There Are Seven Objectives That We Hope to Achieve in This Chapter. First of All We Need

Slide #1

Welcome to this online course on poverty analysis. It consists of eight modules, of which the first is the current one on measuring poverty. This module provides basic background, some of which may be familiar to you, but we need to review it before we move on to the other parts of the course, which are listed here on the left.

Slide #2

There are seven objectives that we hope to achieve in this chapter. First of all we need to define what we mean by poverty, and we also need to explain why we need to measure poverty in the first instance. This is relevant because poverty is expensive to measure. The next step is to identify the three steps required to measure poverty, and these consist of choosing a measure of welfare, establishing a poverty line, and then presenting and summarizing the results. The two main measures of poverty that require attention are consumption and income per capita, and we need to evaluate the pros and cons of each. After that we need to identify the methods used to construct a poverty line, and then both describe and explain how to summarize data on poverty using the headcount index and the FGT poverty measures. Finally, we will argue that there is no single best measure of poverty.

Slide #3

The World Bank defines poverty as “a pronounced deprivation in well-being”. Of course this begs the question of how to measure well-being, and also of how pronounced the deprivation needs to be in order to constitute poverty. The conventional approach is to suppose that poverty means that you don’t have enough money, and this is the view that we are going to take for the rest of this module. But there is a broader view, going back at least to the work of Amartya Sen, which argues that poverty consists of a lack of capabilities to function in society, and these capabilities constitute more than just money. We deal with this more carefully in Module 2, where we discuss multidimensional poverty. In passing,we should note that poverty is not the same thing as vulnerability; a household might not be poor now, but it might be vulnerable if it risks being poor in the future. Nor is poverty the same as inequality; poverty focuses on those at the bottom of the distribution, not on the distribution of income or consumption as a whole.

Slide #4

It is expensive to measure poverty, and it can only be done using household surveys. Given that it is expensive to measure, we need to be clear about why we are going to the trouble. Maybe the most compelling reason is that it keeps the poor on the agenda, so we won’t lose sight of the fact that there are poor people,and they do need our attention. The second reason for measuring poverty is so that we may target our interventions. Domestically we want to know who the poor are, and internationally it allows the World Bank and NGOs to identify where the world’s poor are, and channel their resources in those directions. We also want to measure poverty in order to monitor and evaluate policies that are intended to help the poor. And, in an ideal world, we would use all measures of poverty to evaluate institutions like the World Bank, which says, “our dream is a world free of poverty”. In this context, the World Bank encourages countries to produce Poverty Reduction Strategy Papers to help them focus on the concerns of the poor, and this in turn requires good measurement of poverty.

Slide #5

There are three basic steps that are needed in order to measure poverty. First we have to define an indicator of well-being, and the typical candidates are income or expenditure per capita. Having done that, we then need to establish the threshold, the poverty line below which one would be considered poor. And then we are able to generate our summary measures of poverty. In subsequent slides we will deal with each of these three steps in more detail, but before doing so we need to say a few more words about the data that are needed when measuring poverty.

Slide #6

Since we can only measure poverty using data collected from household surveys, it is important to understand the strengths and weaknesses of this information. Whenever we use data on poverty, there are a number of questions that need to be asked. The first is, is the sample frame representative? In other words, does the survey cover the entire population, and does it do so in a way that is random? Some survey data cover only part of a population, for instance the urban areas, or working people.

Second, we want to know how large the sample is; with larger samples our results are more reliable. Third, most surveys are stratified, they are not simple random samples. Typically we divide the country into regions and then maybe into units like provinces or counties; and then we may take villages and sample households randomly within those villages. We do stratified sampling because it is more efficient. A very important implication is that when processing the data we then need to use weights that reflect the sampling methods used.

The next question we need to ask is whether clustering was used. It is quite common to sample a group of households in a given village, and this constitutes a cluster. While clustered data do not give biased results, they do yield wider standard errors, for which we need to adjust when we work with the data. We also need to ask whether the data come from a cross section or a panel. And we must ask what indicators were collected – for instance, some surveys collect information on income but not expenditure, while others collect data on expenditure but not income.

