Applying the Consensual Method of Estimating Poverty in a Low Income African Setting

As previously mentioned, there is now consensus that poverty no longer constitutes simply a low income, and that non-monetary aspects or dimensions of poverty are also important to consider. Related to this are a number of changes in the way people’s needs are conceptualised and ascertained. Early poverty surveys (e.g. those by Rowntree in the early 1900s) focused on more absolutist, subsistence needs, such as having enough money to buy sufficient food to prevent starvation, clothing and shelter. Subsequent researchers, like Peter Townsend, realised such minimalist thresholds were insufficient, and instead developed ideas around the concept of relative deprivation. For Townsend,

A frequent criticism of poverty surveys is that they too often rely on the opinions of well-remunerated experts (academics, social workers, etc.), and rarely incorporate the opinions of those with a lived experience of poverty—i.e. poor people themselves—or the wider public (Citro and Michael 1995). Such concerns led Mack and Lansley to develop the Consensual Approach to poverty, which makes the opinions of the general public a central part in both the definition and measurement of poverty. Building on Townsend’s earlier work and his concept of relative deprivation, Mack and Lansley developed and refined methods which asked the general public about their views on what constituted an acceptable minimum standard of living. Focus group discussions (Pantazis et al. 2006a) were used to see what types of items (e.g. warm winter coat) and activities (e.g. celebrate special occasions) people thought no one should have to go without due to a lack of money—i.e. were essential and necessary. Those items which a majority of respondents (i.e. 50 % or more) believed to be necessary were then taken forward and respondents asked (either in the same survey, or in a separate survey as has been done most recently in the UK 2012 Poverty and Social Exclusion Survey) if they had or did not have the items, and if they lacked them was this due to choice (i.e. did not want) or a lack of resources (i.e. could not afford). The items a majority of people thought necessities formed the basis of a deprivation index, on which respondents lacking items considered necessities because they could not afford them, scored (for each item) a 1 (deprived) or 0 (not deprived). Townsend’s work showed that multiple deprivation and poverty are highly correlated, and that below a certain level of income (or resources), the number of deprivations experienced increases rapidly (Gordon 2006; Townsend 1979). It was at this point, or kink, of between two and four deprivations (depending on the study) that a threshold could be set to divide the ‘poor’ from the ‘not poor’ (or the deprived from the not deprived). As Gordon noted, when setting out criteria for identifying an objective poverty line, this should be the point “that maximises differences between the two groups (‘poor’ and ‘not poor’) and minimises the differences within the two groups (‘poor’ and ‘not poor’)” (Gordon 2006: 39).

Over the last few decades, the consensual approach has been used to examine relative poverty in a number high-income countries, including Great Britain (Gordon and Pantazis 1997a, Mack and Lansley 1985; Pantazis et al. 2006a), Northern Ireland (Hillyard et al. 2003), Belgium (Van Den Bosch 2001), Sweden (Halleröd 1994, 1995), Finland (Halleröd et al. 2006), Japan (Abe and Pantazis 2013) and Australia (Saunders and Wong 2011). More recently, researchers have begun to use the method in middle income and low income countries, including in Bangladesh (Ahmed 2007), Vietnam (Davies and Smith 1998), Mali (Nteziyaremye and Mknelly 2001), Tanzania (Kaijage and Tibaijuka 1996), Zimbabwe (Mtapuri 2011) and South Africa (Noble et al. 2004; Wright 2008). In each instance the list of items/activities has been modified to reflect local conditions and customs, and in each the method has been found to be effective and reliable. Here, we report on our use of the method in the context of Benin.

The data used in this paper were provided in the 2006 Demographic and Household (DHS) Survey for Benin. DHS data are regularly used by organisations like the United Nations and researchers to assess people’s living conditions in low income countries (Corsi et al. 2012; Gordon et al. 2003; Vaessen 1996).

