Developing an analytical framework for multiple perspective, qualitative longitudinal interviews (MPQLI)

Abstract

Collecting multiple perspectives data (e.g. from related individuals) in a qualitative longitudinal design can provide rich understanding of the dynamics at play in complex relational systems, and the different perceptions of people involved. However, such approaches are inherently challenging due to the complexity and volume of data involved. So far, little attention has been paid to the methodological challenges of data analysis in multiple perspectives longitudinal research. This paper contributes to the development of a systematized analysis process for multiple perspectives qualitative longitudinal interviews (MPQLI). We present a framework for handling the complexity and multi-dimensionality of MPQLI, describing discrete steps in such analyses, and related aims and insights. We exemplify the suggested strategies with our own research on the transition to parenthood. The proposed framework can increase both traceability and credibility of analysis of MPQLI, and help to realize the potential of multiple perspectives longitudinal interviews.

Keywords:

• Qualitative longitudinal
• multiple perspective
• data analysis
• qualitative interviews

1. Introduction

For both qualitative longitudinal (panel) and multiple perspectives approaches, there is a lack of methodological literature generally, but particularly regarding the analysis of data gathered by using these methods (Holland, Thomson, & Henderson, 2006; Mccarthy, Holland, & Gillies, 2003; Zartler, 2010). Qualitative longitudinal research (QLR) adds a time dimension to ordinary qualitative research, incorporating processes and change. It refers to research that is conducted over a longer period of time in at least two waves of data collection. This approach has been increasingly popular over the last decade and has stimulated much methodological discussion in social research (Calman, Brunton, & Molassiotis, 2013; Holland et al., 2006). In a similar vein, in order to grasp a better understanding of dynamics within a group or unit, many authors advocate for a multiple perspectives approach (Burton & Hardaway, 2012; Carr & Springer 2010; Gager & Sanchez, 2003; Reczek, 2014). In multiple perspectives research (MPR), accounts of related individuals are included, e.g. family members, close friends, doctor and patient(s), teacher and student(s), to gain a more nuanced and more comprehensive understanding of the functioning of interactional systems or groups (like families) or life circumstances of individuals from different perspectives (like doctor and patient).

Combining qualitative longitudinal and multiple perspective data further increases the complexity of analysing the data gathered, far beyond even the initial complexities in using each methodological approach individually. A recognition of these difficulties is the starting point of this paper: Complex and multi-dimensional data requires innovative strategies for data analysis and display (Holland et al., 2006). In order to make full use of the data, a systematic and methodologically sound strategy is required for collection and analysis. Firstly, to increase the traceability and credibility of qualitative research, and secondly to help making the complexity in multiple perspective qualitative longitudinal interviews (MPQLI) manageable.

Addressing this methodological gap, we first give a brief introduction to the potential of qualitative longitudinal research and of multiple perspective interviews. Then, we combine these two approaches and develop an analytical framework for MPQLI and illustrate the suggested procedure with empirical examples. Finally, we discuss challenges and directions for future research.

2. Combining multiple perspective and qualitative longitudinal research

2.1. Potential of qualitative longitudinal research

In qualitative longitudinal (panel) research (QLR), the same people are interviewed several times in roughly fixed intervals (e.g. every two years) or around certain events (e.g. before and after childbirth). In the literature, no specific time-span is defined that would make a study longitudinal. However, the minimum consensus seems to be the elapse of sufficient time between waves for change to appear, because change (or stability) is the core research interest to conduct longitudinal research. The time-span between waves then depends on the specific research interest in a certain research project and goes beyond a before and after measurement used in experiments or time-series in psychological research. Thus, longitudinal research can (a) detect changes over time, and (b) explore the processes associated with change or stability as well as (c) interpret the perspective of the person experiencing that change. In sum, QLR traces lived experiences and processes of change or stability (Calman et al., 2013). It is well suited for studying life trajectories (Henwood & Procter, 2003; Miller, 2005, 2011; Thomson, Kehily, Hadfield, & Sharpe, 2011; Thomson & McLeod, 2015; West, 2012) because it can capture critical moments of change and transitions. Gathering data consisting in individual narratives and sense-making of processes is the key element of QLR. Sense-making can be explored as a process and not just as a snapshot (West, 2012): thus, QLR facilitates an understanding of the dynamics between context and subjectivity.

According to Lewis (2007) qualitative longitudinal studies reveal different types of change:

(1)	Narrative change refers to the unfolding of individual stories across time. Qualitative longitudinal studies refer to narratives rather than chronological change because participants do not tell stories in a neat linear way. ‘People forget things, they think they have already mentioned them, they prefer not to volunteer them until the relationship with the researcher has developed over several interviews, or something only becomes relevant in the context of later events’ (Lewis, 2007, pp. 548–549).
(2)	Participants’ reinterpretation of experiences or feelings that they described earlier; that is, issues that seem very important at one time-point but whose perceived importance may change in the course of a study (Calman et al., 2013).
(3)	Researcher’s reinterpretation could emerge after a series of interviews, when new information shows a participants’ experiences in a different light.

