1The object of study of the social and human sciences is by definition a complex one − the human being, a conscious being possessing values, symbolism, language and a social identity. With such a singular subject of study, whether the research is quantitative or qualitative, synthetic or systemic, retrospective or instantaneous, observation is hampered by the subjective nature of the facts and discourses it sets out to record and understand. In the first place, the way respondents perceive the data collection can influence their responses, especially with long, formalized data gathering operations like those used for collecting quantitative event histories. Objectivation is also a constant concern for the researcher, who must distance the subject studied from his/her own representations and bear in mind “the false neutrality of the techniques” used (Bourdieu et al., 1983). In a way, the matter of perceptions − the players’ and the observer’s − is at the centre of any epistemological (and hence methodological) reflection.

2The quantitative life course surveys, by opening up to qualitative observational input (GRAB, 1999), gave rise to multidisciplinary collaboration and thus made more allowance than other data sources for the possibility of exploring the subtle links between factual data and perceptual information. Because of their biographical nature, these quantitative and qualitative surveys require a more or less calibrated reconstitution of an event history, which is recounted in narrative form, as a “story”.

3In this book, which is about temporal dynamics, these individual data will be useful for examining the variable ways in which events fit into time. This first chapter will take stock of the use of perceptual elements for locating events in time and will present in a wider view the avenues to be explored with regard to observation and analysis.

4So this first chapter addresses the various aspects of the use of perception, taking (unlike subsequent chapters) a range of examples with no particular thematic approach. It points out that, beyond the combination of qualitative and quantitative, these perceptual elements cannot be addressed by an exclusively qualitative investigation or analysis. The chapter also addresses the importance of collecting and making quantitative use of perceptual elements beyond issues of temporality.

5After considering the question of retrospective interpretation of facts, a major preoccupation in surveys that record retrospective perceptions, the chapter considers what lessons are to be learned from comparing factual and subjective data about the same life stage. Light is shed both by comparing modes of observation (interview and questionnaire) and by including two types of question in the same questionnaire1. This shows how introducing perceptions makes it possible to re-examine prevailing statistical categories and also improve them. In particular, considering states leads one to discuss the crossing of thresholds, since it proves crucial to identify when a state began and when it ended (at what point was that state achieved? How does one move out of it?). The question of identifying a network as also addressed, since this is heavily reliant on the respondent’s subjective assessment.

6Moving on from the focus on observation, the use of perceptual elements is addressed in detail, illustrated with a presentation of the measurement and use of perceptions in economics. This section concludes with a recent experiment comparing factual event histories with the way the respondents themselves divided their life course into stages.

7The chapter gives a number of examples based on data from the Biographies et entourage survey (see Appendix, Box 1), one of whose aims was to suggest an approach to data gathering which is at the junction between the quantitative and the qualitative (Lelièvre and Vivier, 2001).

1. Comparing Facts with Perceptions

8Before the collation of data strictly speaking can be considered, this chapter needs to address the question of subjectivity (Blair and Sudman, 1996). The narrative nature of a life event history gives the data, even when they are purely factual, a subjective dimension that is often pointed out as a weakness in terms of data “quality”. This particular attribute of life event histories is the first thing that needs to be addressed. It has stimulated a lot of discussion, and is not a feature of life histories alone. Any factual questioning prompts an exercise of judgement; what the researcher records is never completely determined by what the respondent can or wants to say.

1.1. Retrospective Interpretation

9Because prospective surveys2 are costly (time required to obtain life event histories, demanding monitoring logistics), most quantitative life course surveys are retrospective. Methodological work conducted at the time of the Triple Biographie survey (GRAB, 1999, p. 59-73; Poulain et al., 1991; Courgeau, 1985) showed that life event data gathered with appropriate tools are reliable. There may be errors concerning the exact dating of events, but provided one checks that the chronological order of events is remembered correctly, “memory appears to be reliable where the analysis so requires” (Courgeau and Lelièvre, 1989)3. The effects of memory − a subject extensively covered by Auriat (1996) − are not necessarily a problem, especially where survey tools have been designed to ensure that chronological order is adhered to and event histories are consistent.

10However, it will be useful here to make a brief multi-disciplinary review of research that uses subjective reconstitution, possibly idealized, of a life event history considered as a succession of perceptions.

11Bourdieu (1986) describes the limits of event history reconstitution in terms of a “biographical delusion”; Bertaux (1997) in terms of a “biographical ideology”. This type of reconstitution is what happens with an introspective literary autobiography underpinned by a precise line of argument, as in Rousseau’s Émile, the first autobiography of this type (May 1762). In a life course survey, this tendency is better controlled because of the interview’s spontaneity and its dialogic aspect. Nonetheless, the many mediations between the respondent’s original experience and the account they subsequently give of it will “smooth out” the life course. Bertaux explains this smoothing process with reference to the different levels of subjective mediation involved. One is the “introspective totalization process” (Bourdieu, 1986), which leads the subject to retrospectively attribute a meaning to an event and re-situate it in relation to other experiences in their life.

12Leclerc-Olive (1997) identifies at least three phases of mediation: the moment of the event itself, the long period of making sense of the event, and the moment of relating the event, which commits the narrator to their version of the event. It is these different phases of subjective construction of the narrative that are addressed in the “textualist” approach to life history accounts (Desmazière and Dubar, 1997a). This approach mainly focuses on analysing the discursive form of the narrative with reference to what it tells of the effects of the mediations that have shaped it. Leclerc-Olive compares the narration of a personal life event with the trace left by a historic monument which “celebrates a past fact, in a style and with the meaning of the moment of commemoration, and [which we] look at today, marked by our own history”. The linguistic and even psychoanalytical dimension of life history narratives is also prevalent in clinical sociology which studies the interactions between recurrent, collective psychological and social phenomena such as the “class neuroses” identified by Vincent Gauléjac (1987).

