• No results found

4. METHODOLOGY AND DATA

4.2. Unit of analysis

Setting the boundary and the focused level of analysis is the basic element of scientific research. A unit of analysis is defined as a subject – either ‘what’ or ‘who’ – on which analyses are generalised, thus could be different from the unit of observation (Lewis-Beck et al. 2004). It could vary from the most micro one (which is an individual in social sciences) up to a system that includes many sub-systems. In most of the cases, these levels interlink and thus could overlap and cause confusion in analysing (Long in Lewis-Beck et al. 2004). Household is the principal unit of analysis applied in this research. The main analysis is generalised on the core database originally collected through the standardised household survey. However, a significant level of analysis is also made on different scales as well as crossing all levels from individuals up to the social-ecological system. This does not only ensure a more comprehensive approach to such a large and spreading research context, but also facilitate a substantial triangulation process of results. In a more strategic framing, while findings at the household level allow identifying the vulnerable groups to be targeted, the cross-scale analysis provides a sound background for potential policy implications arguments. A more in-depth analysis of the cross-scale analysis theoretical background is provided in Section 3.1.3.

There is no standardised definition of the household unit in the research community but rather varying across projects of different themes and contexts. This practice results in the variation of household composition as well as household size applications. Even within a specific research context, a household’s arrangement changes over time (Deaton 1997, Beaman & Dillon 2011). Therefore, ensuring consistency in understanding since the designing phase and during the data collecting process (i.e. enumerators training and survey quality control) is critical to the quality of data.

In the Vietnamese context, according to the GSO, the definition applied for collecting data for the national database, particularly the census “Viet Nam household living standards survey” conducted every ten years, “[H]ousehold (or family household) is a person or a group of people living together” (GSO 2016). Although it is also stated in the report that this definition is not necessarily in line with several other research which is commonly

60

dominantly based on the legal household registration book9, yet it could be referred to

when establishing the definition used in this research. The latest census data collected in the year 2009 (GSO 2010) shows a decreasing tendency of household size and substantial difference between urban and rural areas (see Figure 4-1). These trends are reflected quite well in the empirical data (more details in Chapter 5).

Average household size, Viet Nam 2009 Average urban household size, Viet Nam 2009 Figure 4-1: Average household size by provincial level and for urban areas in Viet Nam, 2009

(Source: GSO, Census on Population and Housing 2009:81 )

The definition of a household in this research shares the same approach in which the membership is not identified by the legal registration book, i.e. de jure, but based on the de facto situation of that family (confirmed by respondents). Accordingly, household leadership and membership are defined based on the actual internal agreement among family members. Explicitly, a household does not only include people those are currently present at home, members that are temporarily (or more precisely non-permanently) living in other localities, yet remaining regular connections in terms of time, income contribution and closest members (e.g. spouse, parenthood) with the interviewed household are also counted in the household size. A definition of “extended household size” is also applied

9Household registration book (sổ hộ khấu in Vietnamese) is the main management instrument (public security, economic

planning, and control of migration) in the household registration system in Vietnam for more than 50 years (World Bank 2016). There have been many critiques raised on the pragmatism as well as many sequential social issues (e.g. discrimination to migrants in big cities) of this system; yet the debates are still on-going. Independent (including international) organizations such as World Bank and UN also contribute to addressing this issue.

61

where former family members who at one point in the past (regardless when) moved out to establish their own family; of which in most of the cases it happens after the marriage of household head's children. This data is expected to provide more information about social capital at the household level. Further intuition and application of this variable could be found in empirical analysis sections. Moreover, a certain variation of household composition between regions of Viet Nam, particularly between the Northern and Southern parts, is also discussed in the literature (Bryant 1996). Therefore, having a consistent definition and approach across large research areas is critical to a valid method. The questionnaire for the standardised survey was therefore designed accordingly.

The data pool is retrieved from the household survey, analysis at the household level is indeed generalised on the information from individual interviews. Deaton (1997) also discussed the issue of measurements at individual level versus household level and the importance to make the data available for better assessment. Related to a decision-making analysis, (Grossbard 2011) also discusses between independent individual models of decision-making and of joint decision-making in households whose views were taken into consideration in the research design.

In order to control for the bias and ensure the relevance between individual data of respondents and data at the household level, certain selection criteria with attention on respondents’ representativeness are associated with the stratified sampling technique. In other words, not only the household characteristics, such as main livelihood/occupation, geographical locations, but also individual characteristics of respondents, including female versus male, age, role in the household are considered for proportional sampling. Therefore, the information provided by respondents could represent the household. In this way, the potential bias of data is controlled with the selection process.

Related documents