%20-%20https://sketch.com%20--%3e%3ctitle%3e3ECBBF50-FD08-412A-943D-2CFC5E025A16%3c/title%3e%3cdesc%3eCreated%20with%20sketchtool.%3c/desc%3e%3cg%20id='Stylesheet'%20stroke='none'%20stroke-width='1'%20fill='none'%20fill-rule='evenodd'%3e%3cg%20transform='translate(-926.000000,%20-668.000000)'%20id='BUTTONS'%3e%3cg%20transform='translate(100.000000,%20597.000000)'%3e%3cg%20id='BTN-/-RIGHTS'%20transform='translate(816.000000,%2065.000000)'%3e%3cg%20id='ICON-/-Rights'%20transform='translate(10.000000,%206.000000)'%3e%3cpath%20d='M13.5752841,0%20L1.34943182,0%20C0.59375,0%200,0.586666667%200,1.33333333%20L0,22.6666667%20C0,23.4133333%200.620738636,24%201.34943182,24%20L17.6505682,24%20C18.40625,24%2019,23.3866667%2019,22.6666667%20L19,5.36%20L13.5752841,0%20Z%20M13.5752841,1.89333333%20L17.0838068,5.36%20L13.5752841,5.36%20L13.5752841,1.89333333%20Z%20M1.34943182,22.6666667%20L1.34943182,1.33333333%20L12.1988636,1.33333333%20L12.1988636,5.36%20C12.1988636,6.10666667%2012.7926136,6.69333333%2013.5482955,6.69333333%20L17.6235795,6.69333333%20L17.6505682,22.6666667%20L1.34943182,22.6666667%20Z'%20id='Shape'%20fill='%23595959'%20fill-rule='nonzero'%3e%3c/path%3e%3ccircle%20id='Oval'%20stroke='%23595959'%20stroke-width='1.5'%20fill='%23595959'%20cx='9.5'%20cy='13.5355339'%20r='5.5'%3e%3c/circle%3e%3cpolyline%20id='Path-2'%20stroke='%23FFFFFF'%20stroke-width='1.5'%20transform='translate(9.500000,%2013.035534)%20rotate(45.000000)%20translate(-9.500000,%20-13.035534)%20'%20points='8%2016.0355339%2011%2016.0355339%2011%2010.0355339'%3e%3c/polyline%3e%3c/g%3e%3c/g%3e%3c/g%3e%3c/g%3e%3c/g%3e%3c/svg%3e)
Introduction
For several decades, populations have been counted through national, usually decennial, censuses in which enumerators go to households. Inter-censal population estimates have usually depended on reliable birth and death registration systems. Assessing the numbers living in deprived neighbourhoods, the levels of relative deprivation and providing much of the data required for planning purposes have historically been derived from censuses or inter-censal estimates.Footnote 1
In the more developed countries, there is increasing reliance on data linkage using an identity number, or substituting administrative records such as municipal population registers and regular surveys for the once-in-a-decade census, which is seen as cumbersome, rapidly out-of-date and increasingly difficult to implement. Among forty European countries, the number using the traditional census method has declined from twenty-seven in 2001 to twenty-one in 2011, while the numbers using registers and ‘total’ enumeration via sample surveys has increased from nine to sixteen (Valente, Reference Valente2010). The UK moved away, in the 2001 Census, from direct enumeration to self-report, and this continued in 2011; this is equivalent to aiming for a 100 per cent sample household survey, supplemented in various ways for those not in households (see below).
Census organisations in developed countries recognise the difficulties of counting non-household groups. There is much less recognition, both by survey organisations themselves in the UK (although see Joloza, Reference Joloza2009) and in international official publications, of the possible implications for bias in the sampling frames of household surveys (Yansaneh, Reference Yansaneh2005). This is developed in the next section.
In the third section, evidence is presented about missing population sub-groups and their likely poverty or deprivation status in both developing countries and in the UK, as well as the implications for the coverage of household surveys. The fourth section focuses on the UK, showing how this affects the conclusions of recently published analyses of inequalities, and the validity of formulae for allocating resources to health and social care.
Coverage of censuses
This section shows how censuses may not always provide a complete population count, and how these omissions impact directly on assessments of the numbers who are poor.
Census coverage problems
Population censuses have always faced problems of incomplete enumeration. Non-citizens, cultural minorities or marginalised groups and specific categories of prisoners or rebels have often been excluded for political reasons (Buettner and Garland, Reference Buettner and Garland2008). Although this may now be less frequent and is certainly more transparent, there are still several examples in developing countries and even in developed countries. For example, the status and counting of illegal Mexican immigrants in the US was in dispute in the recent election.