13Meanwhile, the partisans of the “realist” approach to life history narratives remain attached to the postulate that “a person’s history possesses a reality prior to and independent of the way it is related” (Bertaux, 1997, p. 33). The many subjective mediations do not affect the objective succession of events and situations that punctuate the life course. On the contrary, they make the diachronic structure of the narration emerge even more clearly from a comparison between different narratives in a corpus. By comparing the many dimensions of the narratives one can set aside the parts that derive from retrospective colouring of events, and take the factual elements to identify the common core of the different life courses studied − their social dimension (Bertaux, 1997). It is precisely this objective succession of events, regardless of the perceptions that permeate the discourse, that a quantitative life course survey seeks to grasp.

14The quantitative life course surveys very often benefited from additional data gathering (GRAB, 1999, p. 20). Mainly conducted in parallel or a posteriori, the additional surveys were designed to complement or add detail to the data collected on certain themes of the survey, or to gather contextual data. Since then, especially for qualitative explorations, some data gathering systems have conducted in-depth interviews at the stage when the event history questionnaire is being designed4 (Lelièvre and Vivier, 2001). To pursue this matter of perceptions further, the information gathered by the quantitative and qualitative methods must be compared. This approach is now tried and tested (Battagliola et al., 1993; Lelièvre and Vivier, 2001); the lessons from this work prove essential for designing quantitative investigations.

15The section below takes two examples where perception and factual evaluation do not necessarily match. What should be done with the deviations? Do such discrepancies add to the fuzziness or do they enable us to improve the analysis?

1.2 Hébergement (“Housed by Family or Friends”): Observation and Measurement of a Perceived State

16Fuzzy-contoured residential status, which is considered more specifically in Chapter 4, is generally associated with housing insecurity. It is an issue that can be viewed very differently by public authorities, researchers and those directly concerned.

17The survey respondents do not care what the interviewer wants to know; they have their own perception of the facts. As the examples below show, their viewpoint does not systematically lead to an underestimation of the number of people being housed by family members or friends, but it sheds a different light on the presuppositions about housing insecurity that the idea evokes, and sometimes contradicts the researchers’ criteria. Some criteria defining “housed by family or friends” (“hébergement ”) stipulates that host and guest must share the home, and that it must be the host’s main residence, and that the guest is not the official occupant. Being housed by the employer is not included under hébergement (Grafmeyer, quoted by Bonvalet and Lelièvre, 2000).

18Also, when the public authorities want quantitative data on this housing category, they regard it as a crisis situation requiring action on their part. The examples below show how necessary it is to draw up a set of specific criteria to clarify the link between housing insecurity and “housed by” status, as relations of dependency between individuals are not necessarily sufficient.

19André is living “at his sister’s place” for a year. Put this way, one might assume he is living with his sister, and that is what the respondent implies. But without further precision one cannot exclude the possibility that André is living in accommodation lent by his sister while she is living elsewhere.

20The first observation is that for the respondent, “being housed by” someone does not necessarily imply living with them (André feels he is “being put up by” his sister, even though he is in main possession of a dwelling that has been lent to him). Respondents describe their residential situation by reference to their own assessment of the help received or given.

21Conversely, a situation classed statistically as being “housed by family or friends” may not be perceived as being helped out by them, as in the examples of Jacqueline and Anna below.

22Since her marriage, Jacqueline has lived in a succession of civil service dwellings as her husband has moved from job to job as school principal and then head of an education authority. Not for an instant does she regard herself as being “put up by” anyone, yet between her father’s house and her husband’s she has never been holder or joint holder of any title to her accommodation.

23Anna, as an au pair girl, has a room in her employers’ flat. Being housed by her employer, her accommodation is not classed as hébergement. Having become friendly with the family, she continues to live there when no longer working for them; where before she was housed by her employer, she is now living with friends − hébergement. What is interesting here is that Anna thinks of this as an improvement in her situation and does not feel that the term hébergement applies, whereas in legal terms her situation has become more insecure as it depends solely on her friendship with the family.

24The lessons drawn from this examination of perceptions are that a given status can evoke different associations, both for the persons concerned and for the researchers. Hébergement has been studied in terms of housing unit occupancy, in studies of family solidarity and in research on sociability. It can also be viewed in economic terms, as a resource or a cost. And it is a frequent matter of concern to the public authorities from the standpoint of taxation or social insecurity.

25These different perspectives call for the use of different identification criteria (this is addressed more fully in Chapter 5). Identification criteria will also depend on the duration of the situations observed, and at what moment in an individual’s life course the study takes place.

26Thus the perceptual elements indicate the need to adapt identification criteria to the purpose of the study. The different facets of the hébergement status cannot be reduced to a single, unequivocal definition.

1.3. Working Career Versus Main Occupation

27Identifying individuals’ original social class is important in a number of research fields such as studies of social or occupational mobility and of socialization.

28The study of social mobility is based on comparing the respondent’s occupation with their father’s (Thélot, 2004; Vallet, 1999). The question is what criteria to use to identify the father’s socio-occupational category, allowing for possible career development. In a critical analysis of surveys used to assess social mobility, Merllié (1994, p. 173) points out that the moment at which the father’s occupation is to be identified is not stated, or not clearly stated, and may vary from one survey to another5. This means that the data do not compare the situations of “father and son at the same moment in their careers”, and that “intra-generational mobility always blurs the outlines to some extent when looking at inter-generational mobility”. Conventionally, and in particular in studies by INSEE in France, surveys ask for the father’s occupation at the time the respondent left school. Nonetheless, the ideal would seem to be to have information on the entire working careers of both father and son.

29In the Biographies et entourage survey, the whole working career of the parents was recorded6. It traced each person’s various occupations and also any periods when they were unemployed or out of the workforce as a student, voluntary worker or prisoner of war, etc. After reconstituting the father’s and mother’s occupational careers, the respondent was asked to give the “main occupation” of each parent as they perceived it, the question used being

• Q. To sum up, what do you think of as being their main occupation?

30This reveals how each respondent thinks of their parents’ occupation, and puts this in perspective with the working career as a whole.

31The first utilizations of data on the respondents’ mothers’ “main occupation”7 reveal that of the 52% of mothers who were declared as “housewives”, a majority (54%) had had at least one paid employment during their lives. Of the 48% of mothers stated by the respondents to be in employment, 32% had been housewives for some period of time and could have been so declared.

