More people in the developing world are eating out. Measuring this well could change our understanding of poverty and inequality
This page in:
Most of you probably buy lunch during the week, but can you recall what you ate yesterday? How about last week, did you snack in the afternoon? How much did you spend? Answering these questions is not as easy as you think, is it? Food consumed away from home (‘FAFH’ in short) represents an increasing share of food consumption around the world, caused by various factors including increasing urbanization, female labor force participation and evolving food systems that have made food availability easier. From street vendors, restaurants to corner stores, we have myriads of options to address our caloric intake needs during breakfast, dinner or for a snack when we leave home.
As it turns out, capturing this information well in the household surveys we use to measure welfare has not evolved as well. Very few surveys around the world do a good job of asking for detailed information on people’s eating habits outside their homes. As a recent review of surveys from 100 countries shows, the majority actually do a pretty mediocre job. Even when we use very generous criteria, 60% of the surveys do not pass the bar. In fact, a shocking 25% of surveys capture a household’s entire eating out expenditures by asking only one question.
But wait, what does this have to do with poverty and inequality?
Conceptually, data on FAFH can affect poverty estimates via two channels: the household’s food expenditures and the poverty line. For the former, a household that spends money eating out will have automatically have higher expenditures – making them appear wealthier. But to define someone as poor is also a function of whether the overall expenditures are below or above a poverty line. As such, if we want to consider FAFH, we also need to add the cost of “typical” FAFH in the value of the poverty line. This effect is ambiguous since it will depend on the relative cost per calorie of eating away from home with that of eating at home. Because the expenditure and poverty line effects may or may not go in the same direction, the overall effect of including FAFH will depend on the magnitude and direction of each. This also raises an important point: even if a country estimates poverty based only on income – a choice that each country makes on its own, FAFH can still influence poverty estimates through its effect on the poverty line.
New methodological research on improving household surveys that the Living Standards Measurement Study team is doing jointly with the World Bank Group’s Poverty Global Practice suggests that the missing data on FAFH in welfare estimates matters a lot. To understand this, we estimate poverty in Peru in a new paper by using the country’s official poverty methodology to compare a baseline scenario where FAFH data is not accounted for against a world where it is. Peru is a good case since more people are increasingly eating out, with the average Peruvian household having spent 27 percent of food budget outside the home during 2013. Peru is also among the few countries that collect high-quality data on FAFH since 2004 as part of a national expenditure survey, which serves as the basis for poverty measurement. This includes information on the number of meals per week, the type of establishment, the type of meal and the cost – all reported by the individual respondent.
The results are striking: we find that extreme and moderate poverty change dramatically but in opposite ways. Extreme poverty is significantly higher once FAFH is accounted for, driven by the higher calorie cost of food bought outside the home relative to home-made meals, which increases the poverty line (so more people “become” poor). By contrast, moderate poverty is significantly lower when FAFH is taken into account, this time driven by increases in household FAFH expenditures that raise overall expenditures (and thus compensate the effect of the poverty line increase).
What surprised us even more are the magnitudes of these results: in 2010, extreme poverty increased by 18 percent – from a baseline rate of 6.5; while moderate poverty fell by 16 percent – from a baseline rate of 36.6. Similar results hold for the period 2011-2013, and the findings are also consistent - in fact more pronounced - with poverty gap and severity measures. Even consumption inequality fell once FAFH is accounted for, though the impact is not statistically significant from 2012 onwards in our analysis.
There is one more unexpected insight: not only do poverty estimates change, so does who counts as being poor. Accounting for FAFH results in a re-ordering of households along the expenditure distribution. In our exercise, almost 2.5 million people (almost 10% of Peru’s population) moved from poor to non-poor and vice versa, which results in small but statistically significant differences for the typical poverty profile between the two scenarios.
So does that mean that our global poverty or inequality numbers are off? We just don’t know, but they probably are.
So what next? We see three immediate steps.
First, we need to replicate the Peru results in other countries to test if such large changes in welfare measures exist in other settings. Colleagues from the FAO find, for example, that when administrative data from school feeding programs is used to calibrate children’s food intake, Brazil’s food insecurity prevalence is reduced significantly. This is likely to matter also for poverty estimates.
Second, guidelines on good practices related to collecting FAFH in the field are urgently needed (we are working on that), such as making sure that information is reported by each individual (at least adult) in the household, ensure that different meals eaten outside the home are captured (lunch, dinner, snack) or information about meal content.
Third, we need to work closely with statistical agencies to design new methodological experiments to find what works best in collecting cost-effective information on FAFH (let us know if you have ideas!).
The good news is that this agenda is gaining traction: it is part of the work plan proposed by the Inter-Agency and Expert Group on Food Security, Agricultural and Rural Statistics (IAEG-AG), tasked by the United Nations Statistical Commission to guide methodological developments in statistics for food security, sustainable agriculture, and rural development.
So stay tuned. Until then, next time you eat out, remember that every bite counts!
As it turns out, capturing this information well in the household surveys we use to measure welfare has not evolved as well. Very few surveys around the world do a good job of asking for detailed information on people’s eating habits outside their homes. As a recent review of surveys from 100 countries shows, the majority actually do a pretty mediocre job. Even when we use very generous criteria, 60% of the surveys do not pass the bar. In fact, a shocking 25% of surveys capture a household’s entire eating out expenditures by asking only one question.