QLR has the potential to flexibly adjust instruments and specify research questions according to new insights gleaned through the whole research process. The research can respond to change and can be adapted to individual narratives. Because processes or developments may not be clear before the research is conducted, some data collected in the first wave may turn out to be unrelated to the emerging process over time. ‘Flexibility and responsiveness to the data and emerging analysis and interpretation is a key skill for the [longitudinal qualitative] researcher’ (Calman et al., 2013, p. 7).

Beside its potential, QLR is also challenging as a methodology. One complicating feature of QLR is (1) the multi-dimensional and complex structure of the data (Calman et al., 2013). To identify changes and the process of change, the analysis shifts between interviews, across participants and longitudinally within individual narratives. This leads to methodological challenges in developing a consistent structure and focus, and in creating a systematic approach for analysing change and the processes underlying it. The size of data sets can also cause logistical problems.

In addition, (2) the resources required vastly exceed those needed in cross-sectional approaches. Some longitudinal studies will have to run for years, which makes project management very time consuming (Calman et al., 2013). Furthermore, the longitudinal character of such studies does not always fit neatly with the demands of funding bodies: A funded project cycle will most often only cover three to four years. If the subject requires a longer period of study, researchers will have to bid for new funding resources part way through an existing study, which lends enormous uncertainty to the planning of longitudinal research. Furthermore, a longitudinal study requires a long term commitment from researchers and respondents, which can be very difficult to secure.

Another challenge is (3) the absence of analytical closure, which Thomson and Holland (2003) identify as the most challenging aspect of QLR. New data in subsequent waves can potentially render previous interpretations redundant or obsolete (Thomson & Holland, 2003). Because of the continuous re-interpretation and re-examination of data involved in QLR, it can frequently be a difficult matter to find a natural and satisfactory end-point for the analysis of any given process.

2.2. Potential of multiple perspective interviews

In multiple perspective interviews, we refer to information from related individuals (Mccarthy et al., 2003) on a joint matter. These related individuals could be members of the same group, e.g. intimate partners, parent(s) and chil(dren), friends, team members, doctor and patient(s), teacher and student(s). The rationale is that interviews with only one member would fail to offer information to reconstruct mutual influences and differences in sense making. It is assumed¹ that different views or ‘realities’ exist within one group (Gager & Sanchez, 2003) and that different perspectives on the same experience exist between individuals coexisting in complex relationships (Reczek, 2014; Warin, Solomon, & Lewis, 2007). For example, family life ‘can be absurdly disjointed. (…) Once we go inside families, we find that it is an illusion to talk about “the family”, as though it were a single entity. The family is the meeting ground of multiple realities’ (Larson & Richards, 1995, p. 189). Thus, hearing perspectives of different members of a group contributes to ‘a more complex and nuanced understanding’ of their shared experiences (Valentine, 1999, p. 67), and views can be better situated within the social relationships wherein they are construed (Harden, Backett-Milburn, Hill, & MacLean, 2010). Relying on a single perspective cannot suffice when the unit of analysis is a ‘relational unit’ or group or the roles different members play.

Research questions for multiple perspective studies take their units of analysis to be relational units, like families, and the mutual impact of related individuals. Contrasting views allow for insights regarding the dynamics of a relationship, and can stimulate theories that have more explanatory power than the sum of theories based on individual perspectives (Eisikovits & Koren, 2010).

In our empirical example (see below), multiple perspective interviews are conducted with related persons in separate interviews. Separate interviews with members of a group allow for studying individual perceptions and understandings, taking account of their interrelation and the context of their relationship (Eisikovits & Koren, 2010). Participants have more freedom in expressing their own views than they might in joint interviews, and the setting allows for more privacy and confidentiality than in joint interviews (Valentine, 1999).

Challenges in multiple perspective research lie in the interpretation of contradictions and inconsistencies which are common in multiple perspective interviews. To illustrate the problem, we conceptualize potential outcomes of an analysis of such cases: Comparing two (or more) individual interviews, results can (a) converge, (b) complement one another or (c) diverge from respectively contradict each other.

Multiple perspectives reveal the complexity and tension between individual accounts. At the same time, a difficult epistemological question is raised as to how researchers should make sense of a diversity of perspectives, and how they could integrate that multiplicity and tension into a cohesive narrative (Mccarthy et al., 2003). Both similar and contradictory views of participants can legitimise multiple readings of a relationship and facilitate different insights into that relationship. Particularly dissonant data has the potential for new theoretical insights and a rich understanding of relationships, which would not be possible without the multiple perspectives approach. Gaps in testimony or silences in interviews can be due to a lack of communication between people, or they can reflect different positions (Mccarthy et al., 2003, p. 15). Agreement in participants’ testimony, on the other hand, can reflect jointly constructed versions of reality or can be rooted in a script (Mccarthy et al., 2003).

