Keywords

From a methodological perspective, working with the concept of vulnerability represents a double challenge. First, it requires finding the right indicators to capture vulnerable persons and/or the situations that make them vulnerable. Vulnerable populations are often described as ‘hard to reach’. There is an obvious link among precariousness, vulnerability, and disappearance from the statistics of the large surveys addressing the whole population of any country, even a rather small one such as Switzerland. For instance, illegal migrants are among the most fragile populations, not only due to their often dramatic life course but also because they do not have an official existence and cannot access basic social rights and social protection. Second, as highlighted in the introduction to this book, vulnerability also has political and ethical dimensions. Thus, working on vulnerability implies being sensitive to its various meanings that are relevant to research participants and being reflexive regarding the assumptions about vulnerability that guide us as researchers. For instance, people working in an informally regulated sector of activity are generally marked as a ‘vulnerable population’. Such is the case of most musicians, who, far from the large, steady classical orchestras, have to live day by day in an uncertain job market. Often, a part of their income is undeclared, and their total revenue is typically considerably below the national median level.

However, many musicians do not feel vulnerable—they do not view their position as critical, nor do they see a need to mobilise resources to escape from it as soon as possible. If not always chosen, their precarious position can be interpreted as a sign of their commitment to the artistic-romantic ethic of not ‘selling out’. Bohemianism, which may seem a vector of vulnerability from many points of view, is a lifestyle that is positively identified with ‘true’ artist life (see Bourdieu, 1975). Vulnerability is thus also difficult to grasp in this context because ‘objective’ criteria are not sufficient to define it: Considering subjectivity is thus almost a prerequisite to guarantee data quality and avoid misinterpretations. To grasp the two sides of vulnerability, building fine-grained datasets that contain robust information on both ‘objective’ and ‘subjective’ elements that frame individual life courses is necessary. Mixed method research designs can be of great help in reaching these aims, as we will present below.

In this chapter, we illustrate how we developed a mixed methods research design to survey ‘ordinary musicians’ living in French-speaking Switzerland. The project, supported by the LIVES research program, was entitled Musicians LIVES. We first seek to clarify our own working definition of ‘mixed methods’, its roots in a pragmatic approach to knowledge generation and research methodology, and why we see mixed methods research as a promising avenue to engage differently with populations that are reputedly hard to reach for research purposes. We then explain why and how two of us (Perrenoud and Bataille) developed the Musicians LIVES survey and how we designed our sampling strategy. Next, we present the data collection process, combining quantitative and qualitative approaches with a specific set of tools built around our main investigation instrument, the life calendar. In the final part of the chapter, we show how these concerns address several classic issues related to fieldwork among vulnerable, stigmatised or hidden populations and then discuss the broader relevance of mixed methods in large-scale research projects in social sciences aiming to engage with such populations.

What Are Mixed Methods, and How Can They Facilitate Research with Hard-to-reach Populations?

Some years ago, Perrigo Pelto (2015) discovered with surprise that mixed methods were being presented as a new development in the social sciences, with ‘some writers stat(ing) that this methodological approach only got going approximately 25 or 30 years ago’ (p. 734). Pelto himself had spent his long career blending numbers, texts and direct observations and, as early as 1970, had published a methodology textbook for anthropologists in which he presented numerous examples of previous studies taking similar approaches dating back another four decades. However, none of these earlier studies specifically referred to their methodology as ‘mixed’. Why not? According to Pelto, ‘the “qual/quan distinction” was generally not significant or special. They were all doing “science”, in which some phenomena should be counted, and others simply described without any numerical treatment’ (p. 741). Before perceiving the practice of mixing methods across research paradigms as noteworthy, social scientists first had to perceive paradigms that associated particular methods with different epistemological assumptions as meaningful realities. Pelto temporally situated the rise of such a perception within the last decades of the twentieth century.

Aiming to overcome the resulting binary opposition between quantitative and qualitative research orientations, Biesta (2010) discussed the possibilities of pragmatism, a notion that can be read both in its everyday meaning and in its more specific philosophical understanding. In its everyday sense, pragmatism suggests that researchers flexibly opt for the specific methodological tools that are most apt to answering their concrete research questions, rather than for a research paradigm as a package tied to abstract worldviews. It follows that more complex sets of questions sometimes require more diversified toolboxes. Philosophical pragmatism adds an epistemological foundation that views knowledge as neither objective nor subjective but—in a tradition of thought stretching back to William James (1907) and John Dewey (1929)—as grounded in the way we experience the consequences of our actions and are able to reflect on these consequences. For epistemological pragmatists, it is impossible to study the world from a spectator perspective: ‘If we want to know the world, we must interact, and as result, we will know the world only in the way in which it responds to us’ (Biesta, 2010, p. 19).

