Chapter 7 describes how to estimate essential needs for decency besides food and housing. Since it would not be practical to list, agree on, and price all essential non-food non-housing (NFNH) needs, the Anker methodology uses secondary data from a recent household expenditure survey to help estimate essential needs. The ratio of NFNH costs to food costs from a recent household survey is multiplied by the cost of the living wage model diet. This chapter discusses the considerable variability in how countries measure and classify food, housing and NFNH costs in surveys and statistics, and how to adjust for this variability so that the NFNH to food ratio used to estimate NFNH needs is consistent with the Anker methodology. A list of what is included in each expenditure group in CIOCOP (Classification of Individual Consumption According to Purpose) is included in an Annex.
This chapter describes how to estimate the cost of all essential needs for decency besides food and housing. In our methodology, a preliminary estimate of non-food and non-housing (NFNH) costs is made using recent household expenditure survey data by multiplying the NFNH to Food expenditure ratio in these data by the cost of the living wage model diet. This yields a preliminary estimate of NFNH costs. Subsequently, post checks of the preliminary estimate of NFNH costs are done using primary data collected in the study location to help ensure that a living wage estimate is more normatively based and there are sufficient funds for health care and education which are considered human rights around the world.1 Our approach is a practical compromise between the impractical approach of identifying and estimating the cost of each and every NFNH need of workers and families, and the conceptually problematic approach of most other living wage and poverty line methodologies for developing countries that uncritically use the current spending of households indicated by a recent household expenditure survey to measure all non-food needs. The latter approach is especially problematic in poor countries where many households at present have little left over to spend on NFNH needs.
Countries differ in how they measure and classify NFNH expenditure. It is important for researchers to examine the national classification of household expenditures, and to adjust secondary household expenditure data to make them consistent with how food, housing, and NFNH are measured in our living wage methodology. This also increases cross-national comparability of living wage estimates.
Most, but far from all, countries use the internationally accepted classification for household expenditures COICOP (Classification of Individual Consumption According to Purpose) or a similar classification for household expenditures collected from household budget or expenditure surveys. First-level expenditure groups in COICOP are listed below with an indication in brackets as to which of our three expenditure groups they belong: food, housing, NFNH. See Appendix 7.1 for a more detailed listing of COICOP, and Appendix II in ILO et al. (2004) for a highly detailed listing and description.
Although COICOP forms the basis for most national classifications of household expenditures, many countries do not use COICOP. Many countries use an earlier version of COICOP, and many other countries that use the latest COICOP structure make adjustments.2 It is therefore important for researchers to look closely at how national household expenditures are measured and classified, and then adjust reported household expenditure data to be consistent with our living wage methodology. Some common differences between classifications used by countries and the latest COICOP particularly relevant for our living wage methodology include the following. These are discussed in greater detail in Part II of this chapter.
Part II discusses conceptual and empirical issues that affect how non-food and non-housing (NFNH) costs are estimated in our methodology.
NFNH costs are estimated by multiplying the NFNH to Food ratio based on secondary data by the cost of the model diet and possibly adjusting this preliminary NFNH estimate by post checks based on primary and secondary data:
NFNH = (NFNH/Food ratio from secondary data × living wage model diet cost) + possible post checks adjustments for health care and education and transport
This approach is conceptually different from the way that food and housing costs are estimated in our methodology.
The NFNH to Food ratio increases with household income and economic development. This is because the income elasticity6 of food expenditure is well below 1.0 as predicted by Engel’s law7 (Anker, 2011a) and the income elasticity of housing expenditure is only slightly above 1.0 (around 1.2 according to Seale and Regmi, 2006).8 This means both that the food share of household expenditure should decrease with household income and the NFNH to Food ratio should increase with household income. This also means that the NFNH to Food ratio should usually be higher in urban areas compared to rural areas within countries, because urban areas tend to have higher income.
Table 7.1 and Figure 7.1 indicate the relationship between food share of household expenditure and income per capita for 207 countries and territories. Food share falls on average from 48% in low income countries to 15% in high income countries with the non-food to food ratio increasing on average from 1.08 to 5.67 (Table 7.1). Figure 7.1 shows that this relationship is non-linear. It is clear that Engel’s Law, formulated in 1857, is still relevant in the twenty-first century (Anker, 2011a). At the same time, since there are substantial differences in food shares for countries at similar per capita incomes, it is also clear that one needs to be cautious about using food share as the sole basis for estimating non-food costs as is done in other common methodologies for developing countries. Part of the reason for so much variability between food shares for countries with similar per capita incomes is measurement differences (e.g. data are sometimes only for urban areas; alcohol, tobacco, and eating away from home are sometimes included in the food expenditure group; cost or value of owner-occupied housing is ignored in some countries). Food share is also systematically different in certain types of countries (transition economy countries and island states have different food shares ceteris paribus) as well as significantly affected by household income inequality in a country (Anker, 2011a).