The general problem that censuses are not themselves necessarily complete is therefore well understood, and guidelines for what constitutes good census practice have been developed (UN, 2008). Although the guidelines are clear, there can still be problems with enumeration in practice with the precise definition of many important sub-groups.
Housekeepers and servants: Whilst Cinderella is a fairy tale, the exclusion from the census count of undeclared poor servants in rich households in Asia is not. In the UK, domestic migrant workers may be paid as little as 50 pence an hour and the prohibition from changing employer, in force since April 2012, means that their status is effectively that of a bonded labourer (Kalayaan, 2012). They are very unlikely to be counted in either a census or a household survey.
Mobile populations: In developing countries, they may well be among the poorest, migrating from rural to urban areas in search of a livelihood (Pincus and Sender, Reference Pincus and Sender2008). In developed countries like the UK, the young, highly mobile – usually male – populations are also difficult to count (OPCS, 1993), but many of them will be relatively well-off. For different reasons, young men in collective households are under-enumerated in censuses in many countries; and the majority of the 1.2 million estimated to be missing in the 1991 UK census count are obvious examples (McLean and Smith, Reference McLean, Smith, Heath, Jowell and Curtuce1994; Mitchell et al., Reference Mitchell, Martin, Dorling and Simpson2002).
Homelessness and counting the actual rather than the legal populations: These are always difficult to count. Most developing countries do not have reliable data on the numbers of homeless people, and, in some developing countries, people who are internally displaced because of civil war, those exposed to extensive environmental damage and street sleepers are ignored for census purposes because they have no official address (CARDO, 2003). Counting the number of gypsies or travelers in developed countries is also more complicated simply because they are nomadic.
Institutional populations: There are several different types of communal establishments (care homes, factory barracks, hospitals, the military, prisons, refugee camps, religious orders and school dormitories). In developing countries, these groups may not be included at all in census population counts; and even in developed countries there is still little agreement over how some of these groups should be included in the population count. For example, in the US, some States count prisoners where they are incarcerated but others at their last home address (Wagner, Reference Wagner2008). If counted where they are incarcerated, the Census Bureau counts them as special census blocks or special households and their characteristics are not fully reported.
Careful reporting of censuses as per the guidelines acknowledges how well these groups have been enumerated and most categories are included in estimated census population counts of developed countries, but not in the census reports of many developing countries. In the UK, special arrangements for the census were made to count people living in communal establishments, such as boarding schools, prisons, military bases, hospitals, care homes, student halls of residence, hotels, royal apartments and embassies, as well as for particular communities such as rough sleepers, travelers and those living on waterways.
The UK Office for National Statistics (ONS) also carries out post-enumeration surveys to make an estimate of the under-count of households and people (Abbott and Marques de Santos, Reference Abbott and Marques dos Santos2007), although there was no assessment of their success in estimating the non-household population. In general, however, post-enumeration surveys are not very successful in their objectives of identifying missing or duplicated people because they have ‘to secure participation by the people who were missed by the Census, and it has to be very accurate in matching individuals counted by the sample survey with individuals counted by the Census’ (Darga, Reference Darga1998). This is well-nigh impossible: there are too many people who do not want the government to know where they are and it will never be easy to match the results of the post-enumeration survey with the Census because of unreliable or fabricated interviews, or ambiguity or misreporting of usual residence (Breiman, Reference Breiman1994). The results for more recent post-enumeration surveys suggests that there has not been any improvement (Stark, Reference Stark, Beck, Bryman and Liao2004); and the US Census Bureau decided not to adjust the 2000 census count (Ronzio, Reference Ronzio2007).
Using censuses and register data to estimate poverty levels
Measurement of poverty levels
There is a huge literature on how to measure poverty on an international comparative basis spanning several decades (OECD, 1975; McGranahan et al., Reference McGranahan, Richard-Proust, Sovani and Subramanian1972; ILO, 1976) and several disciplines. In the UK itself, there is a long history of measuring poverty in sophisticated ways dating back to Rowntree's first survey in 1899 (Veit-Wilson, Reference Veit-Wilson1986).. But, in the broader context of assessing development progress, for better or for worse, a crude cash measure (US$1 or US$2 a day) has been adopted by most international organisations as the flagship measure, even though it ignores non-food needs, which are mostly monetised in urban but not in rural areas (Mitlin and Sattherwaite, Reference Mitlin and Sattherwaite2012). The argument in this paper, however, applies not only to that crude cash measure, but also to many other dimensions of deprivation, such as lack of education, ill-health, mortality, under-nutrition, poor sanitation, lack of clean drinking water, etc.