2.3. Potential of multiple perspective qualitative longitudinal interviews (MPQLI)

The potential of MPQLI lies in the combination of longitudinal and multiple perspective research: The rationale is to detect different views on a specific matter and (processes of) change or stability in the accounts in relation to each other. This yields insights into the functioning of a group from members’ perspectives at a specific moment in time, but also the change or stability of these perspectives in relation to one-another. The reason to consider multiple perspective interviews is the acknowledgment of the multi-layer and multi-dimensional nature of interactions and the meaning-making thereof. Acknowledging this constructive nature also raises questions about congruence and discrepancies over time, e.g. after a critical event in the joint life of group members. To answer these questions, we need MPQLI.

Specific for MPQLI is that although the fundamental unit of research is the individual, the questions posited often address the relational unit and the change or stability over time; for instance, the nature of the social construction of family reality between individual family members. Thus, we have to address the challenge of integrating individual accounts which are inevitably interwoven to varying degrees. The longitudinal character adds yet another layer of complexity. The individual narrative may change over time, leading participants or researchers to subsequently reinterpret specific experiences or findings within an altered context or frame of reference (Calman et al., 2013). At the level of the relationship, accounts of members of a relational unit and the nature of their relationships, interactions and sense-making thereof can change significantly. What was once complementary might subsequently become divergent or vice versa. This has practical implications for the matching of data, as well as for the methodological integration, combining or mixing of data sets.

3. Multiple perspective qualitative longitudinal interviews (MPQLI): development of an analytical framework

In the next step, we integrate multiple perspectives and qualitative longitudinal approaches, paying special attention to data management and analysis. It is obvious that the quantity of data collected using such an approach is substantial, and its interpretation very complex. There is a potential danger of being intimidated by the sheer volume of the data (Lewis, 2007). Thus, the use of a QDA-Software is highly recommended. The challenge becomes one of realizing the potential explanatory power of the data-set (Thomson, 2007).²

In qualitative social research, different approaches to data analysis exist, but they share a common focus. The principal strategy in qualitative analysis is comparing and contrasting, finding similarities and differences. Bazeley (2013) summarizes the process of qualitative data analysis with the following formula: describe, compare, and relate. Themes are identified, content described and differentiated for groups, and then compared and related to similar cases (Kuckartz, 2014). The aim is generally to move from the specific to the more general, discovering patterns. Through a close inspection into specific cases and issues we aim to discover, explore and generate conceptual descriptions of phenomena (Rapley, 2014). This also holds for MPQLI research.

We now develop a schematic overview of analysis in (1) longitudinal and (2) multiple perspective interviews research to conclude with (3) an integration of both.

3.1. Analyzing qualitative longitudinal data

Social science methodology is mainly based on cross-sectional approaches. Longitudinal research brings new challenges, albeit employing the same principles used in cross-sectional studies. However, the analysis in longitudinal studies is more complex. Analyzing qualitative longitudinal data requires a cross-sectional and a longitudinal dimension. In a cross-sectional analysis, we aim to identify different narratives. Longitudinally we look at the development of a specific narrative over time (Kirkman, Harrison, Hillier, & Pyett, 2001; Thomson & Holland, 2003).

Research questions in a longitudinal study focus on change or progress. To identify change, we have to establish some preliminary criteria for making comparisons. Thus, frameworks for ‘locating meaning and tracking qualitative change are required’. This framework is first developed cross-sectionally,³ and is then adapted and extended during the longitudinal analysis.

To establish criteria for tracing change, we suggest four stages of comparison: First, we start by analyzing (a) individual interviews on a cross-sectional level, creating case profiles on every wave of data collection. These cumulative profiles summarize changes in circumstances and narratives. These profiles constitute the basis for comparisons within the entire sample and create (preliminary) coding schemes. By (b) comparing and contrasting cases with each other, we develop the coding categories further, sharpening profiles and grouping them into clusters; in other words, we establish a typology.

These cross-sectional profiles and typologies are the basis for longitudinal analysis. They offer a comparative framework across all participants and time points. In the longitudinal analysis, we again start with the (c) individual analysis followed by (d) comparisons between participants. In order to follow cases or themes as such, we have to return to the original data for the longitudinal comparison, first on an individual level and then by comparing individual trajectories across the sample. In essence, we first extend the individual profiles from step (a) by including a longitudinal component. In due course, we develop categories for change and the related process. These codes or categories are the basis for a comparison between individual trajectories, but they are adapted, extended or reduced during the analytical process. We can then create typologies of change and relate them with the typology produced in the cross-sectional analysis.

In short, cross-sectional profiling is followed by a longitudinal analysis. This makes change traceable. Table 1 systematizes this approach by illustrating the two dimensions of analysis resulting in four types of comparisons. It also identifies the respective (generic) aims of these comparisons in each step.

Table 1. Dimensions of comparisons in qualitative longitudinal research and related aims.

	Individual cases	Across cases
Cross-sectional	a	b
	• Case profiles • Provisional codes and categories	• Expansion of coding scheme • Criteria for comparison • Typology
Longitudinal	c	d
	• Individual trajectories • Provisional categories for change (Timing, structure and causes of change)	• Types of change and underlying dynamic • Criteria for longitudinal comparison • Clusters of types of participants, types of change, and process of change

3.2. Analyzing multiple perspective interviews

In an analogous way to the requirements of analyzing longitudinal data, multiple perspective interviews carry special requirements in data analysis and necessitate new layers of comparisons. They can take the form of one interview, interviews with respondents with similar characteristics (e.g. mothers), interviews with different ‘types’ of respondents (e.g. mothers and fathers) and interviews with related others as well as interviews from different relational units (see also Boeije, 2002). ‘Analysing multiple perspectives data requires a flexible shifting between taking the singular accounts as individual’s stories, analysing dynamics within and between different accounts, and trying to make sense of divergent and convergent data’ (Zartler, 2010, p. 178).

The levels of comparison can again be grouped along two dimensions: unit of analysis (individual vs. relational unit, e.g. couple or family) and within or between types of respondents. As a result, we identify four different steps in the comparative analysis. First, we start with (a) an in-depth reading and the development of categories for every individual interview. The focus is on identifying consistent, recurring themes to create individual conceptual profiles and develop a code tree and an inventory of central themes.

In the next step, we compare (b) individual accounts within one relational unit (e.g. couple or family). This allows for the generation of an inventory of central themes, a profiling of relational units from both perspectives, as well as reaching a better understanding of the interaction between the related persons. Finally, we move from the individual to an understanding of the relational unit from different perspectives. Comparing both partners’ accounts, we can come to different results than we might from each individual perspective. Furthermore, perspectives can converge or diverge. It is important to go beyond manifest statements and consider latent meaning, because surface accounts might converge where an in-depth analysis shows diverging meaning. Agreement or convergent data should not lead to complacency (Mccarthy et al., 2003; Sands & Roer-Strier, 2006; Zartler, 2010). This data should always receive a second look, for example with one team member taking the role of questioning the convergence (Zartler, 2010).

Sands and Roer-Strier (2006, pp. 242–248) identify five interpretative outcomes from this dyadic comparison:

(a)	Same story, same meaning: Interview participants tell similar stories, at times they even use the same terms, words, or metaphors when talking about the same issues. These narratives can be seen as convergent.
(b)	Same story, different interpretation is a hybrid between convergent and dissonant data as the content converges but the interpretation diverges.
(c)	Missing pieces: consist of information provided by one participant that is essential to understand the story and was not included in the counterpart’s description. These can be considered complementary findings, where different descriptions are integrated to produce a whole.
(d)	Unique information refers to a report that is provided by one informant only, but is not integral to understanding the full picture.
(e)	Illuminating accounts are accounts that are different but not contradictory.

Taking account of these potential interpretative outcomes helps in establishing profiles for relational units. Most importantly, the dyadic level of analysis can and often does alter the individually based interpretation of gathered qualitative data.

In a third step, we return to individuals in that we compare (c) individuals regardless of their relation with other respondents, but along other similarities like age, gender, similarity of experiences and so on. As a result, interviews can be grouped into clusters or types based on specific criteria. We look for regularities to identify conditions for patterns. This way, codes can be extended and a typology on an individual basis developed.

Most often however, the relational unit will be the central unit of interest in multiple perspective interviews. Thus, in a last step we compare (d) relational units with each other. This sharpens criteria and codes for characterizing couples and detecting patterns in relationships. Table 2 summarizes the dimensions and aims of the comparative steps described. The comparisons on these levels have different aims, are based on different kinds of questions, and consequently generate different insights.

Table 2. Dimensions of comparisons in multiple perspective interviews and related aims.

	Individual	Relational unit
Within group of similar respondents	a	b
	• Inventory of central themes • Case profiles • Provisional codes and categories	• Inventory of central issues within a relational unit • Profile of relational unit from two or more perspectives • Understanding interaction between related persons
Between groups of ‘types’ of respondents	c	d
	• Expansion of coding scheme • Criteria for comparison • Typology	• Criteria for comparison • Typology and pattern of relational units

3.3. Analyzing multiple perspective qualitative longitudinal interviews (MPQLI)

While the advantages of MPQLI may appear self-evident, to date there has been little research into the introduction of analytical strategies and challenges to systematically address the underlying complexity of such data. Two specific issues introduce intrinsic complexity within MPQLI data, namely the processing of social interrelationships as individual data, and the fluctuating narrative which is inherent within longitudinal data.

