Table 1 shows which items and activities respondents in Benin consider to be essential for a decent standard of living. Most (over 80 %) consider important basics, such as having access to drinking water, to care when sick and having steady work as essential. A majority also believe that (among other things) having three meals a day, being able to send children to school, having access to electricity and a form of transportation are all essential. The table also shows (final column) the proportion of respondents reporting that their needs for each item were “not at all satisfied”; nearly half (46 %) of people felt their need for electricity was not met, around a third (31 %) felt they lacked access to a mode of transport, and over a quarter lacked access to drinking water (i.e. were more than likely to be using unimproved or unsafe sources of water, like rivers/streams). A total of 22 out of the 26 items were considered by a majority of respondents to be essential. Of these 22, need satisfaction data were available for 16, and it is these which form the basis of the deprivation index, where responses reporting that needs were “not at all satisfied” counted as deprived (scoring 1) and other responses scoring 0. Scores were summed to make a final deprivation index, with a maximum score of 16 and a minimum score of 0.
Table 1 List of items respondents considered essential
Establishing Consensus
The Consensual Approach firstly identifies publicly perceived necessities and then proceeds to find out who lacks these, so to move confidently from the first to the second stage it is important to demonstrate consensus about the list of items in the deprivation index. While there is no reason to assume people will agree on what items take priority over othersFootnote 8 it is important to establish horizontal agreement, i.e. that different demographic groups all agree that a particular item is essential or a necessity.
One way to demonstrate this is through the use of heat maps, where respondent’s answers are shaded; items receiving a higher prevalence of positive responses (e.g. thinking that access to drinking water is essential) are shaded darker, with those with lower prevalence a lighter shade. Table 2 shows the proportion of respondents considering an item essential, by their age and gender. Tables 5, 6 and 7 show the degree of consensus by respondent’s level of education and migrant status, religion and ethnicity, and (for the sake of conciseness) are provided in Appendix “2”. What each table clearly shows is the high degree of horizontal consensus; i.e. what younger respondents think essential is very similar to what older respondents report; what women think are essential are also likely to be thought essential by men, etc. What (slight) differences there may be can be explained on a case for case basis (e.g. religious or cultural prohibition).
Table 2 Heat map of attitudes to items considered “essential”, by respondent age and sex (%). (Color figure online)
Item Validity
Table 1 showed that a total of 22 items out of 26 asked about were considered by more the 50 % of respondents to be “essential”. Of these 22, follow-up questions were asked as to whether respondents felt their needs were met with regards 16 of the items. Table 1 also shows the proportion of respondents reporting that their needs were not at all satisfied. So while 84 % of people thought having access to water was essential, 26 % felt this need was not satisfied at all, suggesting a high level of deprivation of a socially perceived necessity (water). The 16 items regarding need-satisfaction form the basis of a deprivation index: respondents score a 1 for each item of which they are deprived. Respondents could have a minimum score of 0 and a maximum of 16.
When constructing deprivation indices, it is important that each item be both a reliable and valid measure of poverty (Gordon 2006). The overall reliability of the scale is discussed in Sect. 4.3; here we show how the validity of each item was tested against four measures known to relate to poverty. Four different validators were used:
-
(1)
Respondent’s evaluations of their household income status: the probability of being deprived for those who thought household income status “difficult” was compared to the probability of those who thought their household income status either “good” or “more or less OK”;
-
(2)
Respondent’s evaluations of their current financial situation: the probability of being deprived for those going into debt was compared to the probability for those who were able to save either a little or a reasonable amount;
-
(3)
Respondent’s evaluations of the stability of their household income: the probability of being deprived for those considering their household income unstable compared with those for whom household income was considered stable;
-
(4)
Respondent’s quintile on the DHS household wealth index: the probability of being deprived for the bottom 20 % was compared to the top 20 %.
In each of the 64 instances (i.e. 16 items × 4 validators), the probability of being deprived was significantly greater for those known to be disadvantaged compared with those who were not. So, for example, with regards validator 1, respondents who felt their household income status to be ‘difficult’ were nearly 15 times more likely to feel that their needs for the requisite number of meals each day were not at all met, compared to those whose household income status was reported as good or more or less OK. Results for each validator are presented in Appendix “3”. There is clear face validity for the index, given that the items which go into making it up are those which relate to people’s everyday living conditions and their needs for clothing, food, health care and transport.
Scale Reliability
Scale reliability was tested using Cronbach’s Alpha, and was found to be high, with an alpha of 0.885 (Table 3). This can be interpreted as the average correlation between this set of 16 questions (i.e. relating to items in the deprivation index) and all other sets of deprivation questions of equal length (i.e. in this instance, 16 items) (Nunnally 1981; Devellis 2003).
