The Basic Necessities Survey (BNS)
The Basic Necessities Survey (BNS) is a method of measuring poverty that is:
- simple to design and implement. The results are easy to analyse and to communicate to others
- democratic in the way that it identifies what constitutes poverty and who is poor
- rights based, in its emphasis on entitlement
The Basic Necessities Survey (BNS) builds on and adapts earlier methods that have been used to measure poverty by measuring deprivation and which emphasise the “consensual” definition of poverty (See below ). However, the BNS is innovative in the way in which individual poverty scores are generated from respondents’ survey responses.
Basic necessities are democratically defined as those items in a survey instrument that 50% or more of respondents agree “are basic necessities that everyone should be able to have and nobody should have to go without“ Items are weighted for importance according to the percentage of respondents who say an item is a basic necessity (i.e. between 50% and 100%). Household poverty scores are based on the sum of the weightings of the basic necessities they have, as a percentage of the total they could have, if they had all basic necessities.
The BNS was developed by Rick Davies in 1997, and implemented by ActionAid Vietnam in 1997, 1998, 1999 and by the Pro Poor Centre (a Vietnamese NGO) in 2006 (See Reports on Its Use below). Its design built on earlier work by poverty researchers in the UK and Sweden (See Related Research below)
Caveat: The BNS is a survey method, but this section does not explain how to do surveys. It only explains the central part of the BNS method
At the planning stage
1. Develop a menu of things, activities and services that could be considered as basic necessities.
- In the BNS, “Poverty is defined as the lack of basic necessities. Basic necessities are those things that everyone should be able to have and no one should have to go without“
- This is a rights oriented view, focusing on perceived entitlements.
- The items on this menu should ideally be identified through consultation with a range of people, including potential survey respondents. Small group discussions may be the best means of generating potential menu items.
- Examples of things, activities and services could include: having a bicycle, going on a holiday for a week once a year, and being able to obtain vaccination for one’s children at a health centre in one’s district.
- The items on the menu should be easily and reliably observed, not vague or subjective, like “high self-esteem”, or “a sustainable livelihood”
- They can include things, activities and services available within the respondent’s community as well as being available within the respondent’s house. For example, a health centre that is open seven days a week
- The items on the menu should range from those that almost everyone might agree are basic necessities to those that few will think are basic necessities, but perhaps more people will think so in the future. For example items like a television or a mobile phone.
During the survey
2. In household survey ask people three questions
- “Which of these items do you think are basic necessities, things that everyone should be able to have and no one should have to go without“
- This can be done by reading items out one at a time, or by asking people to sort cards into two piles, with one menu item written on each card
- “Which of these items does your household have?”
- This can also be done by reading items out one at a time, or by asking people to sort cards into two piles, with one menu item written on each card
- “Compared to other people in this x area (same as area sampled), do you think your household is poor or not poor?”
- This question is optional. It is useful, because it helps to define a poverty line using the BNS survey results, but it is not essential.
After the survey
3. For each item, calculate the percentage of the respondents who say that item is a basic necessity
- Exclude those items where less than 50% of people said it was a basic necessity
- This is a democratic approach. Decision by majority rule. Most people did not consider these items as basic necessities.
- For all the rest give each item a weighting, which is equal to the percentage of people who said it is a basic necessity. This percentage could range from 50% to 100%.
- Add up the weightings for all these items. This figure would represent the raw BNS score for an imagined household that had everything on the menu, which has been democratically defined as a basic necessity.
4. For each actual household, add up the weightings for all the items on the menu (now defined as basic necessities) that they said they had. This is their raw BNS score
- Convert this raw score into a percentage of the maximum possible raw score (step 3 above)
- A household with a low percentage would have very few basic necessities, and a household with a high percentage would have most of them
5. Summarise the results
- Calculate the average BNS % score for all the surveyed households. This describes the average depth or degree of poverty.
- Make a graph showing the frequency distribution of BNS (%) scores, to visualise the distribution of poverty. Typically many will be poor in some respect (lacking a few basic necessities, and a few will be poor in many respects (lacking most basic necessities). However the specific shape of the distribution will vary from community to community, and over time.
6. Optional: Identification of a poverty line, defined as a specific BNS (%) score. This enables a head-count measure of poverty, which can complement the depth-of-poverty measure.
- Identify the percentage of respondents who said they were poor (we will call this x%).
- Move from the bottom of the curve describing the distribution of BNS % scores, (step 5 above) up to the point where now x% of the respondents are on the bottom half of the curve. Read off the BNS % score at that point. This BNS % score can now used as the poverty line. People below it are defined as poor.
- This method assumes that while individual respondents may make errors of judgement about their poverty status, these errors are equally distributed. Some poor people class themselves as not-poor, and some not-poor describe themselves as poor.
- Do different sections of the community have different views about what things are basic necessities? For example, women may have different views to men.
- High levels of agreement were found across age groups and gender in the UK. In South Africa high levels of agreement were found across age groups, gender, urban/rural, families with and without children – statistically significant correlations at level of p<0.01.
- In societies without pervasive mass media, or with more fragmented and isolated communities, there might be less consensus on expectations.