This focus on knowledge-generating social interactions is of particular relevance in regard to the study of social phenomena, or populations, that are traditionally viewed as ‘hard to reach’. In a thought-provoking discussion, Hardy and Chakraborti (2020) wondered whether these populations might not be described more accurately as ‘easy to ignore’ (from the perspective of established knowledge builders). This is certainly true of informal workers, as studied here, who tend to be invisible to public administrations, official statistics and GDP-type approaches to public wealth and productivity. Hardy and Chakraborti hence problematised the idea that certain groups are inherently hard to reach and argued instead that ‘reach’ and ‘access’ characterise social relations, not individuals. Consequently, transforming the social relationship between a research team and specific groups in society will alter the former’s impression of the latter’s reachability. To those researchers willing to circumvent common gatekeepers and directly ‘connect with and capture the lived reality of diverse communities’, Hardy and Chakraborti promised to eventually ‘recognize that the “hard to reach” are, in fact, eminently reachable’ (p. 52).

If entire communities can be ‘easy to ignore’ or ‘eminently reachable’ depending on the particular social relations established with them, a similar point can be made about the individuals composing these communities and—to disaggregate even further—about specific moments in the lives of these individuals. A typical shortcoming of quantitative approaches is that the requirement to work with samples prone to statistical treatment necessarily leads to some form of aggregating individual situations and hence bears the risk of overriding relevant variability across cases. As noted by Verd and Lopez Andreu (2011), a critical advantage of mixed methods designs is therefore that they allow for complementing the ‘variable-centred approach’ that is characteristic of quantitative analyses with a ‘case-centred approach’ through rich qualitative contextualisation of individual situations.

More specifically, Verd and Lopez Andreu (2011) advanced two types of arguments, of particular relevance for the present discussion, for adopting mixing methods from a life course perspective. First, from a life course perspective, ‘outcomes’ are always time-bound. A combination of life event surveys and life story interviews is ideally suited to reconstituting particular sequences of events that lead into and out of temporary states of, for example, well-being or vulnerability. Second, whether a given person is vulnerable at a given moment in time depends not only on a stock of resources objectively available to that person at that moment but also on the conversion factors ‘that can hinder or facilitate the transformation of resources—understood as means—into effective freedom’ (p. 7), as well as on the personal and collective values according to which certain ends are desirable. A lack of means to attain these factors and values represents a meaningful restriction of freedom. Both values and conversion factors (which include factors as diverse as, for example, personal knowledge, social norms or local infrastructures, p. 11) can be difficult to grasp with survey data or qualitative observations alone, as they typically span multiple scales and require a more holistic understanding of the contexts of individual and community lives.

In the present contribution, we follow the call to ‘do mixed methods research pragmatically’ and try to remember what Feilzer (2010) called researchers’ ‘“duty” to be curious and adaptive’ (p. 14). As we view the projection of too much epistemological meaning onto the ‘quantitative/qualitative’ binary as adding more confusion than clarity to current methodological debates, we prefer to approach mixed methods simply as the particular subcase of multimethods research (Anguera et al., 2018) that involves both numeric and nonnumeric tools to exploit a diversified set of research material. In line with the pragmatic approach that guides us, the remainder of this chapter gives due space to a detailed description of specific research practices that are aimed to answer concrete questions within a particular social setting. We then discuss some potentially transferrable suggestions for using mixed methods to facilitate research with hard-to-reach/easy-to-ignore populations that can be derived from a reflection on these practices.

Sampling a Hard-to-reach Population: The ‘Ordinary Musicians’

Ordinary Musicians as a Vulnerable and Hard-to-reach Population

When we started the Musicians LIVES project in 2012, our idea was to survey the work, employment and living conditions of those we call ‘ordinary musicians’. The notion comes from Perrenoud’s former ethnographic work in France (Perrenoud, 2007), and in many ways, Musicians LIVES was a development of this first approach. Ordinary musicians constitute the overwhelming majority of the people who try to make a living from their musical activities but are neither rich nor famous. This is the most common case, as only an infinitesimal minority of artists reach large commercial success or cultural consecration (Perrenoud & Bois, 2017). This kind of musician mostly plays ‘popular music’ such as jazz, rock, pop, or songs, but some also play ‘classical’ music; however, they do not have a permanent job in an orchestra. These musicians are seldom intermittent wage earners, but most of them live as independent workers who are self-employed.

We started from the idea that ordinary musicians had ‘vulnerable’ careers, in the material as well as in the symbolic sense. Indeed, the results of the ethnography in France during the early 2000s and the international literature (Thomson, 2013; Throsby & Hollister, 2003; Webster et al., 2018) have shown that musicians’ income is generally lower than that of the average population, highly variable from one year to another and even from one month to another, and often at least partially undeclared. Ordinary musicians’ material work conditions are often tough: night work, gear handling, car or truck driving to go home after the gig, etc. In addition, musicians’ employment status is often ambivalent, as they can be intermittent wage earners or self-employed. In a typical situation, the bar owner who hires a band for a night can be seen as both the employer and the client of the band. Symbolically, in most of the entertainment gigs that musicians obtain, they are often caught in an unclear work relationship ‘between arts and craft’ in which they have to manage professional identities as both an ‘artist’ and a ‘service provider’ at the same time (Perrenoud, 2006, 2007; Perrenoud & Bois, 2017). Their material difficulties also often lead musicians to hold daytime jobs, mostly part-time, which emphasises the difficulty of perceiving them(selves) as ‘professionals’. These material and symbolic forms of uncertainty make ordinary musicians vulnerable. However, such uncertainties do not imply that musicians are equally vulnerable at different moments of their careers, nor that every musician necessarily passes through phases of strong vulnerability. Nor do they imply that life configurations that might be characterised from the outside as vulnerable are necessarily experienced as such by the musicians themselves. The study therefore notably aimed to clarify how inherited and acquired resources filtered the experience of unsteady careers across a diverse range of musicians and how material and symbolic factors interacted to make the precariousness more or less bearable, if not almost appealing, for musicians from different backgrounds.

As evoked in the introduction to this chapter, the link between vulnerability and hard-to-reachness has been clearly established. Ordinary musicians, who are too few in number to appear as an occupational group in large surveys such as the Swiss Household Panel, do not have a single and steady employment status, need no licence to work (Hughes, 1996), are accustomed to undeclared revenues, and work at night in entertainment places such as bars, clubs, and restaurants. Thus, the occupational group boundaries are very difficult to draw: There are no lists or repertoires of ordinary musicians; consequently, our first issue was to circumscribe our population.

Starting from a Definition: What is a Musician?

For the reasons mentioned above, we chose to build our own sample. As Musicians LIVES started with an ethnographic phase of observation and informal interviews in 2012 and 2013, we could rely on the findings from this first step. This qualitative fieldwork clearly showed that defining someone as a ‘musician’ was not directly linked to income issues. ‘Professional’ musicians are not just those who earn their whole living by playing music. The individuals defined as ‘musicians’ by most of the people we met during this ethnographic work are, foremost, those who play and collaborate with other musicians through projects—such as playing in the same band or recording an album together. Therefore, in line with the interactionist tradition in the sociology of work and professional groups (Hughes, 1996; Becker, 1963), we decided to base our sampling strategy on a simple ‘relational’ definition because it seemed to be the one that makes sense for individuals of our target population: A musician is someone who is seen as such by his or her peers, someone to contract and play with and who is embedded within a larger musical collaboration network.

Online Network and Respondent-Driven Sampling: Two Approaches Yielding Contrasting Results

From this starting point, using a standard random sampling method was not feasible, if only because there was no pre-existing sampling frame from which ‘ordinary musicians’ could be drawn. Therefore, we decided to experiment with alternative sampling methods, especially different network sampling strategies. Network sampling methods comprise a set of different methodological approaches (see Heckathorn & Cameron, 2017), all of which are underpinned by a simple idea: There is a finite and small number of links between every individual of the same society (Forsé, 2012). Thus, if you ask somebody to ask somebody to ask somebody to place you in touch with one of the people whom he or she knows, you will be able to potentially reach any person, from any starting point, through only a few iterations. Could such a process thus be referred to as snowball sampling?

We started a first sampling trial following this idea. We selected seven people among those we met during the ethnographic survey (cf. Fig. 21.1). These people (or ‘seeds’, represented by a black square in Fig. 21.1) were meant to be as diverse as possible.

Fig. 21.1
Two different tree network groups. The web survey has 11 and the R D S survey has 7. Seed, I T W, overlap, and refusal nodes are marked.

Web and RDS diffusion trees

Full size image

We sent these seven people an online questionnaire that asked them to make a list of the musicians with whom they had played for pay during the last 12 months. To grow our snowball sample, we needed several personal contacts for each of these people. The very first persons we had met—the ‘seeds’—agreed to do so. Therefore, we had a highly satisfying first iteration that resulted in dozens of new contacts. However, we had never met these new contacts. Therefore, when these people received an e-mail from the university asking them to contribute to a survey, giving them a link to a questionnaire in which they were asked to give the names and contacts of all the people with whom they had played during the year, most of them refused. The majority did not click on the link, some clicked but did not complete the questionnaire, and at least one of them was so upset that he sent us an insulting letter asking how we could dare to be so intrusive.

Consequently, we decided to adapt our sampling strategy to make it ‘respondent-driven’ (see Heckathorn, 1997; Salganik & Heckathorn, 2004). This time, we asked our seven ‘seeds’ to contact themselves three people in their network, with whom they had played for money during the last 12 months, and to ask them if they would agree to meet a member of our research team. Seeds were selected for their durable integration into the musical professional space and their various profiles, from the specialists in private parties or corporate events to the regulars of avant-garde venues, from the punk-rock singer to the Baroque flute player, from the baby-boomers to the millennials—but all made a living by playing music. This strong integration of our ‘seeds’ within their professional networks was needed to launch the first recruitment waves optimally at the core of the professional space. As shown in Fig. 21.1, this approach was much more fruitful than online sampling. Only 34 interviewees (represented by a blue square in Fig. 21.1) were recruited through the online survey. The recruitment chains were very short (only one or two iterations). There was a wide majority of nonrespondents (N = 161, represented by a red spot in Fig. 21.1). Comparatively, the respondant driven sampling (RDS) survey met with considerably more positive response, and we ultimately gathered 123 completed questionnaires through 3 to 8 waves of recruitment. A more detailed presentation and discussion of our sampling method can be found in Bataille et al. (2018).

When Ethnography Helps Keep the Sample Diversified

Network sampling is typically used to study populations of which the size or social composition are not well known in advance. Deep ethnographic fieldwork can be of great help to obtain a better idea of the inner partitioning of the targeted population (McCreesh et al., 2012) and thus help the researcher make better-informed choices, particularly regarding the selection of the seed respondents. In the present case, unlike, for example, most studies using respondent-driven sampling in a public health context, we did not seek to create a sample that was statistically representative of the reference population but that adequately reflected the diversification of social situations experienced by ordinary musicians in Switzerland. It was therefore critical, regarding our respondents’ social background, to carefully check that the seeds represented a large diversity of profiles. Nonetheless, seed selection was only the starting point of a dynamic process during which the sample composition evolved, and it was equally critical to monitor that social diversity was not lost in this process. We knew from our ethnographic work in France and Switzerland that ‘ordinary musicians’ are rarely recruited among the top of the upper class or the more ideotypically working class families, instead coming most of the time from the various fractions of the middle class. The main internal variation comes from their greater endowment of symbolic resources—especially diplomas. Figure 21.2 represents the rate of women (A) and respondents regarding their social background (B) recruited at each wave of the RDS recruitment process, with a 0.9 confidence interval. These results indicate that people with an upper-class social background were cited more often and were more inclined to answer our survey during the first two waves. Interestingly, however, in the fourth wave and after, our respondents’ social background again became increasingly diverse. In other words, continuing recruitment beyond the second of three iterations turned out to be a decisive factor in reaching a sufficiently diverse range of social situations, which matched our research objectives.

Fig. 21.2
Two graphs of estimated percentage versus recruitment waves for gender and social background. It depicts the mean lines for women and mean lines for people of lower, middle, and upper backgrounds.

Recruitment dynamics regarding gender and social background

Full size image

Combining qualitative and quantitative methods hence allowed us to design an effective sampling strategy. However, the use of mixed methods was also our main approach for the data collection about musicians’ lives, occupational activities, careers and lifestyles.