But wait, what does this have to do with poverty and inequality?
Conceptually, data on FAFH can affect poverty estimates via two channels: the household’s food expenditures and the poverty line. For the former, a household that spends money eating out will have automatically have higher expenditures – making them appear wealthier. But to define someone as poor is also a function of whether the overall expenditures are below or above a poverty line. As such, if we want to consider FAFH, we also need to add the cost of “typical” FAFH in the value of the poverty line. This effect is ambiguous since it will depend on the relative cost per calorie of eating away from home with that of eating at home. Because the expenditure and poverty line effects may or may not go in the same direction, the overall effect of including FAFH will depend on the magnitude and direction of each. This also raises an important point: even if a country estimates poverty based only on income – a choice that each country makes on its own, FAFH can still influence poverty estimates through its effect on the poverty line.
New methodological research on improving household surveys that the Living Standards Measurement Study team is doing jointly with the World Bank Group’s Poverty Global Practice suggests that the missing data on FAFH in welfare estimates matters a lot. To understand this, we estimate poverty in Peru in a new paper by using the country’s official poverty methodology to compare a baseline scenario where FAFH data is not accounted for against a world where it is. Peru is a good case since more people are increasingly eating out, with the average Peruvian household having spent 27 percent of food budget outside the home during 2013. Peru is also among the few countries that collect high-quality data on FAFH since 2004 as part of a national expenditure survey, which serves as the basis for poverty measurement. This includes information on the number of meals per week, the type of establishment, the type of meal and the cost – all reported by the individual respondent.
The results are striking: we find that extreme and moderate poverty change dramatically but in opposite ways. Extreme poverty is significantly higher once FAFH is accounted for, driven by the higher calorie cost of food bought outside the home relative to home-made meals, which increases the poverty line (so more people “become” poor). By contrast, moderate poverty is significantly lower when FAFH is taken into account, this time driven by increases in household FAFH expenditures that raise overall expenditures (and thus compensate the effect of the poverty line increase).
What surprised us even more are the magnitudes of these results: in 2010, extreme poverty increased by 18 percent – from a baseline rate of 6.5; while moderate poverty fell by 16 percent – from a baseline rate of 36.6. Similar results hold for the period 2011-2013, and the findings are also consistent - in fact more pronounced - with poverty gap and severity measures. Even consumption inequality fell once FAFH is accounted for, though the impact is not statistically significant from 2012 onwards in our analysis.
There is one more unexpected insight: not only do poverty estimates change, so does who counts as being poor. Accounting for FAFH results in a re-ordering of households along the expenditure distribution. In our exercise, almost 2.5 million people (almost 10% of Peru’s population) moved from poor to non-poor and vice versa, which results in small but statistically significant differences for the typical poverty profile between the two scenarios.
So does that mean that our global poverty or inequality numbers are off? We just don’t know, but they probably are.
So what next? We see three immediate steps.
First, we need to replicate the Peru results in other countries to test if such large changes in welfare measures exist in other settings. Colleagues from the FAO find, for example, that when administrative data from school feeding programs is used to calibrate children’s food intake, Brazil’s food insecurity prevalence is reduced significantly. This is likely to matter also for poverty estimates.
Second, guidelines on good practices related to collecting FAFH in the field are urgently needed (we are working on that), such as making sure that information is reported by each individual (at least adult) in the household, ensure that different meals eaten outside the home are captured (lunch, dinner, snack) or information about meal content.
Third, we need to work closely with statistical agencies to design new methodological experiments to find what works best in collecting cost-effective information on FAFH (let us know if you have ideas!).
The good news is that this agenda is gaining traction: it is part of the work plan proposed by the Inter-Agency and Expert Group on Food Security, Agricultural and Rural Statistics (IAEG-AG), tasked by the United Nations Statistical Commission to guide methodological developments in statistics for food security, sustainable agriculture, and rural development.
So stay tuned. Until then, next time you eat out, remember that every bite counts!
The FAFH survey brought my attention. I share your thoughts about how poor is the data available in many countries.
In order to adequately measure FAFH expenses vrs a family income, the are many challenges you already pointed out.
My opinion about the challenges is: to begin with... FAFH definition may significantly defer from economic status. To many families, that is not even an option. To others, it could be just for special celebrations.
If you are considering poverty issues, you should consider job location (is it at some factory that subsidized meals ?).
It will be interesting to attach the FAFH survey to the location where people go for medical assistance. When you go to the doctor, you are totally honest. For example at Costa Rica there is free medical access through the centers named "ebais". These ebais are located all through the country.
Good points Vanessa. Indeed, the issue of location is a complicated one since the options are limitless as you point out. We did not dwell on this on the post, but here is a simple conceptual diagram of the potential sources for FAFH and related measurement issues that one would need to consider. As you can see from the left panel, there are many different types of FAFH “events” with as many locations that they can happen. This indeed creates a complexity on how to design the survey in a way that can capture these in an efficient way. Who answers the question though will be equally important, since one big difference between FAFH and eating at home is that FAFH is an individual-specific event and so we do not have much faith asking an “informed person” about what everyone else ate away from home.
The second issue you raise is also important: we do not know much about “what” people eat, and this may also influence our assessments. So much more on this needs to happen, which we are working on…
The hope is that as we move forward, we test variations of modules that allow us to get as close to “truth” as possible in these dimensions.