32To examine more precisely how “mother’s life course” relates to “mother’s occupation as declared by the respondent”, in those cases where the respondent could have stated either “in employment” or “housewife” (1,238 of the mothers in the survey), the researchers compared the length of time spent under each status8. They found that 79% of the mothers declared as “housewives” had in fact been housewives for longer, in all, than they had been in employment. The same observation holds for women declared as “in employment”: 62% had spent more time in employment than as housewives. However, more than a third of these women (37%) had spent more time as housewives than in employment. In these cases, the respondent had focused on their mother’s paid employment rather than on her time as a housewife even though this had lasted longer.

33These few very preliminary figures show that respondents make assessments when answering such questions. If other parameters and connections between “working career” and “declared main occupation” are taken into account, a better understanding can be achieved of the principles governing respondents’ choices between two alternative declarations. At what moment in the respondent’s life course did the parent have the occupation declared − during childhood or adolescence? Do declarations vary according to the social hierarchy or prestige of the occupation? Do they vary according to the respondent’s social category or gender? We see here that it might be useful to compare the parent’s occupational history with their main occupation as perceived and reported by the respondent. The two types of information can be compared, and also used for additional purposes.

2. Perceptions Help with Quantitative Categorization

34At this stage it will be useful to show the informational potential of subjective elements. The examples below show how this kind of information makes it possible to “open up” the defining criteria and so identify states or categories that the nomenclatures make no provision for. This can pave the way for reviewing and improving the nomenclatures.

2.1. How Qualitative Exploration Improves Statistical Categorization: Female Heads of Household

35The quantitative research process sometimes produces theoretical categories that are debatable in the light of the social complexities they are supposed to cover. As regards recording life event histories, studies by Didier Demazière and Claude Dubar (1997b) among others have shown that individuals fall into categories that do not always match the theoretical occupational identity categories usually employed by labour sociologists. Working directly on respondents’ discourse in their life course narratives, they have revealed endogenous categories that are more complex than, and clearly distinct from, the theoretical panoply of their discipline.

36Qualitative methods are offering an increasing number of suitable tools for capturing and processing individual perceptions. Rather than rejecting these tools, the quantitative approaches should take inspiration from them or at least incorporate them into their own approach. Below we illustrate how a static category, a state, if considered in the context of an event history revealed through qualitative interviews, can be used to review the categories employed in quantitative analysis.

37A study by Christine Tichit (2002) in Cameroon was designed to look at the question of interpreting the figures for percentages of female household heads in Africa. In the literature then available on the subject, these percentages had been used to back up a variety of different and sometimes conflicting arguments. Some saw them as indicating women’s increasing independence (Bisillat, 1996), others as evidence of the destructuring of the family (Vimard and N’cho, 1997) others again as evidence of increasing insecurity in the recession that has spread across Africa over the past twenty years (Bessis, 1996; Marcoux, 1997). In fact, the various interpretations were based on a range of different definitions of the status and function of a “household head”. The statistical definition, designed to identify all households and all their members for census purposes, applies simple operational criteria, far more limited than those used in interpreting the percentages of “female heads of household”. Moreover, the analyses were extrapolated from minimal cross-sectional data from censuses, population surveys, consumer surveys, etc.

38To get beyond these contradictory interpretations and test the category definitions, women meeting the statistical definitions of “household head” were interviewed in a longitudinal survey to link the event history described by the cross-sectional variables (cumulative fertility, occupation, etc.) with their residential history and in particular, how long they had been living independently. Life course narratives were collected and alongside this, factual events were systematically identified by methods directly inspired by the quantitative life course surveys (Ageven record sheet)9, which made it considerably simpler to analyse the explicit and implicit content of the narratives. The different types of female life courses that emerge from this, and especially from the respondents’ discourses about their life courses, shed new light on the question of female heads of household.

39The result is a deconstructed and reconstructed category that distinguishes between female household heads and single mothers. There is a considerable overlap between the two, since over 80% of female household heads in Cameroon are single mothers. But whereas single mothers are often thought of as being always household heads, the narratives show that this is not necessarily so. Some women were already single mothers before they were household heads. In some cases they and their children lived in other households where they were very far from being the head. But single mothers living in other households are statistically invisible in censuses and in most surveys. Their existence can only be revealed from a detailed analysis of household composition.

40Quantitative data (in this case from the Demographic and Health Surveys: DHS) make it possible to grasp these dependent single-mother situations provided the kinship links between household members are processed in detail. These results therefore prompted a return to the available data to assess the demographic significance of “dependent single motherhood” by identifying secondary nuclei in households covered by DHS surveys.

41The first unexpected finding was that in fact only one-third of single mothers in Cameroon are also household heads. Referring back to the perceptions and life experiences of single motherhood, it appears that these women’s dependency or independence is almost inseparable from the question of economic activity. Being economically active or having a source of income is a precondition for becoming a household head. That is the basic difference between single mothers who live in other households and those who are household heads − a situation they experience as offering more independence and economic freedom.

42Female heads of household who have the means to be residentially independent are thus the least insecure category of single mothers. The conclusion from this is a more nuanced view of the insecurity of female household heads since (i) only a third of single mothers can even afford to live independently, and (ii) female household heads are only the tip of the iceberg as regards single motherhood.

43This detour to look at single motherhood reveals the limited usefulness of the “female household head” category as an indicator in this context; single motherhood cannot be reduced to the matter of the woman’s residential independence and proves to involve an unexpected economic aspect. Clearly, the blind alleys encountered in analysing states (represented here by one of the classic categories in cross-sectional quantitative analysis) can benefit from using life event history data that make it possible to identify how these states were constructed over time. They can also benefit from respondents’ own perceptions – explored qualitatively here − which make it possible to review the quantitative data and enrich the analysis by taking account of the full complexity of the states in question.