A Mixed Methods Survey Combining Ethnography and Statistics

The Fieldwork Protocol

Establishing a constant protocol for each of the six people who performed the fieldwork was a long and difficult part of the project. Perrenoud and Bataille supervised the process and conducted approximately a dozen interviews each, but four assistants were hired to reach a total of 123 interviews in the year between 2014 and 2015 (Perrenoud & Bataille, 2019).

At first, we hoped that the self-completed online questionnaire would work as a sampling tool and that we could, at a second timepoint, meet some of our respondents for in-depth interviewing. As online sampling failed, we rapidly abandoned that idea. We nevertheless aimed for a protocol that could allow us to obtain accurate standardised data and to conduct rich interviews. To do so, we decided to develop a life calendar that would work as a real mixed methods research device.

The calendar was printed on an A3-size sheet that we then placed on the table to share with the interviewee. On the X axis (see Fig. 21.3), a double timeline showed the respondent’s age and years as a musician. On the Y (idem) axis were several thematic sections regarding the respondent’s musical career and life course in general (especially bands in which the person had played; musical and nonmusical education and training; musical jobs as performer, composer, teacher, etc.; employment outside the music business; biographical landmarks such as getting married or the birth of a child). The interviewer and the interviewee each had a pencil and could fill in the calendar while they chatted. For the bands in which the interviewee had played, we had a chart on which we could precisely note information such as the name of each band, the type of repertoire (original compositions or covers), the type of venues where it played (for instance, concerts in halls, entertainment in bars, or as anonymous background music for commercial animations), the musical style, the band lineup, and the average number of gigs per year (see Fig. 21.3). At the end of the interview, we had some complementary tools, such as a short sociodemographic questionnaire to ensure the collection of certain basic information that could have been forgotten during the interview and a dynamic revenue bar chart on which the interviewee could trace with a marker the composition of his or her musical revenue (gigs, studio sessions, teaching, royalties, merchandising, etc.) at different sequences throughout his or her career. Then, we finished with the networking portion, first asking the interviewee the names of the people with whom he or she had played during the 12 past months (ultimately obtaining a 1300-person network) and then asking them to contact three people in this network to ask them to be part of the survey.

Fig. 21.3
The images are of comments for hand-written calendar notes from 1986 to 2014. The formation, travail non musical, and travail musical, with various classifications of these terms mentioned.

A filled Musicians LIVES life calendar

Full size image

Therefore, our life calendar essentially functioned both as a device for obtaining standardised data for quantitative analysis and as an elicitation tool for the semistructured interviews, each of which lasted approximately two hours and was recorded. It took approximately two months for the team to develop a cautious and precise protocol for using this life calendar and complementary tools. The four people who were hired were master’s students in the social sciences and musicians, mostly with a certain experience of the trade (as for both Perrenoud and Bataille), so we tested our survey protocol on one another until we reached a satisfying routine, which proved to be efficient in the first musician interviews. We rapidly obtained results regarding the main characteristics of musical work and the different types of musical careers in the Swiss context (Perrenoud & Bataille, 2017, 2019; Bataille & Perrenoud, 2021) that we could further compare to patterns found in previous work, especially in France (Perrenoud, 2007).

Beyond General Social Characteristics: Analysing Variability and Subjectivity

Over the course of approximately one year, six of us conducted 125 interviews with musicians in French-speaking Switzerland. At the end of each interview, we completed an input mask on LimeSurvey to obtain all the standardised data concerning the social background, career and network of each interviewee. We did not have the means to transcribe all the interviews, but each interviewer had the sound files that he or she had recorded and could transcribe whole interviews in some cases and specific parts in others.

Standardised quantitative data were analysed with R software and its extension for sequence analysis, TraMinR. In our first approach, these analyses allowed us to highlight that, when considered as one group, the ordinary musicians who took part in our study displayed distinct social characteristics: Compared to the larger Swiss population (as reflected in the Swiss Household Panel), they tended to have a higher educational level and a lower income than the average. Regarding gender, we also found—as expected—a vast majority of male musicians. Flat sorting to perform such descriptive work was straightforward, and we could quite clearly characterise different types of social conditions among the ordinary musicians included in our study. Nevertheless, understanding the meaning of some objective disparities within the group and illustrating the different types of musical careers and the stratification among the diverse ways to ‘be a musician’ became far more complex. To accomplish these goals, we had to obtain the meaning of some variations from one individual case to another, which required an understanding of the occupational culture. To obtain such an understanding, we needed a qualitative approach.