Do official statistics cover the poor?
For developing countries, the census documents for several of the large-population countries have been examined for any commentary about omissions, difficulties or problems encountered. The only omissions acknowledged are institutional populations (by Brazil), insecure areas (by Indonesia and Kenya) and foreign officials and refugees (by Thailand). There is very little commentary in their reports. An extensive web search carried out for other commentary resulted in several news items – often substantial – about reported difficulties in implementation, but very few academic articles, of which most were about Bangladesh (Angeles et al., Reference Angeles, Lance, Barden-O’Fallon, Islam, Mahbub and Nazem2009), India (e.g., Agarawal, Reference Agarawal2011) and Vietnam (Pincus and Sender, Reference Pincus and Sender2008).
In European Union countries, Gerull (Reference Gerull2009) shows that, although the 2011 Census round aimed to gather ‘reliable, detailed and comparable data on the population and housing’ (European Union, 2008: preamble paragraph 1), and recognised the difficulty of counting the homeless in particular, register-based censuses cannot identify and count those living on the street.
Coverage of household surveys and implications for poverty measurement
Overall levels
The problem in assessing the overall level of poverty or of deprivation is that household surveys are an inappropriate instrument for obtaining information about the poorest of the poor, especially in developing countries. This is because household surveys, with rare exceptions, typically omit by design:
-
1. those not in households because they are homeless;
-
2. those who are in institutions (including refugee camps); and
-
3. mobile, nomadic or pastoralist populations.
In addition, household surveys will typically under-represent many of those in fragile, disjointed or multiple occupancy households (because of the difficulty of identifying them), those in urban slums (because of the difficulty of identifying households and interviewing individuals) and may omit certain areas of a country because they pose a security risk. If one wanted a practical – as distinct from a theoretical – definition of the ‘poorest of the poor’ in developing countries, the above population sub-groups could hardly be bettered.
These problems for household surveys also occur in developed countries, but to a lesser extent: there are fewer homeless; the geographically mobile tend to be the young and upwardly socially mobile rather than the income poor; there are fewer fragile and multiple occupancy households; and urban slums are much smaller. The only real issuesFootnote 2 in developed countries are the omission of the traveller population and of the institutional populations from household survey sampling frames.
The lack of recognition of these coverage problems, when using estimates from household sample surveys, particularly although not exclusively in developing countries, is that there has been no systematic attempt to estimate the size and distribution of the ‘missing’ population groups.Footnote 3
In developing countries, because of the lack of statistical infrastructure, it is difficult to make precise estimates of numbers in these groups. In the next section, drawing on an extensive literature and web search, the likely magnitudes of the ‘missing’ populations in developing and developed countries are outlined, together with a reasonably precise estimate for the UK.
Comparative levels
The interest for the planner, however, is often how to distribute or redistribute resources appropriately between regions or across sub-groups of the population. There is no theoretical reason to believe in general that there will be more or less under-counting between regions or across sub-groups, so a planner could assume that the comparison of the estimated values was a reasonable approximation to the true relativities. In many cases, however, there are country-specific reasons to believe that there will be different levels of under-counting between geographical regions, so that comparison of the estimated values will be a biased estimate of the true relativities.
Counting the poor or deprived that are potentially ‘missing’ from sampling frames of household surveys
Globally
There are several groups that may be excluded from censuses or the sampling frame of household surveys: some may be richer than the average, which biases the estimated poverty rate upwards, and some will have an income distribution similar to the rest of the population rather than the majority being poor. For example, those caught up in civil wars will usually be a cross-section of the population, although, among those fleeing a country, there may be a small bias towards the rich. However, even if registered in the destination country, they would be excluded from a census or the sampling frame of a standard household survey. Economic and environmental migrants (Myers, Reference Myers1997) may include the more ambitious and therefore not the poorest (but not necessarily the richest either). Enumeration conventions may exclude temporary immigrants or non-nationals in censuses, who may not be well-off. There is rarely a credible source to estimate their numbers and their wealth and only if most of a ‘missing’ group was large and nearly all in either the top or the bottom income or wealth quintiles would this substantially bias the estimated poverty rate.Footnote 4 The focus here is on groups for which there are credible sources, and that are normally among the poorest.