To develop an analytical framework for the analysis of MPQLI data, we start with a theoretical conceptualization of potentially relevant dimensions for the analysis. The multiple levels of analysis are positioned together across two dimensions: Unit of analysis (individual vs. couple) and time (cross-sectional vs. longitudinal). Research questions can thus be addressed on four different levels: individual cross-sectional, individual longitudinal, relational unit cross-sectional, and relational unit longitudinal. The potential research outcomes of these comparisons multiply. Table 3 brings the two dimensions, unit of analysis and time, together and illustrates some of the possible connections between different research aims and the appropriate level of comparisons.

Table 3. Dimensions of comparisons in MPQLI data and related aims.

		Unit of analysis
		Individual	Relational unit
Time	Cross-sectional	a	b
		• Case profile • Typology	• Profile of relational unit • Comparison of perspectives • Insights into interaction between related persons • Common themes and individual perspectives
	Longitudinal	c	d
		• Individual history • Change and process of change • Comparison between individuals • Typology of change	• Case history of relational unit • Dynamics within the relational unit • Development of perspectives in comparison

The analysis on an (a) cross-sectional individual level generates case profiles or typologies for individuals. (b) Cross-sectional data on the relational unit allows for comparisons of perspectives, insights into interactions, common themes and individual views. As a result of the comparison of perspectives we can identify the similarity or convergence of accounts, complementary/congruent views, contradictory accounts, and silences/gaps. (c) The longitudinal dimension adds information on change, process of change and individual history. As mentioned above, change could consist in change of narrative, respondent’s reinterpretation, or researcher’s reinterpretation (Lewis, 2007). To explore the patterns and processes of change and establish a typology of individual change, we also compare individuals (independently of their affiliation with a relational unit, e.g. fathers and mothers or younger age groups and older age groups). (d) For the relational unit, change could be identified in all members or only in one. By identifying the differences in change between unit members, we gain insights into the case history and dynamics of this relational unit and the development of related perspectives over time.

For multiple perspective interviews, Zartler (2010) recommended a stepwise process, beginning on an individual level, followed by a dyadic and/or relational unit level. In the case of longitudinal data, this stepwise process is extended to include a time component. Thus, for MPQLI, we recommend the following steps visualised in Figure 1. (1) At the cross-sectional individual level the data for each individual is analysed. One case represents one interview. The aim here is to create case profiles. (2) The cross-sectional comparison within a relational unit is based on the members of this group as unit of interpretation. (3) Relational units are then compared amongst each other to evaluate similarities and differences between couples and to create a profile of relationships. Then we move to the longitudinal dimension: (4) Initially, we explore individual change over time through a longitudinal comparison on an individual level. (5) Then we explore change and dynamics within a couple by comparing related respondents from a longitudinal perspective. (6) Finally, we compare relational units longitudinally to identify patterns that are group specific (e.g. couples with similar experiences or background) and to create a typology of processes within relational units.

Figure 1. Steps in the comparative analysis of MPQLIs.

Depending on the research question, some steps can take a more or a less important position within the analysis than others. Furthermore, these steps do not necessarily form a linear process. However, the steps as such will remain, even if the sequence or priority differs.

From our experience, these decisions necessitate a good understanding of the data. It is not necessary to compare everything with everything else, but it is important to have a systematic plan (Boeije, 2002). To utilize adequately the multi-layered data, a systematic plan with a clear outline of analytical steps taken, dimensions of comparisons and clear stated aims is crucial. Otherwise, the volume of the data may easily become unwieldy, and any analysis can lack in traceability.

4. Example from our own empirical research

In this section we illustrate the analytical steps and related insights we have described with examples from an Austrian multiple perspective qualitative longitudinal study with first-time parents.⁴ We conducted three waves of data collection with eleven couples in separate interviews: during pregnancy, six and 24 months after birth. In total, we collected 66 individual, problem-centered interviews (Witzel, 2000) between 2013 and 2015 in Vienna (Schmidt & Rieder, 2016; Schmidt 2017; Schadler, Rieder, Schmidt, Zartler, & Richter,  2017).