For instance, we clearly found that ordinary musicians frequently had several sources of income from their musical activities. Musicians can be paid to play in public but also to record in the studio, to give music lessons or to conduct a choir; they can also earn royalties through copyrights or selling merchandise. We observed cases of extreme contrast: Some people made a living almost exclusively through public gigs throughout their entire career, while others had various revenue sources such as frequent studio sessions, some copyrights, and perhaps music lessons. Some people had ‘daytime jobs’ in addition to the music business, and others earned a living almost exclusively by teaching music throughout their career, with fewer onstage musical activities. Those different profiles appeared to be largely determined by the social class fraction of origin and the ensuing cumulative advantages or disadvantages: owning a cheap or expensive instrument, having studied composition or not, having the idea and the desire to compose music or not, pairing with well-known musicians or with local fellows, receiving a parental grant during the long and uncertain beginning years of the career or having to work to earn a living, etc. However, they were also partially dependent on experiences and events encountered across the life course, for example, dedicating oneself exclusively to music after a negative event in the nonmusical professional sphere or deciding to turn mostly towards music teaching after having a first child.

Among those different profiles, types of musical activities, and income sources, we sought to determine which were the most attractive and the most aspirational for musicians. An objective analysis in terms of employment quality (UNECE, 2015) would assume that musical teaching is more steady, more often officially declared, and with a higher minimum wage than what can be found on the more-or-less underground trail of the musical bars. Such an assumption has even been a source of misunderstanding for some colleagues who, as sociologists of work, found that musical teaching was much more suitable and enjoyable than spending a life on the road travelling from gig to gig.

However, with our data collection design, we were able to move between statistics and individual cases and to ‘zoom in’ on respondents’ comments regarding their objective bundle of tasks and the way they were making a living at different moments of their career. The subjective comments were unequivocal: Teaching music was a second-choice activity that was less prestigious than playing onstage. The stage is the place where the musicians’ professional identity is stated and asserted by the interaction with an audience. Whatever the musical genre or style, we found quotes in the interviews that expressed, about a given period of time on the life calendar, sentiments such as ‘It’s sad, in that period I only have teaching activities’, or ‘At that time, we didn’t get as many gigs, I had to take more teaching hours’.

This complementarity of quantitative and qualitative approaches allowed us to show that, in contrast to common ways of evaluating employment quality, the occupational group we studied was marked by a peculiar professional culture inherited from the romantic Bohemian mythology. In that specific ideology, vulnerability was seen as the normal state of the ‘artist’, who is never certain of what will come next in terms of ideas, work, and income. However, as atypical as it may seem, this kind of romantisation of social vulnerability was previously described by Boltanski and Chiapello (1999) as a major part of the ‘new spirit of capitalism’ and seems today to extend to all sectors of the labour market (Menger, 2002; Perrenoud & Bois, 2017; Bataille et al., 2020). The ideological advent of the ‘entrepreneur’ as a new and hegemonic anthropological paradigm in ultraliberal societies relies on that romanticisation of vulnerability: On online service job platforms, for example, everyone is supposed to commit to the ‘adventure’ of ‘independent’ work to ‘be creative’ and self-motivated in a general context of increasing precarity.

Discussion: Learning to Mix Methods to Unlearn That Certain Groups Are Easy to Ignore

Research methodology is inherently about inclusion and exclusion: Choosing a specific method, often unwittingly, implies the acceptance of ignoring certain facets of the social world that might have become visible with other methods. In our view, mixed methods open fruitful avenues for research practices that include a broader range of perspectives (see Elcheroth et al., 2019) when constructing the ‘intersubjective worlds’ that pragmatist epistemologists see as the foundation of all relevant knowledge generation (Biesta, 2010). There are direct and indirect reasons for this. As with other types of multimethod designs, mixed methods are likely to make visible phenomena that would have gone unnoticed within a single-method design. Moreover, they add an element of flexibility and adaptability to the research process, which can be a critical indirect benefit when the aim is to reach populations that are reputedly hard to reach with standard procedures (see Laganà et al., 2013).