Homeless
Rather obviously, household surveys omit the homeless and street children.
Adults: UN-Habitat (2003) estimated the number of homeless people worldwide to be between 100 million and one billion, depending on how they are counted and the definition used. The difference in estimates is essentially a distinction between those without any roof over their heads at all (the smaller estimate of 100 million), who will almost certainly be omitted from all household surveys, and the much larger numbers in informal – usually illegal – squatter settlements with no security of tenure and at risk of immediate eviction. Agarawal (Reference Agarawal2011) documents the scale of the official under-count of ‘slums’ in cities in India; and it is common in many developing countries for construction workers and their families to live in temporary camps close to their construction sites (CARDO, 2003). Censuses will, and thorough household surveys will usually attempt to, cover informal settlements, but with varying degrees of success.
Children: Estimating numbers is fraught with difficulties. Over twenty years ago, UNICEF estimated that there were about 100 million street children (UNICEF, 1989) and has repeated that figure on several occasions. A recent thorough review (de Benitez, Reference De Benitez2009) concluded that the real number – depending on the definition – is much lower. Many would be counted in conventional censuses – and carefully designed household surveys. where informal settlements are included if the children live with their families, but work on the streets. Equally, many will not be included, but it is very difficult to make an estimate.
Institutionalised populations
Most categories are included in the estimated population counts of developed countries. Household surveys sometimes attempt to include institutions, but sampling frames may not exist where they are reported in aggregate as a special census block. One could use the aggregate number as a sampling frame – as long as they are in one location – but this rarely happens.
Care homes and hospitals: Those in hospitals will mostly be poorer because morbidity is associated with poverty worldwide (e.g., Lopez, Reference Lopez2002) and hospital care is costly.
The World Health Organisation (WHO, 2011) reports twenty-nine beds per 10,000 population globally, ranging from nine in the African region to sixty-two in the European region, with several countries not reporting. While occupancy rates in developed countries are typically between 80 per cent and 90 per cent, hospitals in developing countries are often very overcrowded with patients doubling up or sleeping on floors. A reasonable estimate would be that there were 25 million missing on any one day from household surveys worldwide (based on 18.9 million beds and an average overcrowding rate of 30 per cent). Most would have been included by the respondent in a household survey; but not single persons, because there is no respondent.
Military: Those in the military are not included in the household surveys of any country. The latest CIA (2012) estimate is that there are about 92 million in the military of the world; their estimate for the UK is 400,000.
Prison: Walmsley (Reference Walmsley2009) estimates the total prison population of the world at about 9.8 million, mostly as pre-trial detainees (remand prisoners) or as sentenced prisoners. Almost half of these are in the United States (2.29 million), Russia (0.89 million) or China (1.57 million); these are counted in population censuses (although not always enumerated individually). Those in prisons will usually be poorer. In the United States, with 5 per cent of the world's population but 20 per cent of the world's prison population, the incarceration rates in 2008 among African-American drop-outs from high school was 37 per cent compared to less than 1 per cent of the general population (Western and Pettit, Reference Western and Petit2010).
Refugees
Official refugees will not routinely be counted in annual national population censuses because they are not considered as part of any nation's population. In the UK, however, official refugees and asylum seekers not in detention centres will be part of the household population. The United Nations High Commissioner for Refugees (UNHCR) publishes figures annually on numbers of registered refugees, internally displaced persons and stateless persons. The overall world total is 36.5 million, but these figures do not include the large number of illegal immigrants. For the US, Passel and Cohn (Reference Passel and Cohn2010) estimated there were 11.1 million unauthorised (mostly Hispanic) immigrants in the US, of whom 10–15 per cent would be omitted from any count (including national population counts) and, of course, from the sampling frames of household surveys. Furthermore, some of the remaining 9.5 million would be reluctant to answer surveys.
Highly mobile, nomads and pastoralists
In developing countries, there is no reliable information available on the number of nomadic pastoralists, including sea-faring mobile communities, worldwide. There are various country-specific estimates, based on a wide variety of different definitions. The only internationally comparable source is that for all pastoralists (including sedentary pastoralists) compiled by the International Livestock Research Institute (see Thornton et al., Reference Thornton, Kruska, Henninger, Kristjanson, Reid, Atieno, Odero and Ndegwa2002), based mostly on rangeland livestock numbers, and these are larger than other estimates of (mainly) nomadic pastoralists. Based on estimates from work in the Horn of Africa (Carr-Hill, Reference Carr-Hill2005), between one in ten and one in five of rangeland nomadic and sedentary pastoralists together will be missing from population counts, giving an estimate in 2011 of between 21.8 and 43.5 million missing. While it is difficult to assess their income and wealth, and there are some who are rich-in-kind (or asset rich), the majority are usually poor in all senses.