2.2. Integrating Perceptions to Grasp the Diversity of the Situations Encountered: When Is a Couple not a Couple?

44In the Biographies et entourage survey, the questionnaires were restricted (except for the summaries, to be discussed in a later section) to fact gathering, to the exclusion of retrospective interpretations such as remembered motivations. But while recording perceptions is not here aimed at revealing cause and effect, it does make room for hitherto unexplored categories. As a rule, people’s marital lives are addressed through the event history of marriage or live-in coupledom. However, taking a life course approach and reconstituting the conjugal history in order to understand the interactions between coupledom and occupational and residential histories, the multiplicity of types of union and couple situations emerges (see Chapter 2).

45With such diversity it is necessary to envisage other modes of coupledom alongside partnerships institutionalized by marriage or formalised as cohabitation.

46The Biographies et entourage survey included four questions to pin down marital status at the time of the survey and any other unions in the respondent’s life history.

• Q. What is your current marital status?
  • Never-married
  • In first marriage
  • Separated de facto
  • Legally separated or divorced
  • Remarried
  • Widowed
• Q. Whether married or not, are you currently living in a couple?
  • Yes – Yes part-time – No
• Q. How many times have you been married?
• Q. Apart from spouses, how many people have you been in a couple with for more than a year?

47Identifying a relationship as “being in a couple” gives a common denominator and hence a degree of homogeneity to the various unions recorded. But the degree and type of formalization varies between marriage, cohabitation and the more subjective sense of “being in a couple” with someone. Left to the free appreciation of the respondent, the idea of “being in a couple” is not subject to any criterion predefined by the researcher10. All that counts is the respondent’s perception: if he or she considers that a relationship is or was coupledom, that relationship is counted.

48This approach addresses the reality of life courses more accurately. It goes beyond simply recording events and states that are objectivized by clearly defined markers like marriage and cohabitation. It takes into account events and states that are equally tangible but whose contours are fuzzier and more subjective. However, the stumbling block here is a differential treatment of time and dating methods.

49In the Biographies et entourage survey, the start of married or cohabiting unions is identified, as usual, from an event that seals the union − marriage, moving into a common home, etc. But the start of an unformalized union cannot be dated from any unequivocal, clearly identifiable event so researchers must rely on the subjective judgement of the respondent − the date which, from their point of view, marked the start of the relationship, be it the first encounter, the first sexual intercourse or any other marker or set of markers which, for the respondent, founded the relationship as a couple relationship.

• Q. Marital events and dates

50Although this is an important first step, there are two arguments for recording the “start of the couple relationship”, whatever the type of coupledom in question. The procedure adopted here introduces a degree of heterogeneity in the ways the duration of a relationship is measured, as the start date varies depending on the type of union: the wedding, the start of cohabitation, or the start of the love affair. But it may reasonably be thought that few couples marry or move in together as soon as they have met. The couple relationship is thus very likely to have an impact on other spheres of the life course before the event taken as the start date, in formalised and unformalized unions alike. To achieve more uniformity and relevance in the information gathered, it seems more consistent to collect the “date the relationship started” whatever the type of union in question. The growing number of unformalized unions makes this all the more useful. The landmarks the respondents use themselves to identify the start of their relationship could therefore usefully be made explicit.

2.3. Reconstituting the Circle of Close Friends and Family

51To reconstitute the respondent’s contact network, a series of people are systematically entered and designated according to type of relationship with the respondent: family (conjugal ties and blood ties) and residential connections (all persons with whom the respondent has lived for at least a year). But retrospective assessments are also made, to extend the contact circle to others who have played an important role in the respondent’s life. This means retrospectively indicating all those who make up the circle of close friends and family − an additional step in taking account of respondents’ perceptions.

52Initially, these questions were explored in surveys designed to reconstitute social networks. The Proches et parents survey (Bonvalet et al., 1993) indicated those within the network whom the respondent counted as proches (close friends and relations). That survey was conducted by INED in 1990 and had three aims: to extend knowledge of the extended family, to explore the affinity network and to study the social customs of the network of family and friends. In the questionnaire, the respondent first described the help given and received (for studies, housing or employment) through their life course, addressed the universe of “close friends and relations”, and then information was gathered for each member of Ego’s family and their spouse: their age, place of residence, etc. To explore the affinity network, the survey’s designers decided to ask respondents to list their “close friends and relations”:

• Q. Now we are going to talk about those people you regard as close friends and relations. Can you list them, whether friends, family members or in-laws?

53They can thus identify three types of network: the extended family, close friends and relations, and the mutual help network. The three do not necessarily coincide. These questions proved very useful (Bonvalet et al., 1999; Bonvalet, 2003) and the highly informative analyses of them suggest it would be useful to make longitudinal studies using this kind of question based on subjective criteria.

54Here we examine how questions based on subjectively perceived criteria can fit into in quantitative life course surveys,

2.4. From “Mother” and “Father” to “Mothers and Fathers”

55In the Biographies et entourage survey, the “parents” category, usually considered to include only the biological “mother” and “father”, soon proved inadequate for describing the real life courses encountered. The first versions of the questionnaire simply asked for the dates of birth and, if applicable, dates of death of the “parents”, without specifying anything about that term. This implicitly shows the pre-decided, unambiguous type of reply that was expected. But the first tests revealed that more than one set of parents could emerge (for example adoptive parents who succeeded biological parents who had died as deportees) and that neither set of parents was more important or legitimate than the other. No relevant choice could be made between them in a life course survey covering the respondent’s contact circle. In the next test, the question was broadened to all those “who have played a parental role for you”. With this extended scope, adoptive parents could be taken into account and also any additional or alternative parental figures. In fact there proved to be a significant number: of 116 respondents, 37 mentioned at least one such person who had played a parental role in their life. However, this new approach also revealed the complexity of the notion of a “parent”, between a biologically and/or socially instituted framework on the one hand and individual experience and/or perceptions on the other. Thus while a step-parent could be included as playing a parental role, not all respondents with a step-parent mentioned them in this context, even when they had lived with them for many years. And other figures who are not usually thought of as playing a parental role were mentioned, widening the universe of parental referents on an elective basis to include grandparents, uncles, aunts, neighbours, nannies, employers etc.

56In the definitive version of the questionnaire, the “parents” category explicitly combines different types of tie. Some are biological, legal or social (biological parents and, if applicable, adoptive parents, step-parents or guardians). Others are more elective, depending on the subjective experience of the respondent (“people who have played a parental role”) and are identified by various open questions designed to grasp these types of relationship (Lelièvre and Vivier, 2002; Lelièvre et al., 2005; Lelièvre et al., 2008). The questionnaire records the life event histories of the biological parents, adoptive parents and step-parents in turn, and then asks:

• Q. Are there any other people who have played a parental role for you?

57Early analyses suggest that many respondents have had parental or educational referents other than their biological parents by the time they reached 15 years of age.

58According to the table 1, among the population surveyed 21% mentioned at least one person who had played a parental role11, a total of 803 persons being cited as having played such a role. Most are women (68%) and 32% are men. Most (87%) are related to the respondent. Nonetheless, there is a wide diversity of ties covering three generations:

• 45% are of the grandparents’ generation or even the great-grandparents’;
• 25% are of the parents’ generation: uncles, aunts, step-parents;
• 8% were of the respondent’s own generation: sisters, brothers, cousins, etc.;
• 13% of the parental figures had no kinship ties with the respondent. These were nannies, employers, employees of the parents, friends of the family, primary school teachers, neighbours, etc.

Table 1. Proportion of Respondents Having Alternative Parental Figures, Educational Figures or Other Important People Before Age 15

2.5. From Parental Universe to Important Persons: a More Difficult Extension

59The aim of extending the investigation to a more variable close contact circle (as appropriate for each survey), exploring the existence of other figures who are neither explicitly nor systematically identified in the questionnaire proves more difficult.

60Once the data have been gathered on figures of the close contact circle fixed in advance, the respondent is invited to list other people who have played key roles in their life:

• Q. Apart from these people12, are there others − friends, family, in-laws, colleagues, etc., who are or have been important for you, positively or negatively?

61This time no precise role is suggested, the only criterion is that the person is or was “important”. The finding, from 2,830 questionnaires, is of wide variety in the responses, in a much wider and less well-charted universe than that of parental referents alone. The variety is of several kinds: the number of people mentioned, their identities and their roles vary considerably. The question then is to see whether some regularity can be identified within the variations. Other issues also emerge:

• While the aim was to open up the close contacts universe to people who had not already been mentioned, some figures already mentioned were mentioned again. This more open area of questioning seems to be used to qualify, especially in affective terms, people “simply described” earlier (a harsh step-father, an adored spouse, etc.);
• Symbolic figures (God, Tintin, General de Gaulle), moral role models, spiritual guides and imaginary figures appear.

62Here the tension between individual perceptions, the quantitative approach and retrospective elaboration of meaning emerge clearly. Nonetheless, exploring the people mentioned during childhood and adolescence sheds some interesting light: in the first place, 67.5% of the 2,830 respondents mentioned at least one person or group of people who played an important part in their lives during childhood or adolescence13. The groups mentioned are mainly peers (bunch of friends, the choir, the basketball team, etc.) and most of the individuals mentioned are either members of the family circle (48%) or friends (38%). They are generally older than the respondent (63.5%). The circumstances or the reason why they are mentioned suggest that some played an affective or educational role. Then there are the more elective figures (13%) who also cover educational, parental or professional connections (foster family, school monitor, primary school teacher, etc.), those with a contractual connection (family servant, employer, etc.) or with a more symbolic role (e.g. godmother, monk, church minister).

63It is obvious here that while introducing freely declared categories provides information, it is hard to make statistical use of the data, even with a large sample. However, by analysing these fragile data by inter-individual comparison it will doubtless be possible to identify some of the mechanisms of subjective evaluation.

3. Measuring Perceptions in Economics: Subjective Poverty and Well-Being

64At this stage the different terms of the association between quantitative and qualitative have been described, and perceptual elements that do not necessarily emerge from the qualitative questions have been taken into account.

65Data collection apart, perceptual elements have recently made a remarkable breakthrough into that hardest of soft sciences, economics. In fact the question was dear to the founders of modern economics in the 18th century (Smith, 1759), but that tradition had virtually disappeared in favour of a strictly materialist approach in which money was supposed to be the sole quantitative variable. Interest in subjective variables is even beginning to emerge in the more specific field of development economics. Clearly, this had been due to the difficulty of obtaining socioeconomic statistics in countries where statistics is structurally deficient, so that economists were seeking a substitute for the existing information systems, the new approach would have been doomed. But seen as a complement to hard statistics, the new perspective seems far more promising. In situations where it seems most inappropriate to apply as such the categories developed for Northern countries, and where the repeated failure of development programmes has major consequences, it is high time to start finding out what the people concerned really think and take their viewpoints, aspirations and perceptions into account.

66Happiness and Economics. How the Economy and Institutions Affect Human Well-being (Frey and Stutzer, 2002): the title of this book seems at first sight so incongruous that it makes one think of some popular psychology self-development book, tucked in between How to Conquer Shyness and Infallible Tips for Career Success. The radiant smiles of the two Swiss authors in the photography on the back cover seems to confirm that they are delivering the recipe for the famous Swiss happiness to the world at large. But the publisher is the very serious and respected Princeton University Press. The authors (two respectable economists at the University of Zurich, who have published articles on these issues in leading international journals) acknowledge that the very term “happiness” (a feeling, after all) fits ill in the category of scientific concepts. And yet this book reveals a fascinating strand in economics literature that is little known, especially in development economics.

67The authors first question the conventional view of utility in economics and discuss questions of measurement. They address the most conventional economic factors in happiness and then take a more original stance, examining the effects of politics on happiness. They produce some particularly interesting results regarding several fields of economics. First of all, although income counts, it is not the ultimate measure of happiness. In countries where long data series are available (Japan and the United States since the 1950s), aggregate data show a sharp rise in per capita incomes, whereas the satisfaction index has remained constant. How is this to be explained? The apparent paradox disappears if one accepts that happiness is socially conditioned and is fundamentally dependent on expectations.