- Do people’s views of what are basic necessities reflect their bounded realities? If they dont know how others live, how can they have such expectations?
- In the UK the expectations of poorest respondents were not significantly different from those of the richer respondents
- In South Afrcica “If people were only influenced by their own circumstances then there would not be so much agreement about what the necessities in life are between the different subgroups in the population” (Wright, Noble, Magasela, 2007)
- Do people limit their view of what are necessities when they they cant achieve them?
- In South Africa “Almost all respondents who lacked a socially perceived necessity said that they didn’t have it and couldn’t afford it. That is, there is very little evidence of people reporting that they had chosen not to possess any of the socially perceived necessities.” (Wright, Noble, Magasela, 2007)
- The 2006 Basic Necessities Survey (BNS) in Can Loc District, Ha Tinh Province, VietnamA report by the Pro Poor Centre and Rick Davies.
- There is also an online PowerPoint presentation of the results of this survey
- Beyond Wealth Ranking: The Democratic Definition and Measurement of Poverty A Briefing Note prepared by Rick Davies (CDS Swansea for the ODI Workshop “Indicators of Poverty: Operational Significance“, held on Wednesday, 8 October 1997 in London.
- The Basic Necessities Survey:. the experience of ActionAid Vietnam. Rick Davies and William Smith, Hanoi, Vietnam September, 1998.
- Tom Clement’s PhD Thesis “Money for Something? Investigating the effectiveness of biodiversity conservation interventions in the Northern Plains of Cambodia” includes descriptions of his use of the BNS there
- The Purus Project Full Monitoring Plans A Tropical Forest Conservation Project in Acre, Brazil (2012) by Brian McFarland from CarbonCo. This report documents the results of using the BNS as one part of a community impact survey amongts 16 Amazon communities. See pages 28-29,
- Tony Kaseke, Crop production Engineer in Democratic Republic of Congo, is using a version of the BNS in 2011, using guidance provided by the World Conservation Society (WCS_Modified_BasicNecessities Survey). Tony looked at the degree of inequality in surveyed communities by constructing a Lorenze Curve from the BNS data.
- Social and Biodiversity Impact Assessment (SBIA) Manual for REDD+ Projects PART 2 – Social Impact Assessment ToolBoxby Michael Edwards, September 2011. This has a section on the use of the BNS (section 8, pages 54-58)>
- Household Surveys—a tool for conservation design, action and monitoring. USAID & Wildlife Conservation Society Technical Manual 4. August 2007. This includes a four page section on how to implement a Basic Necessities Survey. It describes the BNS as a “wonderfully quick and relatively inexpensive way to measure and analyse household level poverty and to track changes in poverty levels over time”.
- PS (23/07/08): The Wildlife Conservation Society are now pre-testing the use of BNS in Cambodia and Guatamala, with a view to possibly using on a wider scale within their international program.
- See their June 2009 PowerPoint – slides on page 7 and 8, on use in Cambodia. They define the poverty line as follows: “If 50% of subjects say an asset or service is a basic necessity then any household that does not have all basic necessities is below the locally defined poverty line.“
- PS (23/07/2013): See also the USAID & Wildlife Conservation Society 2007 guidance note on household surveys, which includes a detailed section the BNS
- FREEDOM FROM HUNGER Mali Poverty Outreach Study of the Kafo Jiginew and Nyèsigiso Credit and Savings with Education Programs Anastase Nteziyaremye and Barbara MkNelly May 2001 RESEARCH PAPER NO. 7. See section 1.0 Clients’ Relative Poverty by Financial Product: Basic Needs Survey (pages 9-11 on the method, pages 17-25 on the results and pages 81-83 for the BNS survey format). The BNS survey was used to find out if a specific credit package (CEE) was better at reaching relatively poorer households than the credit unions’ other financial products. 498 randomly selected clients from different financial products of two credit union networks were interviewed. The method and results are both well described.
- The Katine VSLA Project Baseline Survey Report. “This report was prepared for the CARE SUSTAIN project in Uganda, by Margaret Kemigisa, a Ugandan Consultant, in December 2008. The SUSTAIN project is promoting Village Savings and Loans (VSL methodology in Katine Sub County, Soroti district… The survey aimed at establishing the overall poverty and welfare status of VSLA clients/house holds & non VSLA clients (control group), livelihood sources of different social groupings of VSLA and non VSLA clients, socio-economic characteristics of the groups that benefit from the Katine VSLA project and financial services available in Katine VSLA operational areas….The survey employed the basic necessities approach and used a control group of Non VSLA clients in Kamuda sub-county; it covered a total of 116 VSLA clients and 116 Non VSLA clients. Suggested contact for further information: ” Barnes, Helen” email@example.com
- The main limitation of this use of the BNS was the inclusion of items that were not clearly defined and so could not be reliably observed to be present or absent by different observers . For example, “education”, “markets”, “enough food”, “drugs” etc. Unfortunately, the survey analysis did not include calculation of BNS scores for each household, and any examination of the distribution of BNS scores across households. Not really an example of good practice.