Table 7.1 Average percentage food and non-food to food ratio by development level
Development levela | Percentage foodb | Non-food to food ratio |
Low income | 48 | 1.08 |
Lower middle income | 37 | 1.70 |
Upper middle income | 29 | 2.45 |
High income | 15 | 5.67 |
Notes:
a Countries were equally divided into four groups based on income per capita in PPP.
b Values based on a regression with per capita income percentage urban, income inequality, whether country was transition economy country, and if country was island state.
Source: Anker (2011a).
Source: Anker, 2011.
Figure 7.1 Food share of household expenditure as a function of income per capita in PPP (purchasing power parity), 207 countries or territories
Figure 7.2 illustrates that the negative relationship between food share and per capita income found in the cross-national analysis in Figure 7.1 is also found over time in countries such as Sri Lanka with increasing per capita income. Notice, however, that food share does not continuously fall every year probably due to short-term spikes in food prices in some years.
Notes: Real GDP per capita was 3.5 times higher in 2012 than in 1980.
Source: Government of Sri Lanka, 2015 (Table H1).
Figure 7.2 Falling food share of household expenditures in Sri Lanka, 1980–2012
Table 7.2 provides information for five developing countries on how expenditure shares for food, housing, and NFNH and the NFNH to Food ratio differ with household income. The NFNH share and the NFNH to Food ratio both increase with income in all five countries and they are higher based on average (mean) expenditure (which is greatly influenced by spending of richer households) than for the median household. This means that:
Table 7.2 NFNH to Food ratio, food share, housing share, and NFNH share of household expenditures by household income decile, 5 developing countries where we adjusted data to be consistent with our methodology
Notes: Reported food expenditure was adjusted, when necessary, to exclude cooking fuel, alcohol, tobacco, and part of food eaten away from home. Reported housing expenditure was adjusted to include cooking fuel when cooking fuel was included in food group. Reported NFNH expenditure was adjusted as necessary to include alcohol and part of food eaten away from home. Data are per household except for Cambodia, which is per person.
Sources: Cambodia National Institute of Statistics, 2009 for Cambodia (special tabulation); Oficina Nacional de Estadistica, 2007 for Dominican Republic (special tabulation); 2008 CPI expenditure weights for South Africa (Statistics South Africa, 2009); Vietnam Government (2012) for Vietnam; Government of Maharashtra (2011) for India.
Expenditure patterns are generally quite different in rural and urban areas. Expenditures for housing, transport, recreation and communications in particular are usually higher in urban areas compared with rural areas. In addition, urban areas usually have higher incomes, which affect spending patterns. This means that the NFNH to Food ratio is almost always higher in urban areas than in rural areas.
Table 7.3 provides data on food share of household expenditure for rural and urban areas for 11 developing countries (Anker, 2011a). Even though these data indicate food share and not NFNH share, they are still useful to illustrate how different rural and urban areas are in terms of spending patterns. Food share of household expenditure is higher in rural areas in all 11 developing countries in Table 7.3. The rural-urban difference for these countries is 20.6 percentage points on average. Food share is also lower in capital cities than in other urban areas.
Since the food share of household expenditure differs between metropolitan areas, other urban areas, and rural areas, the NFNH share of household expenditure should also differ by location within countries. Table 7.4 shows how the NFNH share of household expenditure varies for rural and urban areas of Kenya and Vietnam (see data for India in Table 7.2). The NFNH to Food ratio is higher in urban areas compared with rural areas in all three of these countries, 1.25 compared with 0.48 in Kenya, 1.21 compared with 0.74 in India, and 0.92 compared with 0.77 in Vietnam.
There are several important implications of the large rural-urban differences in household expenditure patterns illustrated above.
Table 7.3 Food share of household expenditure for urban and rural areas, 11 developing countries
Notes:
a Food share for capital city only.
b Difference somewhat overstated, because food share for four countries was for the capital city, which tends to have a lower food share compared to other urban areas.
Sources: Anker (2011) based on CPI expenditure weights.
Table 7.4 Percentage distribution of household expenditure shares for food, housing and NFNH for rural and urban areas, Kenya and Vietnam
Notes:
a Expenditures adjusted for our methodology.
b Reported housing expenditure unrealistically low in Vietnam, because cost or value of owner-occupied housing is excluded.