68More destabilizing for mainstream economic theory, unemployment emerges clearly as a source of unhappiness, because at equal income, the unemployed are less happy than the rest. That runs counter to the idea of a category of willing unemployed who profit opportunistically from an over-generous system of unemployment benefits. Lastly, from their Swiss data, the authors point out that the development of democratic institutions and participation in politics are decisive factors in happiness. For some, some of these results may seem obvious, but in fact they seriously upset the science of economics because they had never been part of the conceptual framework and, above all, there had been no data to validate the hypothesis.

3.1. A Promising Field of Investigation: Poverty and Subjective Well-Being

69In developed countries, the economic literature on subjective well-being and happiness is a fast-growing genre. Economics is taking over a field that had hitherto been addressed only by psychologists and sociologists. The awarding of the 2002 Nobel prize for economics to Daniel Kahneman for his work in this field (see Kahneman, Diener and Schwartz, 2003) shows that this research field is regarded as a legitimate sub-discipline of economics and that in fact it is not as recent as all that. Some renowned economists predict that it has a rosy future ahead, in its own right (“The scientific study of happiness is only just beginning. It should become a central topic in social science”, Layard, 2003), but also for exploring less well-marked avenues of research currently bubbling with new ideas, such as the study of social interactions (“Experimental and subjective data will have to play important roles in future efforts to learn about social interactions”, Manski, 2000).

70This is a new field for developing countries, especially sub-Saharan Africa. Hitherto data have been lacking, but the situation is changing fast. There are structural reasons for the enthusiasm. The economics of well-being has always been at the heart of the economists’ scientific programme. The subjective approach makes it possible to investigate themes of central importance, such as preferences and reasoning in Africa, that were hitherto the reserve of anthropologists and sociologists. But there is also a conjunction of scientific and political factors that should lead a growing number of researchers to find the approach interesting:

• After two decades of structural adjustment, official development aid is now massively focused on poverty reduction. Thus more than sixty developing countries (DCs) have joined the twin initiatives on poverty reduction (Strategic Document on Poverty Reduction) and external debt (Heavily Indebted Poor Countries, HIPC), launched internationally by the Bretton Woods institutions in the late 1990s. This marks a shift from an instrumental approach to policy, based on the Washington Consensus14, to a results-based approach: reducing poverty.
• After a period when the monetary conception of poverty ruled supreme (poverty = inadequate income), the multidimensional nature of poverty is now unanimously recognized, as witness the 2001 edition of the World Bank’s World Development Report. In line with the work of Rawls and Sen, the broadening of the concept of poverty has prompted economists to take into consideration not only monetary resources but also basic needs (health, education), empowerment, social participation or exclusion, and freedoms. The composite indicators developed by the UNDP since the late 1990s (the Human Development Index and Human Poverty Index) incorporate some of these non-monetary dimensions.

71Further, the accent is now on the political economy of reform, and here the participation of the population is an essential condition for success. This orientation is a major encouragement to take into consideration the populations’ viewpoints, and especially the viewpoint of the poor. With the concepts of empowerment, accountability and voicing fully incorporated into public policy, sociopolitical surveys and representative opinion polls are an original method for tackling this vast and as yet little explored field, as an additional contribution alongside the more traditional qualitative approaches (Narayan et al., 2000). Aside from their contribution to knowledge of poorly-understood social and political phenomena, these tools can play an important part in the conduct of public policy by revealing the views of the population and increasing the negotiating power of social groups who have been traditionally excluded from the decision process. More systematic use of opinion polls, which can be used to combine subjective and objective data (practices, degrees of satisfaction, etc.) should, among other things, encourage the application of human rights and help towards human development, playing a part in policy development. Statistical surveys are even more obviously useful for listening to and passing on “the voices of the poor” (Razafindrakoto and Roubaud, 2002, 2005a). As a rule, their interests are not reflected in political spheres, elected or otherwise, and the intermediate bodies that are supposed to represent them (producers’ cooperatives, users’ association, etc.) do not have the power to make good their demands.

72Subjective, survey-based approaches to well-being and poverty make it possible to address a huge range of questions and draw up a wide array of indicators linked to poverty and the poor. These include perceptions of poverty and its causes; demands and preferences; attitudes and behaviours; satisfaction rates, etc. They have a sound theoretical base (not reduced to the matter of income, and not imposing set choices of preferences). The data are also much easier to collect than monetary data. One of the main limitations of income data is their unreliability, especially in developing countries, where statistical capacities are weak. Below is a sample of questions on subjective poverty and well-being from the questionnaires used in various developed countries (INSEE, 2005) and also in Africa and Latin America (Herrera, Razafindrakoto and Roubaud, 2007a).

• Q. Given your household’s income, do you consider that
  • 1. you live well,
  • 2. you manage ok
  • 3. you have to be careful
  • 4. you are hard up
• Q. What is your household’s current situation?
  • 1. you manage to put quite a lot aside
  • 2. you manage to put a little aside
  • 3. you just manage to break even
  • 4. you have to draw on your savings
  • 5. you have to borrow money
• Q. What is your estimate of the minimum necessary for your household to live on?
• Q. Are you satisfied as regards your household’s needs in terms of
  • A. food
  • B. shoes and clothing
  • C. housing
  • D. healthcare, medicines in case of illness
  • E. children’s education
• Q. If you consider all aspects of your life, do you count yourself happy? (scale 1 to 7: “very happy” to “not at all happy”)

73Despite their positive qualities, the subjective approaches do pose measurement problems. Some researchers therefore have reservations about their use. Results can be more sensitive than in other fields where the questions are more precisely formulated. But in our opinion, the main limitation of this type of approach is the problem referred to in the literature as preference attrition, with the political implications that entails. As a rule, the poor, the most disadvantaged and dominated social groups, tend to be satisfied with living conditions considered unacceptable by those who are better off. For example, the minimum acceptable income (see question above) is an increasing function of households’ real incomes. And estimations show that, all else being equal, the poorest declare themselves happier than the rest. This result is explained by a form of social conditioning that has been denounced by Sen (1984) among others: “The most blatant forms of inequalities and exploitations survive in the world through making allies out of the deprived and exploited. The underdog learns to bear the burden so well that he or she overlooks the burden itself. Discontent is replaced by acceptance… suffering and anger by cheerful endurance. As people learn to adjust… the horror look less terrible in the metric of utilities”. Can it be legitimate to class as “poor” two households whose incomes differ by a ratio of 1 to 100, simply because they are dissatisfied with their living conditions? By allowing the definition of poverty and its threshold to vary from person to person according to their respective preferences, the subjective approach allows individuals to determine for themselves what is good or necessary. But at the same time it ratifies, without correction, phenomena of attrition or self-limitation of aspirations which increase with increasing poverty.