We would also like to know how the data were collected; for example, were households asked to keep a diary recording their expenditures, or were they simply asked for the information in an interview? Finally, were the data cleaned well? Was the quality control good? Data quality varies considerably from survey to survey, although the living standards measurement surveys, developed by the World Bank, have set a very high standard.

Slide #7

The first important decision we need to make when measuring poverty is the choice of our measure of well-being. Ideally we would like to measure utility, which is a rather abstract measure of happiness or satisfaction, but we can’t do that directly. Instead we use a proxy for utility, typically income or consumption per capita.

Income is widely used, and it equals consumption spending plus any changes in our net worth. One problem is that income is quite difficult to measure in many poor countries, where the bulk of the population likely lives in rural areas, holds multiple jobs, and derives much of its income from farming, where recordkeeping is typically weak. It is also especially difficult to measure income from raising animals. We would like to measure poverty in a lifetime sense, but income can vary widely from year to year, and a household with low income this year may not be fundamentally poor. It is also widely accepted that income is significantly under-reported by households.

Slide #8

The other main candidate for measuring well-being is expenditure per capita. Here we need to find out how much a household spends on food and on other goods and services, and of course we need to include the value of goods and services that the household produces itself for its own consumption.

It has been widely observed that expenditure is more stable than income – households are often able to smooth their consumption from year to year – and so expenditure per capita tracks lifetime income relatively well. This idea is captured in the diagram here, which shows how income and consumption might vary over one’s lifetime; thus in any given year consumption is a better guide to one’s lifetime experience of income or poverty.

There are practical problems with measuring expenditure; a typical household will consume hundreds, perhaps thousands, of items, and gathering information on all of these can be extremely difficult, especially in rich countries where the range of goods is so much greater. We also have a standard problem that people under-report their spending on luxury items, or on so-called sin goods like alcohol and tobacco. Finally, all measures of expenditure are highly sensitive to the way in which the questions are asked; the more detailed the questions, the higher the reported level of spending. Between 1993 and 1998 the reported spending on tobacco doubled in Vietnam, but this was mainly because the questions used in the 1998 survey were more detailed.

Slide #9

This is a nice slide that Toots Albert put together some years ago. It summarizes the pros and cons of using income vs. consumption as a measure of welfare, and it is worth pausing for a moment to study this table in some detail. As a general rule, consumption per capita is used to measure well-being in poor countries because it is relatively easy to measure there, while income in such places is relatively difficult to measure; and income per capita is used to measure welfare in rich countries, where salaries can be measured easily enough but expenditure is highly complex.

Slide #10

There are some technical considerations that need to be addressed when measuring income or expenditure. Just three of these are worth mentioning for now. The first concerns durable goods, which are things like televisions and motorbikes. The motorbike that you bought a few years ago is providing you with services here and now, and those services should be counted as part of your income as well as part of your spending. This is especially clear and important in the case of housing. If you own your house, it is providing you with an important service, and is simultaneously part of your income and your spending. It is as if you are renting your house from yourself. Weddings and funerals also represent major expenses, but are often excluded from spending for the purposes of measuring poverty, which is not entirely satisfactory.

Slide #11

If we have chosen to use expenditure as our measure of well-being, we typically divide it by the number of members in the household to give us expenditure per capita. This is a good first approximation, but it is not ideal, because within a household different individuals have different needs. Also there are economies of scale, meaning that it is usually cheaper for a group of people to live together than to live in separate households.

One solution is to measure expenditure per adult equivalent. Instead of simply adding up the number of adults, we use a measure of equivalent adults, and there are different ways to do this. The OECD scale counts the first adult as having a weight of one, while subsequent adults have a weight of 0.7, and children have a weight of 0.5 each. And elegant formulation would take the number of adults plus alpha times the number of children and raise this to the power theta. Alpha is a number like 0.7 and reflects the weight of a child compared to an adult, and theta would be a number less than one and would reflect the effects of economies of scale. Of course the practical problem is what should all these weights be. In an important study,Deaton and Zaidi argue that that is no truly satisfactory method for estimating economies of scale in consumption. Thus adult equivalents rely on arbitrary assumptions, although of course it is just as arbitrary give everybody a weight of one.

Slide #12

Although we have emphasized that expenditure or income per capita are the standard measures of welfare that are used in most poverty studies, there are other measures of welfare that have been used from time to time. One could look at the number of food calories that are consumed; in this case a family is considered to be poor if it does not get enough calories every day.