Difficult to reach
Fragile and disjointed households: There are of course several problems in defining household units in social surveys, especially in developing countries, because the Eurocentric definition assumes that the residential, production and consumption units are the same (van de Walle, Reference Van de Walle2005).
Where the household structure is ambiguous, the task of the census enumerator or survey interviewer in either identifying the household head or reference person and/or counting the numbers in the household is more complicated.Footnote 5 In AIDS-affected developing countries, these have included elderly household heads with young children; grandparent households; large households with unrelated fostered or orphaned children attached; child-headed households; single-parent (mother or father) headed households; cluster foster care, where a group of children is cared for formally or informally by neighbouring adult households; itinerant, displaced or homeless children; neglected, displaced children in groups or gangs, etc. There are several ad hoc surveys, but no systematic way of identifying or counting these different types of household, and they are usually excluded from national household surveys.
Urban slums: UN-Habitat defines a slum household as ‘. . . one or a group of individuals living under the same roof in an urban area, lacking one or more of the following four amenities: (1) durable housing; (2) sufficient living area; (3) access to improved water; (4) access to improved sanitation facilities’. Recent estimates (UN-Habitat, 2011) are that there are more than a billion living in urban slums in developing countries, but the poorest urban populations are often simply not included in data gathering. In particular, the estimates of relativities between areas within urban conglomerations will often be an under-estimate of the true relativities, but currently it is impossible to know by how much, or how this varies between countries, or over time.Footnote 6
Data sets such as Demographic and Health Surveys (DHS) disaggregate by ‘urban’ and ‘rural’, but go no further. Thus, slum populations and the poorest squatters are statistically identical to middle-class and wealthy urban dwellers. Worse, the poorest urban populations are often not included at all in data gathering. (UN-Habitat, 2003: emphasis added)
A small minority of households in the slum areas of developing countries’ cities are therefore uncounted in any census (and therefore not included in the UN-Habitat database) simply because of the difficulties in enumerating in chaotic urban slum conditions. Moreover, even where they are counted in censuses, many would in practice (partly because of the same chaotic conditions and partly because of interviewer reluctance) be excluded from achieved samples in surveys. A range of estimates have been made, assuming that between 1 in 10 or 1 in 5 of the urban slum populations are uncounted (which are probably conservative estimates, given UN-Habitat's remark). These imply that there are between 43.5 and 87.1 million missing from the sampling frame of household surveys.
Some censuses – such as those in Bangladesh and India – have made special efforts to comprehensively cover those in slums (Schurmann, Reference Schurmann2009), but these are not typical and, because they are usually special counts of the numbers in specific marginalised groups, are not part of the sampling frames of household surveys.
Security: Given the security situation – or simply difficulty of transport – in many countries, it can often be difficult for the implementing institutions to carry out a fully representative survey or census. This is frequently a problem in developing countries, but not in general with developed countries and certainly not in the UK.
UK
Only some of the categories examined above are relevant for developed countries, and these are considered below.
Homeless
There are few ‘informal settlements’ in developed countries, and the only issue is the number of ‘actual’ homeless. In the UK, the law defines someone as being homeless if they do not have a legal right to occupy any suitable accommodation (Shelter, 2007: 2). The local authority has a duty to those who are eligible for assistance, legally classified as homeless (as above), in priority need (e.g., pregnant women or those who are responsible for dependent children) and are unintentionally homeless (Shelter, 2007: 3). In 2014, 59,700 households – who were assessed as being in priority need and unintentionally homeless – were in temporary accommodation (meeting the above criteria and waiting to be re-housed). In addition, there are an estimated 1,250 rough sleepers (Homeless Link, 2011). Research for Crisis (2004), however, estimated that there were about 350,000 households who are actually homeless but find a temporary solution by staying with family or friends, and so are omitted from official figures.