Benin, located on the west coast of Africa between Togo and Nigeria, had in 2006 an estimated population of between 8 and 9 million. It is classed by the World Bank as a low-income economy, and in 2011 was ranked 166th out of 187 countries by the United Nations Development Programme’s Human Development Index. In 2005, the year closest to the data used in the paper, life expectancy at birth was 53.9 years, the mean number of years of schooling was 2.9 and the gross national income per capita (adjusted for purchasing power parity) was $1,340 (UNDP 2013). In 2007, around one-third of the population had incomes below the national poverty line. As with most countries in Africa, the population of Benin is socially and culturally heterogeneous; over a dozen distinct languages are spoken across the many different tribal and ethnic groups, which follow a mix of Muslim, Christian and African Traditional religions.

The 2006 DHS for Benin provides data collected separately though the Integrated Modular Survey on Household Living Conditions, run by the National Institute of Statistics and Economic Analysis. This module ascertained respondents’ opinions about what items/activities they considered to be necessary for a decent standard of living (INSAE 2007). Questionnaires were translated into the main languages of Benin (French, Adja, Bariba, Fon, Dendi, Ditamari and Yoruba), and interviews conducted mainly in French but also in Adja, Bariba, Fon, Dendi, Ditamari, and Yoruba.² Respondents were asked about their perceptions of necessities and also the degree to which they felt these needs were met for them and their families. Questions were asked about a range of items, covering food, clothing, housing, health/hygiene, transportation, education and leisure, work and relationships.

The survey used a multi-stage stratified sample, and was administered to 17,511 households, with one adult respondent per household. The sample reflected the roughly 40:60 urban-rural split in population and was representative at sub-national level. Full details of the sample are available from the national DHS report.³

Respondents were asked, from a list of 26 items/activities (covering food, clothing, housing, transport, etc) which they considered to be essential (“indispensable” in French) to have to achieve a decent standard of living.⁴ Included on the list were:⁵

Respondents were then asked (for a sub-set of the original items) the degree to which they felt their own and their households’ needs were satisfied.⁶ The sub-set of items/activities about which follow up questions were asked included the number of meals per day, daily consumption of cereal/root vegetables, daily consumption of vegetables, daily consumption of meat or fish, meals on holidays and special occasions, clothing, shoes, the quality of respondent’s homes, access to drinking water and electricity, furniture in the home, availability of cleaning products, access to health facilities, medicines for illness, personal hygiene products (e.g. soap), the means of transport used, education for Children, leisure (or holiday), relationships with family and friends, and being able to provide assistance to parents in time of difficulty.

Studies like the 2012 Poverty and Social Exclusion Survey in the UK follow a two-stage process; people’s perceptions about what items/activities they consider necessities are ascertained in one survey, and a second survey (on a different sample) asks respondents if they have/do the items/activities asked in the previous survey, and if they lack or do not do them is it out of choice or because they cannot afford to own/do them. Those reporting they lack an item (which a majority at the first stage considered a necessity) because they cannot afford it, are considered deprived. Others doing similar work, such as Saunders et al. (2007) in Australia, used a single survey to ascertain both opinions about necessities and whether or not people lacked them.⁷ This paper follows the approach taken by Saunders et al. (2007), in using data where both sets of information were collected in a single survey. We consider only responses which state items/activities as ‘essential’ (i.e. “indispensable”) to be conceptually equivalent as ‘Necessity’ in other consensual studies, as counting items defined as either “plutot nécessaire” (quite necessary) or “indispensable” (essential) would result in almost universal agreement across all items on the list. The difference between quite necessary and essential was also suggested by the questionnaire wording which asked respondents if the item was: (1) essential or (2) quite necessary (3) not needed. Quite necessary was therefore used as the mid-point between essential and not needed, and we therefore decide to consider only the items considered essential. The adoption of this stricter threshold ensures we err on the side of caution, and are more certain that items counted as essential are really considered such by respondents.

Relative risk ratios and their 95 % Confidence Intervals (CI) were computed to show whether differences between groups of respondents were statistically significant, and also the size and direction of these differences (Morris and Gardner 1988). Relative risk ratios show the probability, or risk, of one group thinking an item to be essential compared to another group. A relative risk of 2 means twice the risk, a risk of 0.5 implies half the risk, etc. Where confidence intervals cross 1 means there is no difference between the two groups (Gordon 2013).