Table 3 Scale reliability
Scale Validity
The validity of a scale can be assessed by seeing if it exhibits statistically significant associations with a set of independent variables known to be correlated with poverty (Pantazis et al. 2006b). For example, it would be expected, from Townsend’s theory of relative deprivation and Mack and Lansley’s concept of “consensual poverty”, that someone who is ‘deprived’ would also be more likely to consider her/himself to be subjectively poor (Bradshaw and Finch 2003), to have a lower level of household resources or assets, or have an unstable income or household financial situation. We tested the deprivation index for Benin against the four validators described above, each of which we know are correlated with poverty. In each instance, the mean deprivation score was highest (i.e. signifying a higher level of deprivation) for respondents identified by the validators as being in the worst circumstances (Fig. 1).
Prevalence of Deprivation in Benin
Having explained how items were identified and the deprivation index or scale developed, Fig. 2 shows the pattern of deprivations across the whole sample. While just over a third (36 %) of respondents reported that their needs with regards the 16 items on the deprivation index were met to one degree or another, and thus were classified as not experiencing any deprivations, around two-thirds felt their needs for at least one item were not at all satisfied. The pattern is as one would expect, with the proportion of respondents deprived decreasing as the numbers of deprivations increase. Around 6 % of respondents were deprived of ten or more items.
Poverty/Deprivation Threshold
Townsend (1979) showed there is a clear relationship between the resources people have, and their ability to avoid the consequences of poverty, deprivation. Previous studies (Gordon and Pantazis 1997c) of poverty using the consensual approach have used household income as a measure of resources which people use to protect or cushion themselves against deprivation, and each show there is a point on the distribution below which the experience of multiple deprivations increases much more rapidly. Below this set level of resources (income or other), people are no longer able to satisfy their basic needs, and the result is multiple deprivation and undeniable poverty.
The DHS data for Benin do not collect data on household income or expenditure so we cannot do a similar exercise to identify the kink or threshold. What we can do, however, is use the raw scores of the DHS wealth index (Rutstein and Johnson 2004) as a proxy for household resources. The asset index uses information about household assets (e.g. ownership of land, vehicles, consumer durables, etc) and the provision of basic services (e.g. access to electricity, piped water, sanitation) to provide households with a score on a continuous scale. These scores can be used to rank households in a distribution, or grouped into categories, like quintiles. As one would expect, there is a clear relationship between the asset index score and deprivation, with respondents experiencing no deprivations having significantly higher asset index scores (Fig. 3). Below a certain point of household resources (on the y-axis), the number of deprivations experienced (on the x-axis) increases sharply, and this is where one would consider setting a poverty line or threshold (if one was using income on the y-axis). In this instance, based on a visual assessment, we would consider respondents experiencing four or more deprivations to be below the asset-index based poverty line, and would consider all such households as poor. ANOVA and Logistic Regression analyses proposed by Gordon (2006) to identify thresholds in the relationship between income and deprivation suggest a threshold of three or more; however, given that income and the Wealth Index are conceptually very different measures, we err on the side of caution and we set the poverty threshold at four or more deprivations. Such households accounted for just under one-third (31 %) of households in Benin, which is similar to the proportion of households living below the national poverty line of Benin, of 33 %;Footnote 9 were the threshold set higher, at 5+ deprivations, 23 % of households would be classed as poor. Just over a third (36 %) of households did not experience any deprivations.
Country Profile of Multiple Deprivation
Table 4 shows how multiple deprivation is patterned across different geographic regions and socio-cultural groups in Benin. Focussing solely on those experiencing 4+ deprivations, prevalence rates in rural areas are twice those of urban areas. There are considerable regional differences, with nearly half of households in Collines experiencing 4+ deprivations; this is in contrast to one in eight households in Littoral. As expected, the prevalence of multiple deprivation is highest for those in the poorest wealth index quintile, with a clear gradient apparent. The fact that 7 % of households in the top quintile report experiencing 4+ deprivations suggests the wealth index is classifying some deprived responds as relatively wealthy, which we consider problematic. Researchers have questioned the methods used to create the wealth index and its ability to make meaningful or reliable comparisons between countries or over time (Falkingham and Namazie 2002; Howe et al. 2008). Our analysis shows the wealth index may identify as wealthy some very deprived households. That said, in low income countries like Benin, where deprivation with regards some basic needs (e.g. access to electricity, or to safe drinking water) is generally high, such a finding is not entirely unexpected.
Table 4 The patterning of multiple deprivation in Benin, 2006 (%)
There appears to be little difference between most age groups, although older respondents (aged 65+ years) do have higher than expected prevalence rates of 4+ deprivations. Given data were collected at household level we cannot comment or assess the extent of intra-household poverty or inequity or comment on gender differences in poverty rates. That said, such issues could be addressed in further work, which would employ an individual-level questionnaire, as has recently been done in the UK 2012 Poverty and Social Exclusion Survey. In terms of ethnicity, it appears respondents from Betamari and related groups are worse off than their compatriots, while those from Dendi and Bariba groups experience relatively low rates of 4+ deprivation. The expected relationship between education and multiple deprivation is confirmed, with a clear gradient apparent (i.e. those with no education have far higher rates of 4+ deprivation).