The main documents which informed the design of the BNS in 1997 were:
- Frayman, H (1991) Breadline Britain 1990s. Booklet by London Weekend Television.
- Mack, J., and Lansley, S. (1985) Poor Britain. Allen and Unwin. London.
- Gordon, D., Pantazis, C (eds) (1997) Breadline Britain 1990s. Ashgate Publishers Ltd.
- Hallerod, B. (1994) A New Approach to the Direct Consensual Measurement of Poverty. Social Policy Research Centre Discussion Paper No. 50. University of New South Wales. Hallerod’s “Proportional Deprivation Index” uses weightings from all items, including those seen as basic necessities by less than 50% of respondents.
- Hallerod, B. (1994) Poverty in Sweden: A New Approach to the Direct Measurement of Consensual Poverty. UMEA Studies in Sociology No. 106. Umea University. Umea.
Other more recent documents of interest
- Consensual Poverty in Britain, Sweden and Bangladesh:A Comparative Study Ahmed, A.I.M.U. Bangladesh e-Journal of Sociology, 2007, 4(2), pp.1-22. “The study focuses on the construction of a normative deprivation index for Bangladesh, which in the traditions of Townsend (1979), Mack and Lansley (1985, 1992), Halleröd (1994) and Gordon et al. (2000), underscores items the lack of which would constitute poverty. …The data for this study come from a sample survey of 1,914 respondents, 1,207 males and 707 females, from all over Bangladesh in 2000″
- Findings from the Indicators of Poverty and Social Exclusion Project: A Profile of Poverty using the Socially Perceived Necessities Approach (2008) Gemma Wright. Department of Social Development, Republic of South Africa. NB: Wright et al use Hallerod’s Proportional Deprivation Index, which includes items that less than 50% of the respondents see as necessities. A poverty line cut off point on the PDI score distribution was identified using self-reported poverty status (as with the BNS) but using a method described as “quite technically complex”. See also:
- Towards a Democratic Definition of Poverty: Socially Perceived Necessities in South Africa (2007) Gemma Wright, Michael Noble and Wiseman Magasela. Cape Town: HSRC Press.
- Developing a Democratic Definition of Poverty in South Africa Journal of Poverty, Volume 11, Issue 4 January 2008 , pages 117 – 141
- The use of weighted checklists to assess the performance of services having more than one dimension to their performance. Such as health centres or schools. These can combine measures of expectations and actual performance, and combine both in an aggregate performance score
- Schreiner’s Simple Poverty Scorecard (SPS). See my comparison of the SPS with the BNS
- The Grameen Foundation has subsequently developed, and widely promoted, the Progress Out of Poverty Score, based on Schreiner’s method
- There has been some questioning about the applicability of the Progress out of Poverty Index, particularly from NGOs. “For instance, there is only one PPI Index for India, yet everyone knows that poverty lines are much different in the North vs. South and rural vs. urban”
- The Grameen Foundation has subsequently developed, and widely promoted, the Progress Out of Poverty Score, based on Schreiner’s method
- BBC report of a recent study by Joseph Rowntree Foundation (JRF) includes a video of people in UK classifying items as essentials, or non-essentials. See also another BBC item on the same report, called the Basics of Britain
There may be times when development project staff want to quickly identify the poverty status of a household, without going through a whole survey process. For example, to check the entitlement of a household to be included in an activity, or to check whether their status has changed as a result of inclusion. In these situations all that may be needed is knowledge of whether a household is poor or not, not a more specific measure of their degree of poverty.
One way of doing this is to use a shorter checklist, which is derived from prior survey findings. One way of structuring such a checklist is in the form of a Decision Tree. A series of questions are asked and depending on the answers, the household concerned is classified as poor or not poor. The issue then is how to design such a Decision Tree. One means of doing so is to use a Decision Tree algorithm, usually found within more generic data mining packages such as RapidMiner. I have written a detailed account of the uses of Decision Trees as representations, and how to develop them, here. A Decision Tree algorithm can be used to find associations between survey responses, for example between the items a household said they have, and their overall BNS poverty score. It then represents the results in the form of a Decision Tree.
Here below is a Decision Tree produced by RapidMiner. It was developed and tested in two stages. First, the model (the tree) was developed using a training data set (half the original data set) and then it was tested against the data in the remaining half of the data set (the test data set).
The tree consists of selected items from the BNS survey, which are linked by lines that have one of two values (0 or 1), indicating if the item was possessed or not. The square boxes at the end of each branch are groups of respondents possessing the combination of items in the branch of the tree that leads to them. Ideally they will all be of one type, but many Decision Trees are not perfect and they will often contain a mix. That mix is in effect a probability statement that respondents with those possessions will be poor or not. For example, at the end of the far right branch there are 129 poor and 382 non-poor respondents. In other words, there is a 75% (382) likelihood that respondents with that branch’s combination of possessions will not be poor
The same software is also able to measure the overall accuracy of the whole model, and even see if it can improved by tweaking various parameters. The model below has an overall accuracy of 87%, taking into account the possibility of both false positives (poor but not really poor) and false negatives (not-poor but not really not-poor).