To exemplify the use of the analytical framework suggested, we focus on respondents’ perception and anticipation of task share in childcare before first-time parenthood, and any anticipation of changes over time as the child grows up. For illustrative purposes, we selected two couples with similar sociodemographic characteristics and views of gender roles and care work. Tina and Tom, Couple 1, Maria and Max, Couple 2, were about the same age (30 and 29 respectively 32 and 34 years old); all worked full-time (before and during the pregnancy). Although their positions at the outset were comparable, the couples developed differently. Going through the analytical steps proposed, we illustrate this change by giving labels to individuals and couples to make the change and added-value of MPQLI more concise.

(1) Cross- sectional comparison on an individual level

Comparing the four individuals in Wave 1, Tina, Tom, Maria and Max were very similar with regard to their perspectives on their planned arrangements for parental childcare: All four considered egalitarian distribution of care and household tasks essential and arranged their joint daily lives accordingly. They all planned to continue equally sharing tasks after the birth. There were financial, political, practical, ideological and professional motivations for this cooperative, shared responsibility.

Thus, in Wave 1, we characterized each of these respondents as egalitarian norm and (prospective) task share. The expectation and practice of these couples with respect to actual and anticipated task sharing were egalitarian.

(2) Cross-sectional comparison on a relational unit level

Looking more closely on the relational level, comparing Max with Maria and Tom with Tina, we found converging accounts in Wave 1. Couple 1, Tina and Tom both favoured a fifty-fifty arrangement with 12 months of parental leave for each of them, with the respective partner being full-time employed. In Couple 2, Max’ and Maria’s perspectives also converged but not as neatly: Both Maria and Max planned to take parental leave for at least one year each, with each of them working part time afterwards. However, Max subliminally questioned the feasibility of this arrangement from a financial perspective while Maria strongly argued in favour of shared responsibilities, saying ‘Why should I do all that on my own?’ (Maria, Wave 1). Within both couples, the partners converged in their political and ideological convictions that it was important to ‘create equal career prospects’ (Tom, Wave 1) for mothers and fathers and to regard fathers’ leave-time as an ‘extremely valuable experience’ (Tina, Wave 1).

The comparison of related individuals within the couples (comparing Maria with Max and Tina with Tom) showed, that both partners held similar role and task share expectations. Thus, we described the couples as convergent egalitarian norm and (prospective) task-share.

(3) Cross-sectional comparison between relational units

The cross-sectional comparison of Couple 1 and 2 showed that both couples agreed on an egalitarian task-share after birth. The justifications for this egalitarian approach were similar and mainly based on an egalitarian political ideal but depended also on finances, in the sense of optimizing their income generated during parental leave.

Thus, comparing couples and their arguments for parental leave arrangements identified Couple 1 and 2 as rational idealistic justification for egalitarian task-share.

(4) Longitudinal comparison on individual level

Now, we add the longitudinal dimension: Comparing individual case profiles over time showed that in Wave 2 (six months postpartum) Tom had enjoyed his daddy month (1 month after birth) he planned in Wave 1. He found it tough to re-enter his full-time job afterwards. In order to spend more time with the family, he had reduced his sports activities. In Wave 3, he had just started working after his parental leave (four months earlier than initially planned), which he considered challenging, but he was optimistic that things were about to become easier.

In Wave 2, Tina was enjoying her parental leave, being at (mental) distance from professional work. In Wave 3 (two years postpartum), Tina was working full-time and had changed her job due to problems with her employer. She had got used to Tom being on parental leave, and had the impression that ‘the workload is distributed in a very egalitarian way’ (Tina, Wave 3) with her partner.

Maria on the other hand felt ‘much more exhausted than [she] would have thought before’ (Maria, Wave 2). She was on parental leave in Wave 2 and 3, as her attempt to re-enter the labour market failed because she could not cope. Also, her expectations regarding egalitarian task share changed: ‘I don’t need to have a fifty-fifty allocation of housework (…) but I still insist on a little bit of equality’ (Maria, Wave 2). In Wave 3, Maria considered it ‘natural’ not being employed with a small child. She now understood this arrangement to be ‘ideal’ and considered herself ‘fortunate not to be forced to earn money’ (Maria, Wave 3).

Although Max took his daddy month immediately after delivery, he continued working. In Wave 2, he worked full-time and still considered taking parental leave, however, only for two months over the summer. Part-time work after a longer parental leave – as envisaged in Wave 1 – was no longer planned. Max felt ‘massively restricted’ (Max, Wave 2) because he had reduced his leisure time activities due to a lack of energy. Overall, he increasingly experienced a strong constraint in his quality of life. Max was happy to be able to provide for his family and to earn ‘substantially more money than my wife’ (Max, Wave 2), but still expressed his view that both partners should work 30 h (although he was in full-time employment also in Wave 3).

Comparing Max and Tom revealed that, despite starting from very similar positions in terms of their view of the ideal distribution of tasks and family generally, their views diverged more and more over time. In Wave 3, for example, they had opposing views regarding leisure time: Tom regarded the reduction in individual leisure time activities as an increase in quality time with the family. Max interpreted this as a decrease in individual quality of life that additionally provoked arguments with his spouse.