3.2. Other Fields of Application: Business Cycle Surveys and Expert Surveys on Governance

74The introduction of perception indicators in economics is not restricted to the measurement of subjective well-being. Nor is it new. Two significant examples can be given by way of illustration. Since the 1950s, opinion polls (of individuals, households or companies) have been increasingly widely used in all countries. Today they constitute a fully-fledged survey system in national statistical information systems (Fayolle, 1987; Cling, 1990). These polls are particularly useful for providing rapid indicators on such quantitative variables as incomes, consumption, production etc. before the conventional statistics become available. As such they are used in economic forecasting. For example, the monthly household survey that is used in all European countries asks about income, consumption, purchasing intentions, etc.

75Contrary to received wisdom, these kinds of data are no more complicated to process than traditional economic statistics. Standard instruments from quantitative sociology can be used, as can the body of tried and tested procedures from economic opinion polls, developed by business cycle polls particularly. One example is the opinion balance method. For a question concerning trends in a given phenomenon that may be rising, stable or falling, the opinion balance is the difference between the percentage of respondents who answered “rising” and the percentage who answered “falling”. These subjective questions may concern trends in a quantitative variable whose level the household knows (e.g. “living standard”, which corresponds to the purchasing power of the available income) or in a more general phenomenon such as the country’s degree of democracy. In the first case the assumption underlying the use of the opinion balance is that the scale of the phenomenon is correlated with its distribution. Thus the faster incomes are rising, the more numerous will be the households that benefit from that rise. More precisely, under certain hypotheses about the distribution of the trend in the variable within the population (normal law) and respondents’ response behaviours (the hypothesis that there are response thresholds above which respondents say a variable has risen or fallen), there is a linear relationship (S = ax + b) between the opinion balance and the actual trend in the variable concerned over the reference period. Except when the opinion balance is close to zero (owing to a response bias due to optimism or pessimism among the respondents), one may consider that the sign and level of the balance is a usable test of the robustness of the hypotheses concerning the link between perceptions and reality. In particular, studies show that trends in monthly opinion balances are closely correlated with consumption trends. Country to country comparisons of opinion balances are a delicate matter because response behaviours differ from country to country, so that for a given level of the quantitative variable concerned (living standards in this case), the opinion balance may differ from country to country.

76The second example is more recent, but is a field of application has been growing fast in recent years in the development field. This is the use of expert surveys on matters of economic and political governance. The best known is doubtless the Corruption Perception Index (CPI) published yearly since 1995 by the NGO Transparency International. It is used to classify countries worldwide on a scale, according to their degree of corruption (see e.g. World Corruption Report, Transparency International, 2005). But while with time and hindsight, with the calibration of data and the standardization of processing methods, the analysis of business cycle surveys has become well-tried and reliable, the handling and exploitation of the results of expert surveys is conversely still risky and hesitant.

77Fiercely criticized at first, they have now acquired some degree of legitimacy. Two kinds of argument have been employed in the literature to justify their validity. Correlating indicators from mutually independent databases shows that they do indeed inform the same phenomenon. And a growing number of studies shows that these indicators have a significant impact on the reality in question, such as growth, investment, international trade etc. But there is nothing to show that there is a link between these perception indicators and actual levels of corruption. An original experiment showed the weaknesses of expert surveys. It was run simultaneously in eight African capital cities, the principle being to couple household surveys with a “mirror survey” among specialists from North and South on the same issues (Razafindrakoto and Roubaud, 2005b). On the subject of corruption, comparing the two sources shows that the experts systematically overestimated the degree of corruption citizens had to endure. While an average 13% of the population stated that they had been direct victims of acts of corruption in the past year, the experts’ estimated rate was 54%. Nor was this bias uniform; there were large differences in the ranking of the different countries by household surveys and expert surveys. In fact there was no correlation between the two variables measuring corruption (one in the expert survey and one in the household surveys), but the “mirror survey” correlated closely with the other international databases on the subject.

78If this kind of database was only used within the narrow field of research (and the flourishing industry of econometrics of growth), the harm would be limited, though it is always harmful to spread unfounded ideas. But in fact the corruption index directly affects the attribution of development aid. Eligibility for the Millennium Challenge Account (MCA), set up recently by the United States to help poor countries, depends among other things on the quality of a country’s governance as judged, among other measurements, by the expert surveys (Radelet, 2004).

3.3. Perceptions of Well-being and Life Course Surveys

79Experiments in measuring subjective well-being through household surveys has shown that this is a useful approach. The indicators are robust, both in cross-sections (see special issue of Économie et statistique on this subject, Herpin and Verger, 1997; and for DCs, Razafindrakoto and Roubaud, 2005c), and by panel (Lollivier and Verger, 1999; Lokshin and Ravallion, 2002; Herrera, Razafindrakoto and Roubaud, 2007b). The question here is whether this approach can be transposed to life course surveys.