Sources: KNBS (2007), Vietnam Government (2012).
It is important for a living wage to be seen as reasonable by workers, employers, governments and laypersons. For this reason, expenditures for goods or services that many people would feel are unnecessary for a living wage are sometimes excluded when estimating NFNH costs. Excluding expenses that are not needed for a decent living standard helps make it clear that a living wage estimate is frugal and reasonable.9
At the same time, we feel that the number of items excluded should be limited so as to avoid appearing petty, overly moralistic, or culturally insensitive.10 As American President Franklin D. Roosevelt (1944) said, ‘Liberty requires opportunity to make a living – a living decent according to the standard of the time, a living that gives man not only enough to live by, but something to live for.’ We recommend limiting unnecessary expenditures to tobacco, narcotics, ‘excessive’ alcohol consumption, and additional cost associated with owning and operating personal motor vehicles compared with exclusive use of passenger transport when passenger transport is considered acceptable for decency.
The preliminary estimate of NFNH costs in our methodology is based in large part on household expenditures from a recent household expenditure survey. It is not based on normative standards. This means that there is a risk that insufficient funds could be included in NFNH for decency. This is especially likely in poor countries where many and sometimes a majority of people do not have decent health care, education, etc. To help correct for this possible problem, rapid assessment post checks are done for the two NFNH expenditures that are human rights (education and health care). NFNH costs are increased when indicated as being necessary based on rapid assessment post checks. Since there is considerable variability in how countries measure and classify household expenditures and this can significantly affect NFNH estimates, our methodology requires researchers to carefully inspect and adjust the available household expenditure data as needed so that they conform to how our living wage is estimated. This improves cross-country comparability and living wage estimates. It is worth noting that other common living wage and poverty line methodologies do not consider whether or not sufficient funds are included for non-food expenses, nor do they look at how available household expenditure data are measured or classified.
A four-step approach is used to estimate NFNH costs for a living wage. The first step requires researchers to select appropriate household expenditure data from a secondary source – for rural or urban areas and percentile of the expenditure distribution that is felt to best reflect the expenditure pattern for a living wage. Step 2 adjusts the household expenditure data to be consistent with our methodology. Expenses considered to be unnecessary for a living wage are excluded and household expenditure data are adjusted to be consistent with how food, housing, and NFNH costs are estimated in our living wage methodology. Step 3 makes a preliminary estimate of NFNH costs for a living wage based on steps 1 and 2 by multiplying the adjusted NFNH to Food expenditure ratio by the cost of the living wage model diet. In step 4, rapid post checks are carried out for health care and education (and possibly transport), and adjustments are made to the preliminary NFNH estimate when necessary to ensure that sufficient funds are available for these.
Because expenditure patterns are so different in rural and urban areas (see Section 7.7 above), data for rural areas should be used when estimating a living wage for a rural area and urban data should be used when estimating a living wage for an urban area. Every effort should be made to obtain special tabulations when data needed for a location and/or income percentile are not published.
Before using recent household expenditure data to estimate NFNH costs for a living wage, it is necessary to decide which part of the income distribution would best represent typical spending of workers who would earn a living wage. The 40th percentile is generally recommended for most developing countries. The 40th percentile household is poorer than the median (50th percentile) household and has less than average (mean) expenditure. It would be reasonable to use expenditures for the 50th percentile for countries with very high poverty rates, and the 30th percentile for some middle-income developing countries. For further discussion on this, see Section 7.5 above.
7.8.1.1 What to do when household expenditure data are not available by income level In some countries, data on household expenditure by expenditure group are available only for average (mean) household expenditures. But the pattern of household expenditure is known to differ by income, with NFNH higher for mean household expenditure (which is greatly affected by the spending of richer households because every dollar or rupee is counted equally in the calculation of mean expenditure) compared with expenditure of the median household. Analysis in Appendix 7.2 indicates that it is difficult to have hard and fast rules on how to adjust the NFNH to Food ratio when only mean expenditure data are available because there is a good deal of variation in this difference across countries. Despite this variability, making an approximate adjustment to the mean NFNH to Food ratio would be better than making no adjustment. So we suggest reducing the mean NFNH to Food ratio when necessary because of lack of data, by the following percentages: 20% for median, 25% for the 40th percentile, and 30% for the 30th percentile.