Another approach would be to look at the share of consumption that is devoted to food. We know, by Engel’s Law that as households become richer, the share of spending devoted to food falls. So we could argue that, for example,poor people are those who devote at least 50% of their spending to food. A third approach is to measure nutritional outcomes – for instance, we could look at the body mass index of adults, or the height for age Z score for children – and somebody would be poor if they are malnourished by such a standard. This works well enough in poor countries, but is not so satisfactory in middle-income countries, where many poor people are overweight, not underweight. Finally we could ask people to assess whether they themselves are poor, or whether their neighbors are poor. The Social Weather Stations project in the Philippines has, for decades, been asking people to rate how poor they consider themselves to be. This self-rated measure of poverty tends to be relatively high, and stable over time, even if real incomes are rising.

Slide #13

Once we have our measure of welfare we need to establish a poverty line. Poverty lines may be relative or absolute. The relative poverty line measures the poor as those living in, for example, the poorest quintile, that is to say the poorest 20% of the population. By this standard the poor are always with us. The European Union often measures the poor as those whose income is less than 60% of the median income in the country where the individual lives.

An absolute poverty line sets a poverty line that simply doesn’t change. This is essential if one is comparing poverty over time. The World Bank’s dollar-a-day standard,about which we will say more in Module 6, is an example of this. Many of these absolute poverty lines look at the cost of basic needs; what does it cost to provide a minimum necessary diet and other essentials such as clothing and shelter? The so-called food poverty line asks whether you have enough spending to simply buy sufficient food to get by. The United States is the only rich country that uses an absolute poverty line. It was established in the early 1960s at the level of three times what was considered to be the cost of an adequate food diet; since then it has been updated only for inflation.

Slide #14

This is an interesting table, dating from about a decade ago, that looks at what types of poverty line were used by World Bank researchers who were examining poverty in Africa in the 1990s. The authors found is that in about half of the cases, researchers used relative measures of poverty; and in the other cases, only some use the classical cost-of-basic-needs approach. Indeed it is striking how much variety there is in the methodological approaches used.

Slide #15

One of the most attractive approaches to establishing the poverty line is the so-called cost-of-basic-needs method. The most basic need is that of food, so we generally start by setting a minimum necessary amount of food intake measured in kilocalories per person per day. Then we ask how much it would cost to buy enough food to supply this number of calories. Of course the cost of food varies a lot depending on what sort of food you eat so we typically look at the type of diet consumed by somebody at or near the poverty line. And then we need to add to this the cost of essential nonfood items such as clothing and shelter. Having established the poverty line, we update it over time by revising the prices of food and on nonfood items.

Slide #16

Both of these tables provide some information from Vietnam, and specifically from the living standards surveys that were undertaken in 1993 and 1998. The top table divides the population into quintiles, starting with the poorest fifth of the population, as measured by expenditure per capita. The first thing to note is that Engel’s Law applies, so as we go from poor to rich, the proportion of expenditure devoted to food falls, and the number of calories consumed rises. The last column is interesting: it shows the cost per calorie in dong – that’s the Vietnamese currency – and it is clear that the rich spend twice as much per calorie for their food as the poor. Thus it is very important to put the right price on the calories that are needed when we are defining the poverty line. The bottom table simply shows the poverty lines that were used to measure poverty in Vietnam in 1993 and 1998 using the cost-of-basic-needs method, adjusted for inflation. What is most striking here is that over these five years, as Vietnam was developing rapidly, the headcount poverty rate fell dramatically.

Slide #17

Once we’ve decided on our measure of welfare, and established our poverty line, we need to summarize our data. The most common way to do this is using the headcount index, which is simply the number of poor people divided by the number of people in our survey. Its great virtue is that it is a very easy measure to understand. One of its drawbacks is that it does not pick up the depth of poverty among the poor. This is illustrated nicely in the table. Both countries have headcount poverty rates of 50%, but clearly country A is much poorer because the two people in poverty there are deeply in poverty, unlike the two poor people in country B. It is also worth mentioning that the headcount poverty measure assumes that spending is shared equally among the members of a given household, and this is a strong assumption that is often not true.