Hospitals and care homes
In the UK, those living alone who are in hospital on census night are included in the UK population counts as being resident at the postcode of that hospital, but those living alone in hospital on the day of the household survey will be excluded. In England and Wales in 2010–11, there were 15.5 million Finished Consultant EpisodesFootnote 7 among fifteen-plus-year olds. Of these, 10.5 million were overnight in-patients with a mean length of stay of 5.3 days (NHSIC, 2012), implying about 160,000 adults in hospital on any one night. Given that 29 per cent of adults live alone (Census, 2011), there are a minimum of 48,000 single persons in the sampling frame of a household survey who will be wrongly assigned to the non-response category.Footnote 8
Those in nursing or residential care homes, although counted in population estimates, will also not be included in the sampling frames of household surveys. This will only be a major issue in ‘developed’ countries and its importance will vary according to patterns of family care: This means that, while there are about 418,000 in publicly or privately provided care homes in the UK (Moorfields, 2011), including about 4.2 per cent of the sixty-five-plus population, residential care homes in Greece serve only an estimated 1–2 per cent of the population above sixty-five years of age (PEMFI, 2012).
Military
The Ministry of Defense (MoD, 2012) reports the numbers of service personnel at 179,800 plus an additional 71,000 civilians, with 22,400 and 11,300 respectively serving abroad. The 57,700 civilians in the UK are clearly in households; 42,500 of personnel are in service accommodation together with their families (estimated at an additional 100,000); 85,000 are in barracks and 30,000 are living at home. The total missing from sampling frames of household surveys will therefore be 107,700 (service personnel serving abroad or in barracks).
Prisons
The prison population in the UK in 2011–12 was about 85,000 (MoJ, 2012). The data linkage project between the Department of Work and Pensions and the Ministry of Justice, showed that ex-offenders were much more likely to be claiming out-of-work benefits than the general population (MoJ and DWP, 2011).
Refugees
In the UK, there were about 238,000 official refugees in 2010 (UNHCR, 2010), about 0.4 per cent of the population, and there may be up to 238,000 unofficial refugees. In any case, it is likely that a refugee – even in the UK – would be more reluctant than most to respond to surveys, and, because of their status, they will be poor.
Mobile
Although there are special census counts in the UK, the permanently mobile are usually excluded from household surveys; gypsies and traveler populations, who have much less access to services, are very rarely included. The Council of Europe (2012) estimated the UK's Gypsy, Roma and Traveller populations to be in the region of 150,000 to 300,000; and the UK Census 2011 included a new ethnicity category for Gypsies and Irish Travelers, which was ticked by 57,680, although this was recognised as being a significant under-count (ONS, 2014).
Children
In the UK, as in other developed countries, most children in very disturbed households are taken into care. About two-thirds are then placed in foster care and so are in households; leaving about 22,000 children in Local Authority care (The Fostering Network, 2012) and these latter will not be in households.
Estimates of numbers missing and of the poorest quintile
Summary of numbers missing
The estimates of the numbers missing from the different population sub-groups are summarised in Table 1, with assessments of their reliability categorised as A for authoritative, B for fairly reliable and C for best guess.
Population groups missing from the sampling frames of household surveys worldwide (millions) and in the UK (thousands)
Notes: n.e. = not estimated.
Sources: Homeless/children: Homeless Link (2011), The Fostering Network (2012); Care homes/hospitals: WHO (2011), NHS Information Centre (2012); Military: CIA (2012), MoD (2012); Prisons: Walmsley (Reference Walmsley2009), MoJ (2012); Refugees: UNHCR (2012); Pastoralists: Thornton et al. (Reference Thornton, Kruska, Henninger, Kristjanson, Reid, Atieno, Odero and Ndegwa2002); Slums: UN-Habitat (2011).
These global estimates neither include the homeless, those in fragile or disjointed households, nor those in areas where there are security risks. It could be argued that the homeless would mostly be from urban slums (and the numbers look plausible) so that there would be double-counting; and therefore the best global estimates are between 268 and 439 million, but the other two categories (large but of unknown size) are definitely additional.
The UK estimate of 1,337,000 is much more precise. It is higher than the difference of 1,017,000 between the estimated national population and the estimated household population (National Equality Panel, 2010), but this is mostly because we have included an estimated 238,000 illegal refugees and 83,000 homeless or children in care. Given the technological sophistication now involved in survey methodology, the total is alarmingly high at over 2.0 per cent of the population of approximately 63.7 million.