Other methodologies typically use household expenditure data without scrutiny. Adjustments are almost always necessary in our methodology. These adjustments are discussed in the remainder of this section. Table 7.5 is a worksheet that can be used to adjust the classification of NFNH expenditures found in secondary data so that they can be used in our methodology. Details about adjustments are discussed in subsequent sections. Table 7.6 provides an example.
7.8.2.1 Adjusting NFNH for food eaten away from home Many people eat meals away from home all around the world. Yet, our living wage methodology assumes that all meals are prepared at home, since food costs are estimated using a model diet. This means that it is necessary to take into consideration that meals away from home reduce the need for home cooked meals before using household expenditure data to estimate the NFNH to Food cost ratio. For example, if 6% of household expenditure in a country was spent for meals away from home and 50% of the cost of meals away from home was for services (such as cooking meals, washing dishes, and serving), other costs (e.g. cooking fuel, electricity, rent, dishes), and profit, then 3% (i.e. 50% × 6%) of household expenditure in this example would be for the food in these meals.
Table 7.5 Worksheet for calculating NFNH to Food ratio before post checks using secondary household expenditure data
Notes:
a Percentage of the cost of meals away from home for the food in these meals varies across countries, especially depending on whether meals are sold in fixed establishments or on street (base assumption is 50% of cost of meals away is for the food in these meals for most developing countries, 70% for Asian type street markets, and 30% for developed countries).
b Additional expenses for owning and operating a private vehicle compared with exclusive use of passenger transport varies by country, especially whether motorbike or car is the norm (base assumption is 50% for developing countries).
National statistical offices differ in how they classify food eaten away from home in their household expenditure statistics. While a majority of countries include food eaten away from home in its own expenditure group (called restaurants and hotels in COICOP), many countries include food eaten away from home within the food expenditure group. This means that how food eaten away from home is classified in national household expenditure data needs to be taken into consideration before the NFNH to Food ratio is estimated in our methodology.
When food eaten away from home is included in its own major expenditure group as in CIOCOP, reported food expenditure should be increased by the value of the food in meals away from home and expenditure for meals away from home should be reduced by the same value. Using the above example, 3% would be added to the reported percentage for food and 3% should be subtracted from the reported percentage for NFNH.
When food eaten away from home is included in the food expenditure group (which occurs in around 22% of countries according to Anker, 2011a), value of services and profit in meals away from home should be subtracted from the reported percentage for food and added to the reported percentage for NFNH. Using the above example, 3% should be subtracted from the reported percentage for food and 3% should be added to the reported percentage for NFNH.
We have conducted ad hoc inquiries in a number of countries to estimate the percentage of the cost of meals purchased away from home that is for the food in such meals. We bought meals from vendors frequented by workers and took these meals with us. We subsequently separated out all of the food items in these meals and weighed each food item. For example, a meal might consist of 10 grams of tomato, 20 grams of chicken, 100 grams of rice, and 20 grams of greens. We then estimated the cost of the meal if it had been prepared at home by multiplying the weight of each item in the meal by its price per gram in our local food market and added 5–10% for spices/condiments and fish or chicken stock depending on what was appropriate for a country. We have found that around 50% of the cost of meals eaten away from home was for the food in these meals in the Dominican Republic, Costa Rica, and South Africa. We found that around 70% of the cost of meals was for the food in street food meals in Vietnam, Cambodia, and China. We found that around 30% of the cost of meals away from home was for the food in these meals in the United States. Researchers should use a percentage for the cost of food in meals eaten away from home that is reasonable for their location.
7.8.2.2 Adjusting for other expenditures included in food expenditure group It is common for national statistical offices to include items besides food in the food expenditure group. When this happens, it is necessary to reduce the reported percentage spent for food and increase the percentage for the appropriate expenditure group. Alcohol and tobacco are often included in the food expenditures group in household expenditure data. Cooking fuel is sometimes included in food expenditure. None of these items are included in a living wage model diet. When alcohol, tobacco or narcotics are included in the food group, this should be subtracted from food expenditure and added to NFNH expenditure. When cooking fuel is included in the food expenditure group, it should be subtracted from food expenditure and added to housing expenditure.
7.8.2.3 Adjusting for when workers expected to exclusively use passenger transport Transportation consists of three expenditure groups in COICOP: (i) purchase of personal vehicles, (ii) operation of personal vehicles, and (iii) passenger transport. When it is acceptable/decent in a location for workers earning a living wage to exclusively use passenger transportation, we recommend reducing the percentage spent for transport by the additional costs associated with owning and operating a personal vehicle compared with exclusive use of public transport. We typically assume that motorbikes/motor scooters (that are common in developing countries) are twice as expensive to own and operate as passenger transport – and we think that this is a reasonable assumption for most developing countries.11
The size of an adjustment for transport will be small in locations where relatively few workers own a private vehicle. This turned out to be the case in rural South Africa where households at the 30th and 40th percentile of the income distribution spend only around 1.0% for private vehicles according to secondary data. The adjustment was fairly large for the rural Dominican Republic where households in the lower half of the income distribution spend around 6% for private vehicles because many people own motorbikes.
