation between confined economics and economics at large. In both cases, the usual distinction between theory and practice is blurred. However, the Giddens perspective is mostly focused on questions of identity, gender, and lifestyle, which are almost entirely absent in contemporary performativity approaches and appear to be clearly subject to economic imperatives in contemporary retail banking (in the sense that we see more of an economics in the wild acting through the work of banking professionals than an assimilation of theoretical sociology in the confined sense). This is fundamentally what may lead to a questioning of the scope of reflexivity as defined by Giddens through an acknowledgment that its rules do not apply to modern finance as effectively as they apply to identity issues permeated by gender, class, ethnicity, or age. It is true that the performativity approach is more appropriate to the ethnographic context of banking and finance; however, as it only provides an incomplete answer to the original Giddens problem of expert-lay feedbacks, it seems we should be paying heed to the waning social science input into certain organizational fields and their compliance instead with the terms and conditions of an economics at large.

A shorter version of this paper was orally presented during the 9th Conference of the European Sociological Association (Lisbon, 2–5 September 2009). The present version was revised by Kevin Rose, to whom the authors are obliged. We would also like to thank the editors of European Societies and one anonymous reviewer for their helpful comments on the original manuscript.

Daniel Seabra Lopes is an anthropologist and researcher at SOCIUS-ISEG/UTL. He is presently doing research on credit and banking, based on fieldwork in two Portuguese retail banks.

Rafael Marques teaches in the school of Economics and Management at the Technical University of Lisbon, and is a researcher at SOCIUS-ISEG/UTL, where he undertakes studies in the fields of moral sociology and economic and financial sociology.