Methodology

The study adopts three alternative approaches for studying the family household: (1) the analysis of household size, type and composition, (2) the developmental or family life cycle approach and (3) the use of a macro-simulation model. The first approach is used to examine the static perspective of households and family households in the Philippines. The characteristics of households and of family households that are examined include the size, type (for example, lone-person household, nuclear and extended family household), number of non-nuclear family members, characteristics of household heads and characteristics of other members of the family household.

2.4.1. Analysis of household size, type and composition

This approach makes use of cross-tabulations and of summary measures such as means and proportions. To measure the strength of the association between two variables which are measured on a nominal scale (for example, type of family

household and area of residence), a chi-square-based measure, Cramer's V, is used. This measure modifies the chi-square value so that the new value is not influenced by the sample size (Norusis, 1988: 283). Its value ranges from 0 to 1; a value of 0 corresponds to no association, and a value of 1 to perfect association.

One-way analysis of variance (ANOVA) is also employed to test the null hypothesis that in the true population all group means are equal. An example of group means is the mean sizes of households in the residential groups: Metro Manila, highly urban areas outside Metro Manila, other urban areas, and rural areas. To identify which pairs of groups have different means, multiple comparison tests are performed. For the present analysis, three multiple comparison procedures are specified: Student-Newman Keuls, Tukey's alternate procedure and Scheffe's test. One alternative to carrying out multiple comparison tests is calculating t-test statistics for all possible pairs of means. The reason for not using a t-test, however, is that the probability of finding one or more pairs of means to be significantly different increases with the number of comparisons carried out. That is, the more comparisons performed the higher is the likelihood that at least one pair of means turn out as statistically different even if all means in the population are equal (Kirk, 1968: 78; Norusis, 1988: 268).

2.4.2. Developmental or family life cycle approach

A developmental or family life cycle approach, by contrast, looks at the family as a process rather than a static unit within certain periods of time. The approach is based on the observation that individuals go through different family patterns and household structures over their individual life cycles, and that families and households undergo various types of structures, organisations, and relationships (Hareven, 1977: 340). Such a development is obscured in the snapshot approach. The life cycle approach and the macro-simulation model are employed to deal with the developmental changes and the dynamics of the family household.

The family life cycle is usually defined as the succession of stages through which the typical or nuclear family passes (Rowland, 1991: 3). Duvall (1967: 4)

described it as a way of taking a long look at family life. According to Young (1994: 128), it is the term used to describe the changes in the size, composition and functions of the family over its lifetime. The family life cycle approach therefore is a tool for analysing the developmental perspective of the family.

A forerunner of studies on family life cycle was the work of B. S. Rowntree (1902, cited in Rodgers, 1977: 40; and in Young, 1977: 5). An attempt to understand the pattern of poverty in late nineteenth century and early twentieth century England, the study was one of earliest which analysed the family in a developmental way. However, major credit for quantifying the concept of family life cycle goes to Glick (1947), who made the first detailed investigation of the timing of the main demographic events in the family life cycle. He derived the median ages of husbands and wives at certain key demographic events and used these to describe the experience of a nuclear family. According to him, the family begins with the marriage of the couple, gains in size with the birth of each child, and ends at the death of the last surviving spouse (Glick, 1947: 164-165).

In the literature on family, other expressions have emerged in relation to the concept of family life cycle. The expressions family career and family life course are sometimes used in place of family life cycle, and state in place of stage. Rodgers (1962: 23-24; 1977: 42) argued for the abandonment of the use of the terms stage and cycle,

and proposed that these terms be replaced with the concepts category and career,

respectively. His view that the family life cycle stages are merely analytical categories is in agreement with the distinction made by Francis (1958, cited in Rodgers, 1962: 24) between the terms stage and state. According to Francis, a stage implies a pre­ determined progression which is invariable, while a 'state' simply implies a condition at a point in time.

Thus the present study adopts the expression family life cycle state since it uses cross-sectional data, the 1990 population census data of the Philippines, in its analysis of the characteristics of the family household at each state of the life cycle. A typology

of family life cycle states is constructed for the present analysis, and this is based on the classification scheme proposed by Priest (1982: 78). A detailed discussion of this typology is presented in Sub-section 5.3.2 of Chapter 5.

2.4.3. The macro-simulation model

Macro-simulation models consist of intricate mathematical equations the numerical solutions to which are calculated by the use of a computer program. Sometimes called projection models, macro-simulation models are deterministic because the rates at which demographic events and changes in family status occur are exactly determined by specified input variables (Bongaarts, 1983: 33). Multi-state increment-decrement life tables form a class of macro-simulation models which have made rapid and very promising developments in recent years (Oechsli, 1975; Schoen,

1975; Willekens et al., 1982, all cited in Bongaarts, 1983: 33).

A family status life table is a macro-simulation model which is constructed with the same basic technique used in the calculation of multi-state life tables, such as the marital status life table. Required as inputs are the risks of dying and of transferring between family states to which individuals are subjected. Family status life tables can be calculated either for males or for females, and may contain several hundred states. For a cohort of women, for instance, apart from the four marital states, which are typically used in a marital status life table, the family status life table includes the so- called maternal states, to describe the number, sex, age, and residential status of living children, as well as the women's parity and fecundity status (Bongaarts, 1987: 191).

Bongaarts' computer program FAMTAB generates a nuclear family status life table using an ever-married female member of the family as the marker. In his model, each ever-married woman stands for a nuclear family, as it is based on the assumption that no married children live in parental homes (Zeng, 1987: 9). Zeng (1987) extended Bongaarts' nuclear model into a model that accounts for both nuclear and three- generation families. In this extended model, if an ever-married woman and her children live with her parents or her parents-in-law, then this woman stands for a three-

generation family. One important limitation of Zeng's model is that it ignores cases in which ever-married siblings live together.

To implement his model a computer program called FAMY was developed (Zeng, 1990). FAMY simulates a synthetic cohort of women's marital, parity, maternal and marker status changes under given demographic regimes (Zeng: 1991). The present study uses FAMY to calculate the proportions of women in the Philippines who were in different family statuses and the average duration spent by these women in each family status. The computer program FAMY and its data requirements are described in detail in Chapter 6. The family status life tables that are generated are based on a number of assumptions, and these are discussed in Chapter 6 as well. The output of the simulations is presented in Chapter 7.

This chapter has presented the conceptual framework employed in this study, which describes the inter-relationships between the demographic factors, the cultural norms, the political, social and economic conditions of the society on the one hand and the size and structure of the family household on the other hand. It has also discussed the data and methodology used. The next chapter examines the trend over time of the average size, and the distribution by size and by type of households in the Philippines. It also analyses the differentials in sizes and types of households across a delineated continuum of level of urbanisation.