In sum, the longitudinal perspective illustrated different types of change: Whereas for Tina and Tom, we found relatively little change in their individual attitudes and behaviour, Maria’s accounts illustrated a narrative change as well as her reinterpretation of the importance of the egalitarian distribution of tasks. Her case perfectly exemplified re-traditionalisation. For Max, we also found a subjective reinterpretation of his contribution and parental leave. Thus, over time, Tina’s and Tom’s accounts could both be described as consistent egalitarian norm and practice, Maria’s as re-tradionalising and Max’ as relativist because he viewed the change in his attitude or the impracticability of his anticipated role as a result of unforeseen circumstances.

(5) Longitudinal comparison within a relational unit

Within Couple 1, both Tina and Tom agreed that the daddy-month gave them a perfect opportunity to bond as a family. Later on, their full-time work allowed them ‘not to worry about the income’ (Tina, Wave 3). Tina’s and Tom’s views on the share of household and child-related tasks converged widely cross-sectionally as well as longitudinally.

In contrast, Maria and Max showed considerable inconsistencies over time. For example, Maria (Wave 2) estimated that Max reduced his work hours to 40 percent – mentioning that he would claim it was 10 percent. Max himself said he had reduced his workload to 20 or 30 percent. Moreover, Maria complained about Max’ multiple time-consuming activities beyond his job and his family, while Max complained about a drastic reduction in his time for hobbies and other interests. Over time, Max’ and Maria’s views drifted apart. Approval of traditional gender roles replaced their shared ideal of gender equality in Wave 1. Almost inevitably, conflicts emerged: While Maria considered it increasingly comfortable to be a stay-at-home mum, Max in Wave 3 wished Maria would contribute to the family income.

On the relational, longitudinal level, Couple 1, Tina and Tom, can be described as ‘consistently convergent egalitarian’. Couple 2, Maria and Max, is ‘increasingly diverging from egalitarian to re-traditionalisation and relativism’.

(6) Longitudinal comparison between relational units

In the final step, we compared both couples from a longitudinal perspective. The considerable similarities in Wave 1 categorised both couples as ‘egalitarian norm and practice’. However, this changed substantially over time. Couple 1, Tina and Tom, showed a high congruence and accordance in relation to their parental distribution of care work, both over time and within their partnership. Many of their plans expressed in Wave 1 had been put in practice in Waves 2 and 3 – despite minor adaptations. Both shared a common ideal of having an egalitarian division of tasks.

In contrast, couple 2 developed substantial incongruences over time. Max strongly argued in favour of gender equality, equal opportunities and parental leave in Wave 1, but ended up being the provider for the family in Wave 3. Maria declared particularly high demands with regard to an equal division of tasks in Wave 1. In Wave 3 this conviction was replaced with the attitude of ‘benefitting’ from Max’ job and feeling ‘fortunate’ to be able to stay at home.

In sum, the multiple perspective and longitudinal research design allowed us to uncover the different related processes and changes. The analysis revealed that both couples’ constructions differed substantially over time although they seemed comparable in the first step. As a result of step 3, the label ‘rational idealistic’ changed to ‘implementing ideals’ for Couple 1 and ‘adjusting to re-traditionalisation’ for Couple 2 by adding the longitudinal dimension.

In these couples both the longitudinal and the multiple perspective approach not only yielded more information, but also partly altered the overall interpretation that would have been reasonable for any individual perspective in isolation. The MPQLI approach gave a more realistic reflection and nuanced understanding of the development of individual lives and sense making, but also of dynamics within a couple around the transition to first time parenthood.

5. Discussion

Multiple perspective qualitative longitudinal interviews (MPQLI) have great potential. They not only shed light onto complex interactional systems from the different perspectives of participants within these systems, but also allow researchers to explore dynamics and changes in individual lives and perceptions involved over time as well as within relational units. These approaches are highly desirable and promising but they entail several challenges. They are resource intensive. In data collection, they require respondents’ continued participation over the duration of the study which might be several years. The problem of panel mortality is intensified by the multiple perspective character as related individuals participate. In consequence this means, that if one person drops out of the study at any point in time the data collected previously and from related others is of limited value or cannot be considered.

Regarding the data, MPQLI create a potentially unwieldy amount of multi-layered and complex data. As such, due to a lack of systematic methodological reflexions on data analysis in particular, the potential of MPQLI is rarely fully exploited (Mccarthy et al., 2003). It is a common problem, specifically in qualitative research, that researchers often ‘describe at great length how their studies were carried out, but remain vague when it comes to giving an account of the analysis’ (Boeije, 2002, p. 392). A lack of explication reduces the potential for verification and the credibility of qualitative reports. Thus, the objective of our paper was to suggest a rigorous approach to the analysis of MPQLI data that allows researchers to systematise their work and increase the traceability of their results.