Numbers missing and in poorest quintile
Globally
All the categories discussed above will also be poor with the probable exception of officers, say 25 per cent, of the military
UK
Of the groups identified in Table 1 for the UK, the military have average or above average income, so are not considered here. Those in or just released from prison (90,000) have low or very low incomes, as have most (say 75 per cent) of the 238,000 unofficial refugees (Doyle, Reference Doyle2008), and an estimated 80 per cent of the 150,000 gypsies and travelers (Cemlyn et al., Reference Cemlyn, Greenfields, Burnett, Mathews and Whitwell2009), making a total of 398,000. The income situation of the largest groups – those in hospitals or care homes (678,000) – is less clear.
Hospitals: The estimate above of the numbers missing in practice from the sampling frames of household surveys was at least 48,000 out of the 160,000 in hospital on any one night. The relationship between morbidity and poverty is less strong in the UK mostly because the majority of those in hospitals are older, and age is confounded with morbidity, but less so with poverty. Nevertheless, it is reasonable to assume that about a half (24,000) are from the bottom quintile.
Care homes: About 418,000 are in care homes, most of them from single person households, given that nearly half of all sixty-five-plus-year olds are living alone. Of those, 199,000 are council supported and so all their income is taken as contribution to the costs of the care home except for a £24.40 personal allowance, about a third of the average weekly household expenditure on personal expenses (ONS, 2013). This is about right, so long as the Local Council funding covers the full cost of the home, but at least 56,000 are having to pay top-up fees (Passingham et al., Reference Passingham, Holloway and Bottery2013); a reasonable estimate would be that about half are having to contribute to their own care, which puts them squarely in the bottom quintile. Of the 152,000 initially paying fully for their own care, many will have sold their house to pay for their care, but, with current house prices, most will still be paying for their care after the average length of stay of about two years (Lievesley et al., Reference Lievesley, Crosby and Bowman2011) and, given that care home fees are about £15,000 after benefits, are likely to be on median (individual) income. The 44,000 in nursing homes funded by the NHS are less likely to be wealthy;Footnote 9 indeed, it would be reasonable to assume (as with those hospitalised) that a half (22,000) would be in the poorest quintile.
A reasonable estimate is that 600,200 of the excluded are from the poorest quintile of the population made up as follows: 90 per cent of children in care, homeless and of those in prison giving 155,700; 80 per cent of gypsies/travelers, another 120,000; 46,000 in hospitals, 100,000 in care homes; 100,000 of those funded by councils in residential homes; and 75 per cent of the informal refugees or 119,000 (Table 2). This means the poorest quintile of the real population includes about 45 per cent rather than 20 per cent of the 1.337 million excluded. When looking at population income inequalities they would constitute a non-negligible 4 per cent of the poorest quintile, but only about 1.6 per cent of all the other quintiles.
Estimates of the poorest quintiles worldwide and in the UK missing from the sampling frames of household surveys
Notes: n.e. = not estimated.
Sources: Homeless/children: Homeless Link (2011), The Fostering Network (2012); Care homes/hospitals: WHO (2011), HES (2012); Military: CIA (2012), MoD (2012); Prisons: Walmsley (2008), MoJ (2012); Refugees: UNHCR (2010); Pastoralists: Thornton et al. (Reference Thornton, Kruska, Henninger, Kristjanson, Reid, Atieno, Odero and Ndegwa2002); Slums: UN-Habitat (2011). Calculations as discussed in text. Assessments of the reliability of estimates categorised as A for authoritative, B for fairly reliable and C for best guess.
Implications for analysis in the UK
In the UK and in many other developed countries, estimates of income inequality or of poverty are based on household surveys (such as the Family Resources Survey (ONS/DWP, 2012) or the EU Survey of Income and Living Conditions (Atkinson and Marlier, Reference Atkinson and Marlier2010)). The responses are typically re-weighted to the ‘total population’ using census data because of differential response rates from specific population sub-groups (Crockett et al., Reference Crockett, Afkhami, Rafferty, Higgins and Marshall2011). But population censuses are themselves surveys of a kind. A modern census in developed countries includes the mobile (because they refer to those present in the household on a specific day or night), will cover those in institutions and will, if necessary, carry out special counts of the homeless and of traveler populations. But, for practical reasons, census officials often only provide overall counts of these groups, so that characteristics of the individual members are not available for calculating weights. This means that weighting of household surveys based on census counts will itself be biased.
Estimating inequalities
The most recent comprehensive analysis of inequalities is by the National Equality Panel (2010), based mostly on data from household surveys; accordingly, it suffers from the problems of missing populations described above.