We have illustrated here how these different aspects can be combined in a study designed to make visible a population of ‘ordinary musicians’ that typically remains hidden in the social landscapes described by social statistics. This experience showed that being ‘hard to reach’ is not an inherent characteristic of the target population but depends on the concrete social relations between (professional) researchers and (professional) musicians. Just as informal work is ‘easier to ignore’ (to rephrase Hardy & Chakraborti, 2020) from the perspective of public administrations than for the thousands or millions of people whose livings depend on it and whose lived worlds are shaped by it, ‘ordinary musicians’ might be harder to reach from our university offices than with the help of peer musicians in the field. Consequently, the status of being ‘hard to reach’ can change when the relevant social relations change. In the present case, adaptive elements of the research design, such as integrating qualitative elements into survey interviews (i.e., allowing for open-ended elaborations on relevant topics) or cooperating with earlier research participants to identify further participants, changed the climate of individual interactions with participants, as well as the perceived meaningfulness of the study within a closely connected community. These changing social relations eventually opened doors that appeared closed during earlier stages of fieldwork.

This example shows how the functions of a classic sampling procedure in ethnography—here, snowball sampling—can be creatively expanded for old and new purposes. Here, it allowed the creation of a diversified sample that reflected a relatively broad range of social realities experienced by musicians in Switzerland. While certain accounts proved useful in describing variable social conditions within this sample, they were not designed with the primary aim of making quantified inferences to a broader reference population. When such inferences are a research objective, other ingenious variants of network sampling can help bridge the type of aims described here with those of classic probability sampling. In the context of the LIVES program, for example, controlled network sampling allowed for the creation of a representative cohort sample with an oversample of the (statistically) hidden population of second-generation immigrants (Spini et al., 2019). In that study, network sampling generated further benefits: It allowed us to diminish attrition in subsequent waves in comparison with research participants recruited through classic random sampling (Brändle et al., 2017) and to exploit the tie structure used to generate the sample to describe different configurations of social capital in the reference population (Guarin, 2020).

To conclude, we hope that our contribution helps clarify, first, why it is important to denaturalise and disaggregate ‘hard-to-reach’ populations and, second, how mixed methods designs can help reach these aims. Denaturalising hard-to-reach populations implies thinking more in terms of connecting social relations between researchers and research participants rather than in terms of certain populations being more or less reachable per se. It also means considering the temporality of social phenomena, which is central in the life course perspective: A person or a group can be ‘easy to ignore’ at one moment and ‘eminently reachable’ at a different stage of their personal or collective trajectory. For instance, a musician can experience a humble and marginal beginning, then have some success and become more visible at a point of his or her career, before perhaps becoming old-fashioned and disappearing again from the limelight. The pragmatic orientation underlying mixed methods research provides the necessary flexibility to adjust methods to accommodate concrete research goals and to create research designs that facilitate the social relations necessary to connect with specific target groups at specific moments in time.

Disaggregating ‘hard-to-reach’ populations implies being precise about concrete types of reasons that can make it more difficult to create these connections. Ellard-Gray et al. (2015) distinguished among remoteness, vulnerability or invisibility by calling for different types of responses. When remoteness is the main hurdle, cooperating with social actors who are closer to the target group can help overcome it. When groups are particularly sensitive to stigma or other potential harm, a clear ethical orientation and the capacity to convey it in a meaningful way will be an essential foundation of relationship building. When the group is ‘hidden’—i.e., difficult to identify—the key effort demanded might lie in constructing new frames to find and recognise group members. In practice, the three types of reasons can overlap but frequently do so in unequal proportions. In the present example, the creation of a new knowledge map to delineate a fuzzily defined social group was the main focus, but connecting with concrete group members also required sensitivity to sometimes precarious social conditions as well as specific social skills to bridge different cultural or organisational environments.

Tackling these challenges required methodological creativity, but eventually, it also reified the underlying conceptualisation of vulnerability that oriented the study. The mixed methods design notably allowed the description of professional ethos among many musicians that positively connoted situations that sociologists of work would describe as precarious. If we want to avoid naturalising certain values—say, the values that prioritise full and continuous participation in consumer society—we need to take seriously the plurality of values revealed in this way while still contextualising their emergence and maintenance. Such an approach might eventually lead us to wonder to what extent vulnerability stems from inconsistencies between specific values and opportunities rather than simply from limited resources to reach self-evident ends.