This entails the inherent problem of abstraction. Developing an analytical scheme without relation to a research problem has limited merit. It is obvious that any analytical strategy has to be adapted to the types of data and research interests in any given project. Nevertheless, the great potential of the presented approach lies in offering a systematic framework for the analyses of MPQLI data. The foremost benefit in the suggested analytical steps is the reduction of complexity. The sheer volume of data produced by MPQLI makes a systematic reduction and data management a vital first step in any rigorous analysis.

The framework we propose is constructed from the combination of the two dimensions which characterise MPQLI: Time and relational unit. The suggested approach is structured by a combination of cross-sectional and longitudinal comparisons on the one side, and individual versus within and between relational unit comparisons on the other. Consequently, we identify (1) cross-sectional individual, (2) cross-sectional within and, (3) cross-sectional between relational unit comparisons followed by (4) longitudinal individual, (5) longitudinal within and (6) longitudinal between relational unit comparisons (see Figure 1). Applying this analytical scheme, we gain (a) case profiles and typologies of individuals, (b) profiles of relational units with a comparison of perspectives, as well as (c) individual histories of change and processes of change and (d) dynamics and development within relational units (see Table 3).

We recommend this framework independently of differing methodical approaches in data analysis. We assume that it will apply to various qualitative data analysis approaches as long as these are based on comparisons.

A systematic, stepwise approach to the analysis of MPQLI is necessary to handle the large amounts of data such methodologies generate. This stepwise approach necessitates rigorous planning, preparation and discipline. The six steps outlined can vary in their relative prominence or their sequence depending on current research interests. However, we are convinced that the steps as such will still offer a helpful framework for a broad variety of applications. The steps cannot always be divided neatly, and researchers should make notes when something is uncovered that may relate to another step. Nevertheless, the focus always, where possible, remains on one step at a time and the temptation should be resisted to conflate different dimensions. In following these small, but vital directions, future researchers can utilise their data to its full potential, and benefit from the great explanatory power that an MPQLI research approach provides.

Disclosure statement

No potential conflict of interest was reported by the authors.

Notes on contributors

Susanne Vogl is a Post-doc Researcher at the University of Vienna. Her research agenda focusses on research methodology. She strives for improving existing methods in social science and developing techniques and methods further. Susanne Vogl’s substantive fields of research include the sociology of deviance, children and young people, family and the life course.

Ulrike Zartler is an Assistant Professor of Family Sociology at the University of Vienna. She is a childhood and family researcher, specialised in qualitative research methods, and has used a multiple-perspectives approach in numerous studies. Her main research areas cover divorce and post-divorce families, family transitions, children’s rights, and legal aspects of childhood and family.

Eva-Maria Schmidt is a PhD Researcher at the University of Vienna. She studied Sociology and European Ethnology. Her research interests are parenting, fatherhood and masculinity, gendered transition to parenthood, cohabitation and marriage, work and organization, parental leave, biographical research and qualitative methods in the social sciences. Recently, she completed her work on her PhD thesis on parenting between care and career.

Irene Rieder is a researcher at the University of Vienna. She holds a Masters degree in Sociology. Her research focus is on family sociology, transition to parenthood and qualitative methods.

Notes

1. From a positivistic stance, multiple perspective interviews could also be used for cross-validation of information. However, this is usually not an approach taken in qualitative research.

2. An additional strategy of dealing with the amount and complexity of the data could also be to employ a mixed methods approach. Qualitative data can be analyzed with a combination of qualitative and statistical procedures. If appropriate, statistics can support, corroborate qualitative observations of change (Saldaña, 2003). Amongst others, Miles and Huberman (1994) and Saldaña (2003) advocate for selected studies the transformation of qualitative data into proportions. It can facilitate charting category frequencies and discern change and its underlying dynamics. Word counts can be a first step to look at very complex data and explore central themes (Bernard & Ryan, 2010).

3. This approach obviously follows the logic of the project: a longitudinal analysis is only possible after the completion of all data collection whereas cross-sectional analysis can be conducted earlier, namely after every wave. Furthermore, ideally the (preliminary) results from each wave should inform the next data collection round. However, for practical reasons, it is not always possible even to skim all the data from one wave before the next wave commences. This also points the lack of analytical closure (Thomson & Holland, 2003) mentioned above: the next wave can always challenge the existing interpretations.

4. This work was supported by the European Union’s Seventh Framework Programme (FP7/2007-2013)under Grant number 320116 and by the University Jubilee Foundation of the City of Vienna (Hochschuljubiläumsstiftung der Stadt Wien) under Grant number H-284605/2015.