Appendix 3 of the CASE report (National Equality Panel, 2010) says that the exclusion of the non-household population will not have a major impact on the general picture, because:
The largest groups that are omitted – those in residential care homes for the elderly and those living in armed forces accommodation – have a range of resources that are not so different from others of similar ages. As a corollary, the data we can use on the household population, while incomplete, can still present a fair picture of the circumstances of the population as a whole. (p. 411)
This is probably true of those in the armed forces – in fact they may be slightly better off except when dead – and for self-funding residents of care homes; but certainly not for gypsies/travelers, those in prisons or just released and those residents of council homes who are either supported or wholly funded by local councils. Their analysis has also omitted those living alone who were in hospital at the time of the survey, and those in nursing homes at the time of the survey. Both groups will be poorer than the rest of the population, simply because serious morbidity is associated with poverty. In addition, even if refugees are in the sampling frame, they would almost certainly be very reluctant interviewees.
Implications for resource allocation in the UK
Given the strong association of morbidity with poverty, and the exclusions of the poor and vulnerable, it is inappropriate to use household surveys as a basis for health care resource allocation to populations, as suggested by some authors (Gordon et al., Reference Gordon, Lloyd, Senior, Rigby, Shaw and Ben Shlomon2001; Asthana et al., Reference Asthana, Gibson, Moon, Dicker and Brigham2004; Stone and Galbraith, Reference Stone and Galbraith2006). The magnitude of the bias in using household surveys varies with the outcome variable considered.
For health service allocations, about half a million of the population identified in Table 1 will have an acute or chronic illness, made up as follows: all the rough sleepers and perhaps 10 per cent of the children in care, making 3,500; all those in hospitals and in nursing homes (204,000); between a third and a half of those in residential homes or between 139,700 and 209,000; none of the military; between a quarter and one-third of those in prisons; unofficial refugees and gypsies / travellers of between 104,500 and 139,000. The grand total is between 441,200 and 555,500, most of whom will be in the poorest quintile.
The recent Person Based Resource Allocation approach to resource allocation, to generate fair share budgets for practice-based commissioning, used several years’ data on hospital episode costs for all those registered with a GP (PBRA, 2010), and generated formulae with much higher predictive power than those using area-based indicators. It did not rely on survey data, but excluded ‘people not registered with a general practitioner (such as homeless persons, prisoners, members of HM Armed Forces)’ (PBRA, 2010: 5), or about 390,000 mostly very poor people.
Resource allocation for social care is more complicated but the Relative Needs Formula (RNF) is the most contentious (Gibson and Asthana, Reference Gibson and Asthana2011). Because local authority records only cover clients, they have to be supplemented by other data in order to assess relative needs at an area level. Population household surveys have been combined with a sample of those about to enter care to generate a predictive equation (Darton et al., Reference Darga2006). The issue is which of those population groups in Table 1 not already in care (children in care and adults in residential and nursing homes) might be considered vulnerable: a third of those in hospitals (16,000); a sixth of those in prisons (15,000); a third of gypsies/travellers (50,000); and a quarter of all refugees (119,000); making a total of 200,000 – much less of a problem than predicting the vulnerability of c.3 million single older people.
Conclusions
Population under-counting was highlighted by Setel et al. (Reference Setel, Macfarlane, Szreter, Mikkelsen, Prabhat, Stout and AbouZahr2007) but there has been no recognition that the under-count differentially excludes the poorest of the poor, which means that ‘poverty reduction’ programmes (e.g., Hashemi and de Montequiou, Reference Hashemi and de Montequiou2011) risk ignoring the poorest of the poor. This blindness is a public scandal and could be affecting between 250 and 500 million in developing countries.
In developed countries, patterns or trends in inequality are a major focus of debate, but, like many other poverty-related discussions, debate tends to focus on the numerator rather than the denominator; indeed, the National Equality Panel (2010), although we have shown that it was not comprehensive, was unusual in dealing with the issue at all. Research on inequalities and poverty nearly always rely on household surveys and, although procedures for developing formulae for resource allocation in the UK to date have been based on census and administrative data, they also suffer from missing population problems.
Biases in estimates of population or of changes in population are ignored; indeed, there appears to be a general reluctance to discuss the issue at all. It is urgent to understand the extent and nature of these biases both for social policy planning and research on inequalities.
Acknowledgements
I am very grateful to colleagues and audiences at several presentations, to the anonymous referees for comments which have made the arguments much clearer, and for suggestions of further literature and references to substantiate the estimates.