80In the first place, it should be stressed that progress in this field would be extremely promising scientifically. To combine the two approaches’ strong points (an easily-collected measure of well-being on the one hand and the temporal dimension offered by long-term individual and collective event histories) would be a huge step forward for analysing the dynamics of poverty and hence for development policies. For economists, it would fill a gap in the life course surveys used to date, particularly in developing countries, as they gather no data on well-being, financial or otherwise. Some surveys have tried to capture histories of income, including the 1989 Dakar survey (but only for the main occupation) and the 1996 Yaoundé survey (wage before and after any promotion during the period), but without much success. The experiment was not repeated in subsequent surveys because only a minority of respondents were wage-earners. Moreover, apart from the high proportion of non-responses, many respondents gave only their basic wage rather than income including bonuses (especially in the civil service).

81The use of multiple-round surveys has so far proved unsatisfactory in this regard, especially because of the attrition problem. Where panels have been set up they have rarely been maintained for more than two successive survey rounds. At best, existing panels can only address short-term trends.

82But might life course surveys be relevant in this connection? Designed for dating events and changes of state, might they be useful for recording processes and monitoring continuous variables like well-being? The retrospective nature of life course surveys is a problem for recording perceptions over time. Could statements concerning perceptions of well-being (WB) for situations dating back thirty or fifty years be considered reliable? This retrospective measure can only be a reconstruction ( ), whose connection with the (potential) series of “true values” of subjective well-being (recorded in real time over the life course, ) would have to be calibrated. Clearly, even apart from memory problems, the response is likely to be conditioned by the respondent’s personal situation at the time of the survey and more broadly by the events that have occurred since the period for which the information is sought. As a result, someone who has lived through painful experiences like the death of a child or a separation may tend to think of the earlier period as a golden age although they did not necessarily see it that way at the time.

83But are these shortcomings entirely damning? A priori judgements apart, it has been clearly shown that reconstituting perceptions is more delicate and questionable (revisions, adjustments, misrepresentation, etc.) than reconstituting objective facts. Even measurements of such basic variables as level of education tend to over-estimate, and the older the respondent the greater the bias. On the other hand, information from the life history narratives that underpin the work of anthropologists and sociologists are obtained in similar conditions, but they are not dismissed as invalid. The critical equipment developed for this type of qualitative approach could be used to process similar data obtained from quantitative surveys. At worst, respondents’ statements can be taken as they stand: a measurement of well-being at different stages in the life course, as assessed at the time of the survey. At the very least, this avenue deserves to be tested empirically. And this is exactly what the Biographies et entourage survey does, as we shall see below. Interesting suggestions are made by Philippe Bocquier (modelling and predicting events) and Xavier Bry (modelling gradual transitions) in Antoine and Lelièvre (GRAB, 2006).

4. The Life Course As a Whole: Actual and Perceived Time Periods

84In a life course study, the entity analysed is not necessarily the event but the life course itself, regarded as a complex process. In the Biographies et entourage survey, although the data gathering reconstitutes the events that have punctuated the respondent’s life rather than their motivations, at the end of the questionnaire the respondent is asked to look at the grid in which the chronology of family, residential and occupational events is recorded, and to divide up their life course, with explanations, as follows:

• Q. Can you divide up your life into different periods?
  describe these periods, identifying what it is that differentiates them from each other and what they represent in your life
• Q. For each period,
  • Vg : very good years
  • G : good years
  • NP : years with no problems
  • D : difficult years
  • Vd : very difficult years

85This allows the respondent to identify the periods and nodes of their life course (Tichit and Lelièvre, 2006) and to select the highlight events and fields that have meaning for them.

86In the Biographies et entourage survey the average number of factual landmarks (taking into account family events and changes of residence and occupation) per respondent aged 50 to 70 is about twenty, and the average number of periods identified is four. Accurate to one year15, 37% of factual landmarks coincide with a change of period. This low proportion is due to the fact that there is an average of 20 landmarks, some of them concomitant, but an average of only four periods over the same stretch of time. For example, 57% of dates given as the start of coupledom (including marriages) also mark a change of period; 50% of separations (including divorces) do so, and only 27% of births (23% for men and 30% for women).

87Conversely, only 13% of changes of period synchronize with none of the events included in the survey. It is unsurprising that respondents associate a passage from one period to another with a factual landmark. For example, 81% of changes of period correspond to a change of residence but only 19% to a wedding.

88Lastly, respondents were asked to make an assessment of their financial situation, based on the initial division into periods but altering it if necessary:

• Q. During these periods, how would you say your household’s income enabled you to live
  • 1. well
  • 2. all right
  • 3. all right but we had to be careful
  • 4. it was difficult
  • 5. we had to borrow to get by

89This should make it possible to compare well-being with financial affluence and grasp their perceived dynamics.

90The problem of calibrating this scale, addressed in section 3.3 above, is here only partly solved by the next question, which measures the household’s current income. That information in no way counterbalances the respondent’s comparison of different periods. They may compare an earlier financial situation with their current situation and describe it as adequate, whereas if asked at the time they might have perceived it differently. However, what is sought is an interpretation, so this reading is at least as valid as that of the researcher, who deduces (also retrospectively) from the factual information the time trend in the living conditions of the respondent’s household.

Conclusion

91This chapter has shown that the classic dichotomy between objective data and subjective information is partly artificial. Categories usually regarded as “hard” because they are factual are also based on perceptions. But rather than contrast these different types of information or try to eliminate any notion of subjectivity from the data gathered, the examples given here argue for seeking out and incorporating the perceptual elements as best we can, whether using them in their own right or using them to improve the usual data collecting and analysis categories.

92Certainly, the respondents introduce assessments in their responses. But the researcher also uses categories and divisions, and it would be well to investigate these. The work of data preparation, constructing typologies and coding categories is not neutral, but it is rarely discussed. These reflections suggest we should examine more closely and take more care with these stages of data production. Not in order to abandon hope and see only the limitations, but to take advantage of the fuzziness, measure the content of the fuzziness. To return to the film metaphor used in the introduction, it is a matter of focus. With the help of the perceptual elements, a caricatured “editing” of the sequences can be avoided.

93Lastly, this chapter has emphasized that asking for perceptual elements and including them in the analysis is not the sole reserve of qualitative approaches. One of the lessons from our work is that, apart from the complementary roles the two approaches can play, the use of subjective variables in quantitative analysis is a promising field.