People experiencing poverty (and deprivation), then, are effectively discouraged from demanding radical change or high enough norms or standards to meet expert opinions (or peoples’ needs), and instead accept their circumstances out of necessity (Sen 1992). This, some claim, is a potential source of bias, in that it results in a constrained expression of what social norms really are or should be, and thus renders them unreliable.

Crettaz and Suter (2013) using data on Switzerland, recently found evidence of adaptive preferences (downward adaptation) among individuals affected by income poverty, and that both indicators of material deprivation and subjective indicators related to income satisfaction and financial constraints were affected. However, they note, “bias caused by adaptation processes…varies considerably among the various measures and indicators examined within each of these two groups, and some of them, in fact, appear not to be affected at all” (2013: 148). They found that overall deprivation indexes like those developed by Townsend (1979) and Halleröd (1995) were either only weakly affected or did “not show any bias due to adaptive preferences of poor people” (Crettaz and Suter 2013:149).

Noble et al. (2008) examined whether adaptive preferences and bounded realities affected a democratically defined measure of poverty in South Africa. Such an exercise is particularly important in a country like South Africa, with its legacy of Apartheid which systematically marginalised and disadvantaged a majority of the population. Any attempts at a “democratic” definition of poverty, using the consensual approach, would need to ensure that what emerged had not been dampened down by peoples’ experiences. Controlling for differences between population groups and area (i.e. urban/rural), Noble et al. found that respondents located above a subjective poverty line (i.e., who reported that their household income was greater than the amount required to make ends meet) were more likely than those who were not above the subjective poverty line to define certain items (e.g. ‘meat/fish/vegetarian equivalent every day’) as essential. A recent paper by Wright and Noble (2013) explored the issue of adaptive preferences in South Africa in more detail, and examined whether or not respondents possessing an item were more or less likely than those lacking it to consider it a necessity. While respondents who owned items were more likely to claim it to be necessary, what was also apparent was the fact that those lacking the item did so because they could not afford it, not because they did not want it. So while they found some evidence of adaptive preferences, with poor people less likely to report an item as a necessity, they nonetheless argue convincingly that this does not affect the reliability of the method or its applicability to low income countries.

We used the Benin data to see if there was any evidence of adaptive preferences among people experiencing poverty. If it could be shown that respondents who report their needs for a particular item were “not satisfied at all” (i.e. were deprived of a socially perceived necessity, see Table 1) were systematically less likely than those who felt their needs were met (and thus not deprived) to consider items on the deprivation index as essential, this would suggest some evidence of adaptive preferences, with implications for the overall deprivation index. Controlling for respondent’s age, sex, place of residence (i.e. urban or rural location) and household wealth quintile, we compared the relative risks of those deprived/not deprived thinking whether an item was essential or not (Fig. 4). It is apparent that in the case of Benin, people in poverty (i.e. defined here as those deprived of four or more socially perceived necessities) were marginally more likely than those not experiencing any deprivations (i.e. those not in poverty) to consider all of items on the deprivation index essential. This analysis provides evidence to contest the claim that poor people adapt their preferences downwards with regards what they consider the necessities of life, and suggests instead that they in fact have opinions and views very similar to the non-poor as to what constitutes a decent standard of living.

Open image in new window

To confirm that this result was not due to the use of the overall deprivation index, risk ratios were run for each individual item on the index, comparing those reporting that their needs were not met at all (i.e. and were thus deprived) with those whose needs were met (i.e. who were not deprived). In almost every instance (i.e. 254 out of 256 tests¹⁰), deprived respondents were significantly more likely to consider each of the items in the deprivation index to be essential, compared to those not deprived (see charts in Appendix “4”).

This is not the first time a study of consensual poverty has reached such a conclusion. Gordon and Pantazis (1997b) found no major differences between “multiply deprived” and “less deprived” respondents about the necessity of different items; in fact, respondents who considered themselves “genuinely poor all the time now” were more likely to report certain items and activities as necessary than never-poor respondents, including having carpets in living rooms and bedrooms, having a television and being able to have a night out fortnightly. Nevertheless, it should be noted that the differences between the two groups are extremely small: whether controlling for age, sex, place of residence and wealth or not, deprived respondents are only about 10 % more likely to perceive items as necessities than the non-deprived ones. This however represents strong evidence of lack of adaptive preference among the surveyed population of Benin.