The preliminary estimate of NFNH costs for a living wage is equal to the cost of the living wage model diet multiplied by the adjusted NFNH to Food ratio for the appropriate location and the appropriate part of the income distribution.
The preliminary estimate of NFNH costs for a living wage from step 3 in our methodology is subject to post checks and possible adjustments to make sure that sufficient funds are available for health care and education, because they are considered human rights around the world. Post checks are also sometimes done for transport when this is a major expense. Post checks compare the amount implicitly included in the preliminary NFNH estimate for health care and education to rapid assessment estimates of typical costs for acceptable education and health care. NFNH is then increased when a rapid assessment indicates that there is a big difference. How to do rapid post checks is discussed in the next three chapters.
A hypothetical example is provided below to illustrate the steps involved in estimating NFNH costs. The hypothetical distribution of household expenditure data in this example is shown in Table 7.6. In addition, we assume that the poverty rate is 35%.
Since the poverty rate is 35%, expenditure data for households at the 40th percentile of the expenditure distribution from a recent household expenditure survey was used to estimate NFNH costs. Such households in this example spent 6.0% of household expenditure for owning and operating private vehicles, 8.0% for passenger transport, 1.7% for alcohol, 1.0% for tobacco, and 4.8% for food eaten away from home.
Table 7.6 NFNH to Food ratio based on secondary data from household expenditure survey before and after adjustments, hypothetical example
The adjusted NFNH to adjusted Food ratio was 1.14 (i.e. 42.7/37.3). Note that the unadjusted ratio was 1.23 (i.e. 46.4/37.6).
As shown and discussed in Section 7.5, the NFNH to Food ratio is higher when it is based on average household expenditure data than when it is based on expenditures of the median household or households at the 30th or 40th percentile of the household expenditure distribution. This presents a problem when the only published household expenditure data are for average household expenditures, because in this situation the NFNH to Food ratio estimated would be too high. The first thing a researcher should do in this situation is to make every effort to get a special tabulation of household expenditures by income decile or quintile. We were able to do this, for example, for rural Dominican Republic and Phnom Penh, Cambodia.
This appendix addresses the issue of what to do when after serious efforts have been made, a researcher only has data on mean household expenditure, and whether it is possible to develop guidelines for this situation. To address this, we put together data on the NFNH to Food ratio for the median household and mean household expenditure for: (i) 15 developing countries and territories from an ILO household income and expenditure survey database, and (ii) 5 developing countries where we had done living wage studies and so were able to adjust the household expenditure data to be consistent with how we estimate the NFNH to Food ratio in our methodology.
The NFNH to Food ratio was 25% lower on average (median) for the 15 developing countries and territories in the ILO database when this ratio was based on median household expenditure compared with when it was based on mean household expenditure. It was always lower for median expenditure than for mean expenditure, but there was considerable variability across countries as differences ranged from 9% to 40%.
Table 7A.1 examines the same issue of how the NFNH to Food ratio changes with household income for five developing countries where we were able to adjust the expenditure data to be consistent with our methodology. The NFNH to Food ratio was always higher when based on mean household expenditure than when based on expenditure of median household (and the NFNH share of household expenditure uniformly increased with household income in all five countries). This ratio was around 17% lower on average (median) when based on the median household expenditure compared to when the ratio was based on the mean household expenditure. As with countries in the ILO database, there was considerable variation in this difference across countries as it ranged from 4% to 67%. At the same time, the decrease in the NFNH to Food ratio was much smaller and much less variable as you go from the median to the 40th percentile and from the 40th to the 30th percentile of the household expenditure distribution; the ratio is around 7% lower on average between these deciles.
Table 7A.1 Percentage decrease in NFNH to Food ratio when based on mean household expenditure compared with when based on expenditure of households at 50th, 40th, and 30th percentile of household expenditure distribution using Anker methodology adjustments, five developing countries
Sources: Anker, R and Anker, M (2015, unpublished) for Vietnam, Cambodia and India. Anker and Anker (2014a) for Dominican Republic. Anker and Anker (2013) for South Africa.
Implications of the above analyses are: