Food Policy 72 (2017) 146–156 Contents lists available at ScienceDirect Food Policy journal homepage: www.elsevier.com/locate/foodpol You are what (and where) you eat: Capturing food away from home in MARK welfare measures☆ ⁎ Gabriela Farfán , María Eugenia Genoni, Renos Vakis The World Bank A R T I C L E I N F O A B S T R A C T Keywords: Consumption of food away from home is rapidly growing across the developing world, and will continue to do so Poverty as GDP per person grows and food systems evolve. Surprisingly, the majority of household surveys have not kept Inequality up with its pace and still collect limited information on it. The implications for poverty and inequality mea- Food consumption surement are far from clear, and the direction of the impact cannot be established a priori. This paper exploits Welfare measurement rich data on food away from home collected as part of the National Household Survey in Peru, to shed light on Data collection the extent to which welfare measures differ depending on whether food away from home is accounted for or not. JEL codes: Peru is a relevant context, with the average Peruvian household spending over a quarter of their food budget on O1 food away from home since 2010. The analysis indicates that failure to account for this consumption has im- C81 portant implications for poverty and inequality measures as well as the understanding of who the poor are. First, I31 I32 accounting for food away from home results in extreme poverty rates that are 18 percent higher and moderate poverty rates that are 16 percent lower. These results are also consistent, in fact more pronounced, with poverty gap and severity measures. Second, consumption inequality measured by the Gini coefficient decreases by 1.3 points when food away from home is included – a significant reduction. Finally, the inclusion of food away from home results in a reclassification of households across poor/non-poor status – 20 percent of the poor are dif- ferent, resulting in small but significant differences in the profile of the poor in dimensions such as demo- graphics, education, and labor market characteristics. Taken together, the results indicate that a serious re- thinking of how to deal with the consumption of food away from home in measuring well-being is urgently needed to properly estimate and understand poverty around the world. 1. Introduction nationally representative household surveys have not kept up with the pace and collect very limited information on food away from home Consumption patterns are rapidly changing across the developing (FAFH). Conceptual and practical challenges make integrating FAFH in world, with prepared and packaged meals and meals consumed outside household surveys a complex exercise. For example, we need a clear the home taking an ever growing share of the households’ food budget.1 protocol to capture otherwise confusing items such as meals produced Furthermore, with rising incomes, urbanization, and women entering outside but consumed at home – or vice versa; we need to measure the labor force, among various reasons, this trend is expected to persist meals whose content is unknown to the consumer and which are con- as economies transition to middle-income status (Smith, 2013; USDA, sumed in non-standard quantities. In addition, we are confronted with 2011). the likely high measurement error that arises if we elicit the informa- In spite of its growing participation in households’ budgets, most tion from a household informant – a common practice in household ☆ We are grateful to colleagues at the National Statistics Office (INEI) in Peru, and especially Nancy Hidalgo, Lucia Gaslac, and Oscar Perfecto for their countless help and advice during the data preparation and analysis. The paper benefited from comments from Gero Carletto, Dean Jolliffe, Talip Kilic, Nobuo Yoshida, members of the National Poverty Commission in Peru as well as participants of the 2014 Food Consumption Conference in Rome and the Side Event on Improving Food Consumption Measurement in National Household Surveys at the 46th Session of the UN Statistical Commission. Financial support for this research has been received from the World Bank’s KCP window as well as FAO’s funding through the “Global Strategy for Improving Agricultural and Rural Statistics” initiative. The views expressed in this paper are those of the authors and do not necessarily reflect those of the World Bank or any of its affiliated organizations. All errors and omissions are our own. ⁎ Corresponding author at: The World Bank, 1818H St, NW, Washington, DC 20433, USA. E-mail address: gfarfan@worldbank.org (G. Farfán). 1 FAFH has been found to contribute to as much as 36 percent of the daily energy intake among men in urban Kenya, and 59 percent among market women in urban Nigeria (Oguntona and Tella, 1999; Van’t Riet et al., 2002). Among the younger population, FAFH contributes, for example, to 18 and 40 percent of daily energy intake among Chinese children and school- going adolescents in Benin, respectively (Liu et al., 2006; Nago et al., 2010). http://dx.doi.org/10.1016/j.foodpol.2017.08.020 Available online 06 September 2017 0306-9192/ © 2017 The World Bank. Published by Elsevier Ltd. G. Farfán et al. Food Policy 72 (2017) 146–156 surveys, when the consumption takes place out of the home and distribution of food consumption by income strata changes once food therefore out of sight of the informant.2 consumed at school is taken into account. In particular, they show that As a result, in practice, few countries have addressed these survey proper account for food received through a school feeding program design issues adequately, as shown by a recent comprehensive assess- targeted to the poorer population results in a more equal distribution of ment done by Smith et al. (2014). To assess the relevance and reliability food consumption than previously thought, allowing for a long due of food data, the authors analyzed the questionnaire content of the most revision of the FAO assessment of undernourishment in Brazil. recent nationally representative consumption or expenditure household In this paper, we evaluate the impact of accounting for FAFH on survey from 100 developing countries, which represents 70% of the poverty and consumption inequality estimates in Peru.6 Drawing on developing countries. Among various quality indicators, the coverage rich FAFH data collected as part of the multi-year National Household and detail of FAFH data are analyzed. Following a very lax definition of Survey (ENAHO), we simulate a situation where we move from a world FAFH, which consists of checking whether “any food item in the food where FAFH is not accounted for to one where it is. In the process, we list itself, the title of the section in which it is found, or a question show that from a theoretical point of view the direction of the effect on regarding the item, contains words such as consumed out, restaurant, poverty or inequality cannot be predicted ex-ante. Peru is a relevant consumed away, and the like”, it turns out that 90 percent of the sur- context to study this question since FAFH is fairly widespread and in- veys do consider FAFH in some form. However, when looking more creasing. In 2013, the average Peruvian household spent 27 percent of deeply into the way this information is collected the authors find huge their food budget on FAFH. variation in quality, painting a far from optimal picture in the collection To assess the impact on poverty measurement, we follow the official of FAFH data. For example, a quarter of the surveys aim to capture all methodology adopted by the National Institute of Statistics and related household consumption from FAFH using just one question; one Informatics (INEI) and start with a scenario where FAFH is not ac- in five surveys considers multiple places of consumption; only 35 per- counted for. Then, we use this estimate as the benchmark against which cent takes snacks explicitly into account (when most snacking is ex- the impact of including FAFH is assessed. Peru updated its poverty pected to take place out of the home); and close to half of the surveys do measurement methodology in 2010, and therefore we use that year for not include FAFH received in kind. our analysis. The definition of FAFH included in the ENAHO comprises Poor measurement of FAFH may have far reaching consequences in all food prepared outside the home. We estimate the effect of FAFH on welfare analysis. Food consumption plays an instrumental role in the the poverty rate, the poverty gap, and the severity of poverty. Then, to design and monitoring of development policy at the local, national, and evaluate the effect on consumption inequality we compute the Gini global levels. Poverty, food security, health, and nutrition, lie at the coefficient based on the expenditure distribution with and without heart of the development agenda, and the computation and monitoring FAFH. Finally, we go beyond a summary welfare measure and analyze of indicators that track those welfare dimensions rely heavily on food whether lack of accounting for FAFH changes our understanding and consumption or expenditure data. While data on household consump- characterization of the poor population, by looking at how the profile of tion or expenditure have dramatically increased over the last few dec- the poor changes once we take into account FAFH. ades3, appropriate information on FAFH patterns is lacking, and the Our analysis indicates that failure to account for FAFH has sig- consequences of miss-measurement of food consumption on the as- nificant and sizable effects on poverty and inequality indices and to our sessment and understanding of these major policy areas are largely understanding of poverty in general. First, accounting for FAFH results unknown. Furthermore, as FAFH is expected to gain importance as in extreme poverty rates that are 18 percent higher and moderate economies develop, appropriately measuring this component con- poverty rates that are 16 percent lower than the scenario without FAFH. stitutes an urgent issue or overtime comparisons of consumption pat- The increase in the extreme poverty rate is driven by the higher per- terns and poverty will become less meaningful over time. calorie costs derived from FAFH relative to food prepared at home, To the best of our knowledge, only two papers analyze the im- which increase the cost of the food basket and therefore the poverty plications that failing to account for FAFH can have on food security line. In contrast, the moderate poverty rate falls because the increase in analysis.4,5 In a study from India, Smith (2013) argues that the great measured household consumption, which comes from accounting for Indian calorie debate, originated by an apparent increase in under- FAFH, offsets the rise in the moderate poverty line. These effects are nourishment at the time of falling poverty rates, can be partly explained also consistent, in fact more pronounced, when we compute changes in by inaccurate data on calorie intake due to the lack of measurement of the poverty gap and severity of poverty. Second, consumption in- FAFH. Similarly, Borlizzi and Cafiero (2014) in Brazil show how the equality not only falls among the poor (severity of poverty), but also across the entire population. When including FAFH, the Gini coefficient falls by 1.3 points. 2 In a small-scale study in an urban slum in India, Sujatha et al. (1997) interview Finally, accounting for FAFH also generates a re-ordering of spouses about the men’s dietary intake, and find that women are not aware of the foods households along the consumption distribution. Overall, 41 percent of consumed by their spouses outside their home. Similarly, Gewa et al. (2007) find that mothers of rural school-aged Kenyan children missed 77 and 41 percent of the energy the population changes their relative ranking when measured by the intake originated in FAFH in the food shortage and harvest seasons, respectively (where percentile of the expenditure distribution they belong under each sce- FAFH contributes to 13 percent and 19 percent of daily energy intake in each season). nario. This generates a reclassification of the population across poor/ 3 The 1990 World Bank World Development Report on Poverty relied on data from only non-poor status – about 20 percent of the poor population is different, 22 countries, and no country had more than one survey. Today, there are more than 850 resulting in small but significant differences in the profile of the poor surveys from 125 countries with consumption or expenditure data (Ravallion and Chen, 2011). when measured by demographic and socio-economic characteristics. 4 With obesity increasingly becoming a pressing health issue in some middle-income The remaining of the paper is organized as follows: Section 2 con- countries, the link between eating out and obesity is also drawing attention in the de- nects FAFH to welfare and discusses the impact that FAFH has on the veloping world (Bezerra and Sichieri, 2009; Lozada et al., 2008). poverty and inequality indicators analyzed in this paper as well as on 5 The literature on FAFH in the developed world has a longer history, where a main the profile of the poor; Section 3 introduces the setting and data, in- focus has been on health and nutrition issues. There is widespread interest in studying the differences in the nutritional composition of the food provided by commercial outlets cluding details on the official methodology INEI implements to compute relative to home-made food, aiming to understand the health consequences of eating out (Vandevijvere et al., 2009). In particular, there is a body of research devoted to under- stand the link between obesity and eating out, among other health outcomes (Burns et al., 6 A few papers analyze the impact of different aspects of survey design on total ex- 2002; Guthrie et al., 2002; Kant and Graubard, 2004; Le Francois et al., 1996; Lin and penditures, and poverty and inequality measures (Backiny-Yetna et al., 2014; Beegle Guthrie, 2012; Binkley et al., 2000). There is also interest in establishing food-based et al., 2012; Deaton and Grosh, 2000; Gibson et al., 2003; Jolliffe, 2001; Pradhan, 2001). dietary guidelines to prevent obesity and related chronic diseases (Kearney et al., 2001; The work by Backiny-Yetna et al. (2014) is the only one to look in particular at the impact O’Dwyer et al., 2005). of food consumption data collection methods on poverty and inequality. 147 G. Farfán et al. Food Policy 72 (2017) 146–156 poverty statistics; Section 4 presents the results; and Section 5 con- in the following way:9 cludes. 1 ΔP = [P (z 0,f1 (x ))−P (z 0,f0 (x )) + P (z1,f1 (x ))−P (z1,f0 (x ))] 2 2. Integrating food away from home in welfare analysis 1 + [P (z1,f0 (x ))−P (z 0,f0 (x )) + P (z1,f1 (x ))−P (z 0,f1 (x ))] 2 While it is well understood that welfare is multidimensional, con- In words, the expenditure effect represents the change in the pov- sumption has long been considered a flagship summary welfare in- erty index resulting from the change in the consumption distribution, dicator. In developing countries, where food consumption represents a holding fixed the poverty line. Similarly, the poverty line effect re- high share of total consumption, mismeasurement of food consumption presents the change in the poverty index resulting from a shift in the can have far reaching implications on welfare analyses. In this work, we poverty line, holding constant the consumption distribution.10 study the impact that accounting for FAFH has on monetary poverty and inequality analyses. 2.1.1. Expenditure effect Moving from a world without FAFH to one that measures FAFH has 2.1. FAFH and monetary poverty a relatively straightforward impact on overall household consumption: everything else equal, every household has higher total measured Under a monetary approach, the poor are those individuals whose consumption.11 In other words, the whole consumption distribution resources – measured through consumption (or income) – fall below an shifts to the right, and among the indices most commonly used, this ‘adequately’ defined threshold (the poverty line). Most countries track lowers poverty. two levels of poverty: extreme and moderate. Extreme poverty is as- sociated with a poverty line that reflects the cost of acquiring a food 2.1.2. Poverty line effect basket that satisfies minimum calorie requirements. 7,8 Moderate pov- Since the extreme poverty line is meant to reflect the cost of ac- erty also includes the cost of satisfying other essential needs. In general, quiring a food basket that satisfies minimum calorie requirements, the both lines are set based on the consumption patterns of a selected group impact of including FAFH will depend on the relative calorie costs of called ‘the reference population’. food items included and excluded under each scenario: the poverty line The choice of goods that conform the food basket and their prices is will go up (down) if calorie requirements are satisfied at higher (lower) not trivial, and today there is wide heterogeneity in the methodology costs when including FAFH.12 followed across countries. The fact that FAFH calories are, in general, The change in the moderate poverty line will depend on: (a) the more expensive, may raise further questions as to whether such con- change in the value of the food basket, and (b) the change in the re- sumption should be included. At the end, the choice is up to each lative cost between food and non-food items. Since accounting for FAFH country. One approach, followed by Peru among others, consists of increases the expenditure share allocated to food and therefore de- selecting the basket based on the consumption patterns of the popula- creases the relative cost of non-food items (i.e. for any given cost of the tion whose consumption is around the poverty line, under the rationale food basket, fewer resources need to be added to get to the moderate that those households consume in a way such that their basic needs are poverty line), the direction of the net effect will depend on the direction met. The objective is to use real costs as faced by households, without of (a) and the relative magnitude of (a) and (b).13 imposing constraints inconsistent with their behavior, under the as- Once again, for most common poverty indexes the direction of the sumption that households are rational in their consumption patterns. change in the poverty line determines the direction of the change in the With respect to FAFH, it can be argued that in today’s world eating out poverty index. is not a choice but a necessity. Urbanization, long working hours far away from place of residence, or women entering the labor force, are among some of the factors that explain changes in eating patterns. 2.1.3. Overall effect Furthermore, the increased demand for FAFH has been accompanied by The final effect on the poverty index will result from the sum of the an expansion of options including the proliferation of street foods, expenditure and poverty line effects. If both effects move in the same making calorie prices faced by those at the lower end of the distribution direction, there can be a significant change in the poverty index even much lower than prices at the top. when both the expenditure and poverty line effects are small in mag- There is far less agreement as to how to calculate the moderate nitude. Similarly, if the effects move in opposite directions, the net poverty line. A widely used methodology, followed by Peru, relies on effect can be quite small even if both the expenditure and poverty line the relative participation of food and non-food items in the household effects are substantial. budget to back-up non-food costs. Therefore, FAFH can affect poverty measurement through two 2.2. Consumption inequality channels: the estimation of household consumption and the estimation of the poverty lines. Let P (z ,f (x )) be a poverty index, with z the poverty Well-being not only depends on the levels of deprivation – i.e. line and f (x ) the consumption distribution, a change from a world poverty, but also on the distribution of resources across the population. without FAFH to one with FAFH can be expressed as: 9 ΔP = P (z1,f1 (x ))−P (z 0,f0 (x )) We follow a similar approach to Kakwani (2000). 10 If one measured the changes sequentially, the magnitude of each effect would de- where subscript 0 and 1 refer to poverty without and with FAFH, re- pend on the order in which they are computed. To avoid such arbitrary choice, the de- composition measures the expenditure effect as ½ the change due to the shift in the spectively. consumption distribution under the original poverty line and ½ the change due to the To quantify the impact generated by the change in consumption same shift in the consumption distribution computed under the new poverty line. A si- (expenditure effect) and the impact generated by the change in the milar reasoning applies to the poverty line effect. poverty line (poverty line effect), we decompose the previous expression 11 This assumes that the inclusion of FAFH in the survey does not lower reports of other consumption items. 12 The fact that the reference population changes across the two scenarios may result in 7 This corresponds to the ‘cost of basic needs approach’ which is the most commonly different calorie requirements. Among all simulations done, the change in calorie re- used. Alternative methods include the food energy intake approach and subjective eva- quirements is very low (the maximum change is 4 kilocalories from a total of 2105). luations. 13 An additional effect comes from the change in the composition of the reference 8 Internationally agreed recommendations on calorie requirement are provided by population. In practice, this effect is trivial relative to the impact of accounting for FAFH FAO/WHO/ONU, and differentiate by gender, age, and physical activity levels. on the food share when keeping fixed the reference population. 148 G. Farfán et al. Food Policy 72 (2017) 146–156 Fig. 1. Recent trends of FAFH in Peru.. Source: own calculations based on ENAHO 2006–2013 It is increasingly recognized that high levels of inequality is detrimental away by children, and (c) food consumed away by adults. to development and therefore development policy should focus, not The high quality of the FAFH data makes this survey well suited for only on growth, but also on inequality. The impact that proper mea- this analysis. The survey takes into consideration most of the elements surement of FAFH can have on consumption inequality depends on the reviewed by Smith et al. (2014). It considers both food produced out- degree to which the incidence and magnitude of FAFH consumption side and consumed outside as well as food produced outside but con- varies along the consumption distribution, and cannot be established sumed at home (take-out); it contemplates different modes of acquisi- ex-ante. tion – paid versus in kind; it explicitly considers meals and snacks; and it accounts for multiple sources of FAFH. Furthermore, this survey is 2.3. Re-classification and poverty profile among the few in the world that collect FAFH at the individual level.14 Unless all individuals consume the same amount of FAFH (or con- 3.2. Descriptive statistics on FAFH sume progressively more as we move along the distribution), the shift from a distribution without FAFH to one with FAFH results in a re- FAFH is fairly widespread in Peru, and despite its high incidence it ordering of individuals. If this re-ordering is big enough, the identity of has continued to grow over the last few years (Fig. 1, left panel). In those individuals falling at the lower end of the distribution will 2006, 84 percent of the households reported having at least one change. household member eating away from home. By 2013, almost 9 out of This has two important implications. On the one hand, it affects the 10 households have a member who eats at least a meal or snack out. If identification of the poor population leading to a re-classification across we focus on adult consumption, the trend presents a similar pattern, poor/non-poor status. On the other hand, it changes the composition, only shifted about 5 percentage points downwards.15 and therefore the characterization of the poor population. There are no Not only the incidence but the contribution of FAFH to the food strong a priori predictions as to how the profile would change, aside budget is substantial (Fig. 1, right panel). Between 2006 and 2013, the from the fact that those likely to leave poverty are individuals who eat share of FAFH on food expenditures increased by 21 percent, from 23 to more outside and those who fall into poverty are individuals more 27 percent. This is largely explained by adult consumption, which re- likely to eat at home. Therefore, correlates with eating out such as presents in 2013 almost a quarter of households’ food consumption. household composition, education, and labor market outcomes are In addition, FAFH is quite relevant for consumption levels. Fig. 2 likely to differ across those who change their poverty status. presents for the year 2010, (a) the absolute amount spent on FAFH and (b) the share of expenditures on FAFH over total expenditures, by 3. Setting and data percentile of the expenditure distribution without FAFH. We can see that, while the absolute amount spent on FAFH increases along the Peru is a middle-income country that has experienced sustained expenditure distribution, the share of FAFH on total expenditures is GDP growth over the last decade. Poverty has been steadily declining higher at the lower end of the distribution. Understanding the drivers of over time, with moderate poverty falling from 58.7 to 23.9 percent and FAFH behavior is beyond the scope of this work. However, it is con- extreme poverty from 16.4 to 4.7 percent between 2004 and 2013. sistent with worldwide trends linking FAFH with urbanization and Likewise, consumption inequality has declined to reach a Gini coeffi- labor market dynamics, and it is evident that FAFH is part of everyday cient of 35 in 2013. life among the Peruvian population. 3.1. Consumption data 3.3. Poverty methodology in Peru Data comes from the National Household Survey – ENAHO, a multi- The main source of information for poverty measurement is topic household survey that has been collected annually since 1995, ENAHO. Peru went through an important methodological change in and is the main source of information to monitor the living standards of poverty measurement in 2010, when a new food basket was selected the population. and new poverty lines computed. Therefore, our main analysis corre- The survey has an extensive and detailed household consumption sponds to the year 2010.16 module reported by a household informant. Within the food con- sumption module we can separately identify food prepared and con- 14 Only 17 percent of the surveys reviewed by Smith et al. (2014) collect individual- sumed at home, food produced outside but consumed at home (take- level information, and even fewer elicit the information from each individual respondent. out), and food consumed outside the home by children. Additionally, 15 In the measurement of adult consumption we only include meals produced by the survey has an individual-level module on food consumed outside commercial sources. While the questionnaire asks for consumption provided by social the home applied to each adult household member. Consumption is programs, the valuation of such consumption is controversial, thus we abstract from that component. Information on food consumed at other households is also reported, but in reported by meal event – breakfast, lunch, dinner, or snacks, and in- order to avoid double count of meals INEI excludes that component in the computation of dividuals report frequency, amount spent, and place of consumption, total household consumption. such as street food, restaurant, social program, or work. Following the 16 We also present estimates for the period 2010–2013. After 2010, the poverty lines country’s definition, FAFH includes: (a) take-out; (b) food consumed are updated only to account for price inflation. 149 G. Farfán et al. Food Policy 72 (2017) 146–156 Fig. 2. Distribution of FAFH expenditures, by percentile of the expenditure distribution without FAFH. Notes: own calculations based on ENAHO 2010. Amount spent on FAFH expressed in monthly Soles of Lima Metropolitana for 2010. Fig. 3. Impact of FAFH on poverty indexes, Peru 2010–2013. Notes: own calculations based on ENAHO 2010–2013. Dashed and dotted lines indicate the 95% confidence intervals. The selection of the food basket is based on the household and in- by the National Center for Food and Nutrition (CENAN) under the di- dividual consumption reported in the ENAHO by the reference popu- rection of the National Institute of Health. This table has all 941 items lation, which in Peru corresponds to those individuals between the 20th reported in the household-level consumption module. Additionally, it and 40th percentile of the per capita expenditure distribution. has information on 12 ‘representative food items’ that correspond to the Three pieces of information are necessary for the selection of the individual-level food consumption module. Since this module does not food basket: calorie content, quantity consumed, and value. Caloric have information on content, the 12 food items refer to a representative composition of each food item comes from a nutritional table computed meal for each meal occasion – breakfast, lunch, dinner, and snack, 150 G. Farfán et al. Food Policy 72 (2017) 146–156 Table 1 Impact of FAFH on poverty rates, poverty gap, and severity of poverty Simulations including different FAFH components. Place of consumption… Extreme poverty Moderate poverty At home At home+… All At Home At home+… All Meals out Takeout Meals out Takeout Adults Children Adults Children (1) (2) (3) (4) (5) (1) (2) (3) (4) (5) Poverty rate 6.46 8.15* 6.36 6.19 7.60* 36.55 33.72* 35.37 33.95* 30.72* Expenditure effect −2.94 −0.2 −0.14 −3.25 −9.03 −0.63 −1.12 −10.46 Line effect 4.63 0.1 −0.13 4.39 6.2 −0.54 −1.48 4.63 Overall effect 1.69 −0.1 −0.27 1.14 −2.83 −1.18 −2.60 −5.82 Poverty gap 1.41 1.88* 1.39 1.35 1.74* 11.23 9.94* 10.76 10.30* 8.94* Expenditure effect −0.81 −0.05 −0.03 −0.87 −3.62 −0.29 −0.35 −3.99 Line effect 1.28 0.03 −0.03 1.20 2.33 −0.18 −0.58 1.70 Overall effect 0.47 −0.02 −0.05 0.33 −1.29 −0.47 −0.93 −2.29 Severity of poverty 0.48 0.66* 0.47 0.46 0.60* 4.83 4.20* 4.60 4.42* 3.75* Expenditure effect −0.31 −0.02 −0.01 −0.33 −1.78 −0.15 −0.14 −1.92 Line effect 0.49 0.01 −0.01 0.45 1.15 −0.08 −0.27 0.84 Overall effect 0.18 −0.01 −0.02 0.12 −0.63 −0.22 −0.41 −1.08 Notes: own calculations following the official poverty methodology using ENAHO 2010. Standard errors take into account survey design. * Statistically significant at 5 percent level. differentiating across three sources – street vendors, restaurants, and Table 2 work.17 Summary statistics for poverty simulations including different FAFH components. To get an estimate of the quantity of FAFH by adults, INEI uses the Place of consumption… At home At home+… All information available from the other two components of FAFH: take-out and FAFH consumed by children where respondents do report the Meals out Takeout specific items and quantities consumed. With this information, INEI calculates the median price per kilogram (by quartile of per-capita Adults Children expenditure distribution, location, and meal type), and assigns this (1) (2) (3) (4) (5) value to adult FAFH to back-out quantities. Extreme poverty line 101.9 134.5 102.3 101.4 134.5 With the cost of the food basket, the moderate poverty line is de- Moderate poverty line 241.3 266.6 239.3 234.5 260.1 termined using the Orshansky coefficient or inverse of the food budget Orshansky coefficient 2.26 1.93 2.24 2.22 1.88 Median Kcal cost 0.004 0.007 0.003 0.004 0.004 share – that measures how many times total consumption exceeds food Monthly pc-expenditure 444 501 446 452 512 consumption. FAFH component: Incidence 89.77 37.14 42.93 93.70 Share over total exp. 16.05 0.95 2.23 19.23 4. Results Share (con > 0) 17.88 17.88 5.21 20.50 4.1. Changes in welfare indicators Notes: own calculations following the official poverty methodology using ENAHO 2010. Statistics calculated at the individual level among the reference population. National 4.1.1. Poverty poverty line calculated as a weighted average of the 7 regional poverty lines. We now evaluate the impact that the inclusion of FAFH has on poverty estimates. We focus on the FGT(α) family index, in particular sample design and population weights. the FGT(0) or head-count ratio; the FGT(1) or poverty gap; and the Table 1 presents the full set of results. The left panel presents ex- FGT(2) or severity of poverty. treme poverty estimates and the right panel moderate poverty esti- Our baseline scenario consists of estimated consumption and pov- mates. For each case, the estimates of the poverty headcount, gap and erty lines in a world where no information on FAFH exists (“at home” severity are presented in the upper, middle, and lower panels, respec- estimates). The resulting poverty estimates constitute the benchmark tively. Finally, column (1) of each panel shows the baseline specifica- against which the impact of FAFH is assessed. tion; columns (2) to (4) include a component of FAFH one at a time Next, we proceed to include, one at a time, the three components of while leaving the other FAFH components out; and column (5) sum- FAFH: food consumed away by adults (15 and older), food consumed marizes the overall effect of FAFH that match the official poverty sta- away by children (14 and younger), and food consumed at-home by the tistics. household (take-out). As pointed out in the previous section, FAFH by adult members is by far the largest component of FAFH. The inclusion of all three components simultaneously results in the official poverty 4.1.1.1. Expenditure effect. The expenditure effect provides the estimates. estimated change in poverty statistics resulting from including the For each poverty index, we decompose the overall effect into ex- different components of FAFH in total expenditures, while leaving the penditure and poverty line effects following the specification presented poverty line unchanged.18 in Section 2.1. Standard errors are estimated taking into account the Focusing on the poverty headcount, the results suggest that if in- formation on FAFH was excluded, extreme poverty would have been 17 This information was gathered through conversations with INEI staff. We could not access documentation specifying the process by which these representative items were 18 This effect coincides with the total effect in cases where the poverty line used comes selected. from a different data source and is not adjusted to take into account FAFH. 151 G. Farfán et al. Food Policy 72 (2017) 146–156 6.46 percent and moderate poverty 36.55 percent in 2010. Once all These pronounced differences in the impact on the cost of the food components of FAFH are included, the expenditure effect implies a 3.3 basket across FAFH components translate into sharp differences in the percentage point reduction in extreme poverty and a 10.5 percentage impact on the moderate poverty line. When accounting for adult con- point fall for moderate poverty. Both changes are sizable in an order of sumption, the increase in the cost of the food basket outweighs the fall 30–50 percent relative to the baseline specification. in the expansion factor – Orshansky coefficient, resulting in an increase If we break FAFH into its different components, adult consumption in the moderate poverty line. In contrast, when accounting for child is the main driver of the expenditure effects. Accounting for adult FAFH consumption or take-out the cost of the food basket barely changes, and translates into a decrease of 2.9 percentage points in extreme poverty therefore that effect is outweighed by the fall in the expansion factor, and 9 percentage points in moderate poverty. In contrast, accounting resulting in a lower moderate poverty line. The fall in the poverty line for FAFH consumed by children or take-out generate very small changes is, nevertheless, modest because these FAFH components contribute in the extreme poverty rate, 0.2 and 0.1 percentage points respectively, very little in the overall household budget and therefore have a small and accumulate to a change of about 1 percentage point in moderate impact on the expansion factor. Consequently, the poverty line effect is poverty. As expected, these differences are explained by the relative positive and substantial when introducing adult consumption, but ne- small importance that each component has on the household budget.19 gative and small when including either of the other two components. Indeed, summary statistics reported in Table 2 show that adult con- Once again, the direction and relative magnitude of the effects also sumption is by far the most important component, with an incidence of apply to the poverty gap and severity of poverty. almost 90 percent, and a mean participation in the household food budget of about 16 percent.20 By contrast, significantly fewer resources are assigned to child consumption or take-out (incidence is about 40 4.1.1.3. Overall effect. The overall effect results from adding the percent and participation in the household budget is at most 2.2 per- expenditure and poverty line effects. If they both go in the same cent). direction, they reinforce each other and magnify the final impact on the The same pattern follows when we look at the poverty gap and the poverty indexes. If they move in opposite directions, the net effect will severity of poverty. Accounting for FAFH reduces the distance between depend on the difference between the two, and the magnitude of the the mean expenditures among the poor and the extreme poverty line by overall effect will be smaller in absolute terms than either of the two S/0.87, and the moderate poverty line by almost S/3.99. Furthermore, effects individually. the reduction is relatively stronger for those further away from the In our case, the overall effect of including all components of FAFH poverty line, causing the severity of poverty to fall too. Again, these on extreme poverty is always positive and significant. The poverty rate effects are dominated by adult consumption. increases by 1.1 percentage points, or about 18 percent rise over the benchmark results. The increase on the poverty gap and severity of poverty is even more pronounced, changing by about 25 percent. 4.1.1.2. Poverty line effect. In contrast to the expenditure effect, the Extreme poverty increases because the impact of a higher poverty line direction of the poverty line effect is a priori unknown. Furthermore, outweighs the impact of higher expenditures, and it is driven by adult the extreme and moderate poverty lines need not move in the same FAFH. direction. Table 2 presents the simulated poverty lines under each By contrast, the overall effect of all FAFH components on moderate scenario. We find that when all components of FAFH are accounted for, poverty is always negative and significant. The moderate poverty rate the extreme poverty line increases by S/0.33 while the moderate falls by almost 6 percentage points, or a 16 percent decrease over the poverty line only increases by S/0.19 (a 32 and 8 percent increase benchmark. Similarly, the poverty gap and severity of poverty fall, by over baseline, respectively). The increase in the extreme poverty line is 20 and 22 percent, respectively. Contrary to the case of extreme pov- explained by the higher calorie costs of FAFH relative to food prepared erty, it is now the expenditure effect that outweighs the poverty line at home (see Table 2). When computing the moderate poverty line, the effect. increase in the cost of the food basket is partially offset by the fall in the A second difference relative to extreme poverty case is that now Orshansky coefficient, which implies that non-food items are weighted adult consumption does not explain virtually the totality of the change. more in the moderate poverty line estimate. When analyzing the expenditure and poverty line effects individually, The increase in the two poverty lines translates into about a 4.4 and the absolute magnitude of the effects is always stronger when including 4.6 percentage point increase in the extreme and moderate poverty adult consumption relative to either of the other two components. rates, respectively (Table 1). The reason the effects are similar is be- However, taken together that is no longer the case. The reason behind is cause of two compensating effects. In the first case, the poverty line that the expenditure and poverty line effects move in opposite direc- shifts more but the effect is evaluated in a segment of the expenditure tions when we account for adult consumption. However, both effects distribution with lower population density. In the second case, the move in the same direction when accounting for child consumption or poverty line moves less, but the move is evaluated in a segment of the take-out. Individually, each of the three components has a statistically distribution with higher population density. significant impact on poverty and the three effects are of comparable The effects of FAFH by its different components vary across our magnitudes. scenarios. The differences in extreme poverty rates are driven by dif- In sum, when looking at extreme (moderate) poverty, the inclusion ference in calorie costs: calories from adult consumption are more ex- of FAFH increases (decreases) the number of individuals who are poor, pensive than calories from home-made food, but that is not the case for increases (decreases) the poverty gap among the poor, and increases child consumption or take-out (Table 2). As a result, the poverty line (decreases) the severity of poverty among the poor. Furthermore, the and poverty rate change significantly when we incorporate adult con- impact of FAFH on the poverty gap and severity of poverty is more sumption, but they barely change when we include child consumption pronounced than the effect on poverty rates, especially for extreme or take-out. poverty.21 19 This may reflect a weakness in the questionnaire design to accurately collect this information. See Borlizzi and Cafiero (2014). 20 Since poverty rates are calculated at the individual level, figures in Table 2 are at the 21 individual level. An incidence of 90 percent means that 9 in 10 individuals live in Fig. 3 shows that these results are robust to the year selected. When we replicate households where at least one adult member eat outside during the reference period. poverty indexes for the period 2010–2013, we find a more or less parallel shift of the Similarly, on average individuals live in households where 16 percent of the food budget downward poverty trend that Peru experienced over the last few years. These results are is allocated to expenditure on FAFH by adults. consistent with the fairly stable trend in FAFH consumption since 2010 (Fig. 1). 152 G. Farfán et al. Food Policy 72 (2017) 146–156 Table 3 Table 5 Impact of FAFH on the Gini coefficient Simulations including different FAFH components. Impact of FAFH on the profile of the extreme poor. Difference in means between those who are poor when FAFH is included (FAFH poor) relative to those who are poor when Place of consumption… At home At home+… All FAFH is excluded (at-home poor). Meals out Takeout Sample… Extreme poverty Adults Children If changed poverty status All (1) (2) (3) (4) (5) Gini coefficient 38.46 37.28 38.28 38.45 37.14 Poor Non- Diff At- FAFH poor Diff (0.48) (0.44) (0.47) (0.47) (0.44) to poor home Effect −1.18 −0.18 0.00 −1.32 non- to poor (0.65) (0.67) (0.67) (0.65) poor poor Notes: own calculations based on ENAHO 2010. Standard errors calculated taking into Household size & composition account survey design and sample weights. *Statistically significant at 5 percent level. Household size 5.14 4.80 −0.34 5.01 4.92 −0.09* Children < 15 0.75 0.77 0.02 0.74 0.75 0.01 Women 15–60 0.81 0.80 −0.01 0.79 0.79 0.00 Women 60+ 0.23 0.31 0.08** 0.30 0.32 0.02 Table 4 Men 15–60 0.85 0.69 −0.16*** 0.75 0.71 −0.04*** Change in poverty status when including FAFH. Men 60+ 0.23 0.28 0.05 0.25 0.26 0.01 Extreme Poverty Household's head characteristics Female 0.23 0.21 −0.02 0.22 0.22 −0.01 Including FAFH Single 0.04 0.04 −0.01 0.04 0.04 0.00 Indigenous 0.42 0.54 0.12*** 0.49 0.52 0.03** Poor Non-poor Illiterate 0.36 0.27 −0.09** 0.34 0.31 −0.03** Primary 0.62 0.56 −0.06 0.63 0.61 −0.02* Only at-home meals Poor 1,520,463 393,285 incomplete Non-poor 730,765 26,979,752 Primary 0.19 0.23 0.04 0.20 0.21 0.01 Moderate Poverty complete Secondary 0.09 0.11 0.01 0.09 0.10 0.00 Including FAFH incomplete Secondary and 0.09 0.10 0.01 0.07 0.08 0.01 Poor Non-poor above Dwelling ownership Only at-home meals Poor 8,748,740 2,078,532 Renting 0.03 0.01 −0.01 0.02 0.01 0.00 Non-poor 353,034 18,443,959 Owner 0.78 0.81 0.03 0.83 0.83 0.00 Other 0.20 0.18 −0.02 0.15 0.16 0.00 Notes: own calculations based on ENAHO 2010. Access to facilities Running water 0.38 0.37 −0.01 0.34 0.34 0.00 Bathroom 0.37 0.46 0.10** 0.39 0.42 0.03** 4.1.2. Consumption inequality Sewage 0.11 0.12 0.01 0.10 0.10 0.01 connection To measure the impact on consumption inequality in Table 3 we Electricity 0.56 0.58 0.03 0.46 0.49 0.02** look at the Gini coefficient. We find that when FAFH is accounted for, Phone 0.40 0.36 −0.04 0.27 0.27 0.01 the Gini coefficient falls from 38.5 to 37.1, a statistically significant Geographic region change.22 Once again, this reduction is mainly driven by adult FAFH Urban coast 0.07 0.05 −0.02 0.04 0.04 0.00 consumption.23 Rural coast 0.05 0.02 −0.03** 0.04 0.03 −0.01** Urban sierra 0.05 0.07 0.02 0.05 0.06 0.00 Rural sierra 0.61 0.64 0.04 0.67 0.67 0.00 4.2. Re-classification and poverty profile Urban selva 0.04 0.05 0.01 0.03 0.04 0.00 Rural selva 0.15 0.11 −0.04 0.15 0.13 −0.01** When we account for FAFH, 41 percent of the population changes Metropolitan 0.04 0.07 0.03 0.02 0.03 0.01 their relative ranking – measured by consumption decile. Moreover, the lima re-classification of individuals across poverty status is also substantial Labor market outcomes (Table 4). Among those classified as extreme poor when no FAFH is Log per-capita 5.39 4.75 −0.64*** 4.74 4.63 −0.11*** income included, 21 percent (or close to 400,000 people) ‘escape’ poverty once # Individuals 2.88 2.18 −0.71*** 2.39 2.23 −0.15*** we include FAFH. In addition, more than 730,000 individuals that were employed not poor are classified as poor with the inclusion of FAFH. The corre- # Females 1.20 1.06 −0.14 1.08 1.05 −0.03 sponding numbers for moderate poverty are also large: 19 percent employed (more than 2,000,000 individuals) and 350,000, respectively. # Males 1.69 1.11 −0.57*** 1.30 1.18 −0.13*** employed Does this re-classification affect the overall characterization of the # Employers 0.06 0.06 0.00 0.04 0.05 0.00 poor population in any significant manner? To see this, we compare # Females 0.01 0.01 0.00 0.01 0.01 0.00 poverty profiles when FAFH is accounted or not. First, we compare employers demographic and socioeconomic characteristics between those groups # Males 0.04 0.04 0.00 0.04 0.04 0.00 employers that change poverty status when FAFH is included: those who leave # Employees 1.55 0.44 −1.12*** 0.75 0.51 −0.24*** versus those who fall into poverty. Then, we test whether the change in # Females 0.45 0.14 −0.31*** 0.21 0.15 −0.06*** the composition of the poor population leads to a different poverty employees # Males 1.11 0.30 −0.81*** 0.54 0.36 −0.18*** employees 22 Bootstrapped standard errors are estimated taking into account the survey sample # Self- 1.15 1.10 −0.05 1.14 1.13 −0.02 design and population weights (Bhattacharya, 2005, 2007). employed 23 Fig. 4 presents inequality trends for the period 2010–2013. Proper account for FAFH # Females 0.37 0.38 0.00 0.37 0.37 0.00 shifts the trend downwards. Changes are significant at 5 percent in 2010–11, but not in (continued on next page) 2012–13. 153 G. Farfán et al. Food Policy 72 (2017) 146–156 Table 5 (continued) Table 6 Impact of FAFH on the profile of the moderate poor. Difference in means between those Sample… Extreme poverty who are poor when FAFH is included (FAFH poor) relative to those who are poor when FAFH is excluded (at-home poor). If changed poverty status All Sample… Moderate poverty Poor Non- Diff At- FAFH poor Diff to poor home If changed poverty status all non- to poor poor poor Poor non- Diff At- FAFH poor Diff to poor home self-employed non- to poor # Males self- 0.78 0.72 −0.06 0.78 0.76 −0.02 poor poor employed Household size & composition Observations 296 587 883 1505 1796 3301 Household size 4.38 3.75 −0.63*** 4.78 4.83 0.05** Children < 15 0.68 0.67 0.00 0.76 0.77 0.02*** Source: ENAHO 2010. Women 15–60 0.80 0.73 −0.07* 0.83 0.84 0.00 Notes: all statistics at the household level, population-weighted means; test of difference Women 60+ 0.18 0.33 0.15*** 0.25 0.27 0.02*** in means accounts for sampling design. Men 15–60 0.85 0.73 −0.12*** 0.80 0.78 −0.02*** , , Statistically significant at 1, 5, and 10 percent levels, respectively. *** ** * Men 60+ 0.20 0.24 0.04 0.23 0.23 0.01** Household's head characteristics Female 0.24 0.24 0.00 0.22 0.22 0.00 profile by comparing average characteristics of the entire poor popu- Single 0.06 0.05 −0.01 0.04 0.04 0.00** lation for the two scenarios. All differences are calculated going from Indigenous 0.32 0.36 0.04 0.39 0.41 0.02*** the ‘FAFH poor’ – i.e. poor when FAFH is included, to the ‘at-home Illiterate 0.13 0.14 0.01 0.18 0.20 0.01*** Primary 0.32 0.37 0.05 0.43 0.45 0.02*** poor’ – i.e. poor when it is excluded, and therefore reflect the fact that incomplete the former group eats relatively more at home than the latter. Results Primary 0.21 0.19 −0.02 0.21 0.21 0.00 are shown in Table 5 for extreme poverty and Table 6 for moderate complete poverty. Secondary 0.17 0.13 −0.04 0.15 0.14 −0.01** incomplete Overall, we find that the two population groups that change poverty Secondary and 0.30 0.31 0.01 0.21 0.20 −0.02*** status have different average characteristics. These differences also re- above main statistically significant, though smaller in size, when we compare Dwelling ownership the overall poor populations under each scenario. While the magnitude Renting 0.09 0.08 −0.01 0.05 0.04 −0.01*** of the changes is generally small, they are consistent with expected Owner 0.64 0.70 0.06 0.73 0.74 0.02*** correlates with eating out, especially terms of household composition Other 0.27 0.21 −0.05 0.23 0.22 −0.01*** and labor market characteristics. Access to facilities Looking first at demographic characteristics among the extreme Running water 0.65 0.65 0.01 0.54 0.52 −0.02*** poor, the ‘FAFH poor’ have a different household structure – smaller Bathroom 0.66 0.69 0.04 0.58 0.56 −0.01*** households with fewer prime-age males; the household head is more Sewage 0.52 0.46 −0.06 0.35 0.31 −0.04*** connection likely to be indigenous and is less likely to be illiterate or have primary Electricity 0.86 0.80 −0.06** 0.73 0.70 −0.03*** incomplete. There are no statistically significant differences in the Phone 0.71 0.66 −0.05 0.56 0.52 −0.03*** distribution of dwelling ownership, though there are a few in access to Geographic region services: the ‘FAFH poor’ are more likely to have electricity and a Urban coast 0.16 0.29 0.14*** 0.14 0.14 0.00 bathroom. In terms of geographic location, once FAFH is included there Rural coast 0.05 0.00 −0.05*** 0.04 0.04 0.00*** are fewer poor households in rural areas. Finally, marked differences Urban sierra 0.17 0.06 −0.11*** 0.13 0.11 −0.01*** Rural sierra 0.17 0.39 0.22*** 0.36 0.41 0.05*** arise when looking at labor market outcomes. Results are consistent Urban selva 0.10 0.00 −0.10*** 0.07 0.06 −0.01*** with those households that are extremely poor when FAFH is accounted Rural selva 0.08 0.03 −0.05*** 0.09 0.09 0.00 for – but not otherwise – having lower income per capita and fewer Metropolitan 0.27 0.23 −0.04 0.17 0.15 −0.02*** members employed, in particular males. Differentiating across types of lima employment, it is fewer employees, as opposed to self-employed or Labor market outcomes employers, what drives the results. Overall, the findings suggest that Log per-capita 5.89 5.45 −0.44*** 5.29 5.14 −0.14*** households with more prime-age adults and working members are more income # Individuals 2.50 1.84 −0.66*** 2.34 2.27 −0.07*** likely to consume FAFH and therefore more likely to be classified as employed extreme poor if those resources are not properly accounted for. # Females 1.07 0.85 −0.23*** 1.06 1.04 −0.01* The difference in the profile of the moderate poor is more pro- employed nounced, with almost all characteristics being different across the two # Males 1.43 0.99 −0.43*** 1.28 1.23 −0.05*** scenarios. In terms of household structure, the ‘FAFH poor’ have larger employed # Employers 0.11 0.10 −0.01 0.08 0.07 −0.01** households with more dependents, and fewer prime-age males. In ad- # Females 0.04 0.03 −0.01 0.02 0.02 0.00** dition, household heads are more likely to be indigenous and have employers lower education. In contrast with extreme poverty, ownership status is # Males 0.07 0.07 0.00 0.06 0.06 0.00 also different: the ‘FAFH poor’ are more likely to be owners. They are employers # Employees 1.28 0.48 −0.80*** 0.89 0.77 −0.12*** also less likely to have access to services such as water, sewage, or # Females 0.45 0.14 −0.31*** 0.29 0.24 −0.05*** electricity. Moreover, there are statistically significant differences in the employees geographic distribution of the poor. Finally, consistent with increasing # Males 0.83 0.34 −0.49*** 0.60 0.53 −0.07*** consumption of FAFH as resources increase, ‘FAFH poor’ households employees have fewer members employed and lower income per capita. # Self- 1.04 1.04 0.00 1.09 1.09 0.01 employed # Females 0.46 0.43 −0.03 0.43 0.43 −0.01 (continued on next page) 154 G. Farfán et al. Food Policy 72 (2017) 146–156 Table 6 (continued) Finally, we explore whether these effects change our general un- derstanding of who the poor are. We confirm that by accounting for Sample… Moderate poverty FAFH poverty profiles differ due to the fact that individuals move along If changed poverty status all the consumption distribution. While the magnitude of the changes is generally small in this setting, they are consistent with expected cor- Poor non- Diff At- FAFH poor Diff relates of eating out, and therefore raise concerns over the possible to poor home implications to other settings. non- to poor poor poor The collection of FAFH data in national household surveys poses a number of challenges. For example, respondents do not know the in- self-employed gredients used in the meals and therefore can only vaguely specify the # Males self- 0.58 0.61 0.03 0.65 0.67 0.02*** content of what they consume, meal sizes vary widely and cannot be employed measured in standard units, complicating the accurate report of quan- Observations 1445 284 1729 7533 6372 13,905 tities. Moreover, household informants are unlikely to know precisely the consumption of all other household members. Additionally, even Source: ENAHO 2010. Notes: all statistics at the household level, population-weighted means; test of difference with adequate data on household or individual consumption, the in means accounts for the sampling design. measurement of FAFH usually also requires some information from the , , Statistically significant at 1, 5, and 10 percent levels, respectively. *** ** * supply-side. While this is most relevant for nutrition and health ana- lyses, it is also important for poverty analysis since calories are used in the construction of the poverty line. Due to these challenges, new methodological research is needed to identify best practices in the collection of FAFH data. Yet, some po- tential directions to improve data collection for poverty and inequality analysis seem relevant. First, adding a module on FAFH seems to be a first order priority. To simplify reporting, information could be asked by meal events (with the events tailored to the context); snacks (and possibly alcoholic beverages) should be explicitly taken into account; the module could differentiate across places of consumption, especially across those expected to have different calorie prices and nutritional composition. Addressing the respondent challenge is harder. Whenever possible, such as when surveys already have an individual-level module, information should be reported by each adult respondent. If this proves impossible and information needs to be reported by a household informant, careful thought should be put into the selection of the appropriate person. Finally, household level reports could be complemented with additional information to estimate calorie costs. In all, more work is needed to further test some of these ideas. In all, our findings suggest that ignoring this increasingly important Fig. 4. Impact of FAFH on Consumption Inequality, Peru 2010–2013 Gini coefficient. component of food consumption in household surveys can seriously Notes: own calculations based on ENAHO 2010–2013. Dashed and dotted lines indicate affect welfare measures and our general understanding of who the poor the 95% confidence intervals. are. Given that the direction of the effect is unknown ex ante, ac- counting for FAFH is even more important. Furthermore, as the pre- 5. Conclusion valence and importance of FAFH is expected to increase as economies grow, the measurement of FAFH in national household surveys becomes Food consumption away from home is rapidly growing across the an urgent issue or overtime comparisons will become less and less developing world, yet Household Consumption and Expenditure meaningful over time. Future research should aim to expand and vali- Surveys are failing to properly account for these changes in consump- date these findings in other settings, and deepen the analysis through tion patterns. This is likely to have significant consequences for welfare methodological research aiming to identify how best to collect this analysis, since the measurement of food consumption is critical for information in household surveys. poverty, food security, health, and nutrition analyses, among others. In this paper we study the potential impact that excluding FAFH References consumption may have on poverty and consumption inequality. We first show that conceptually, the direction of the change on poverty and Backiny-Yetna, P., Steele, D., Yacoubou Djima, I., 2014. The impact of household food inequality cannot be established ex ante. Then, we take advantage of consumption data collection methods on poverty and inequality measures in niger. WP Policy Research Working Paper 7090. the rich data on FAFH collected by the National Household Survey in Beegle, K., De Weerdt, J., Friedman, J., Gibson, J., 2012. Methods of household con- Peru to provide evidence on the direction and magnitude of the change sumption measurement through surveys: Experimental results from Tanzania. J. Dev. for Peru by simulating the scenario where no FAFH is accounted for. Econ. 98 (1), 3–18. Bezerra, I.N., Sichieri, R., 2009. Eating out of home and obesity: a Brazilian nationwide Peru is a relevant context for this analysis, as FAFH has been growing study. Public Health Nutrition 12 (11), 2037–2043. over the last decade to reach about 27 percent of the food budget. Bhattacharya, D., 2005. Asymptotic inference from multi-stage samples. J. Econom. 126, We find that the effect on poverty and inequality indices is sub- 145–171. Bhattacharya, D., 2007. Inference on inequality from household survey data. J. Econom. stantial in magnitude and statistically significant. The extreme poverty 137, 674–707. rate increases by 18 percent while the moderate poverty rate falls by 16 Binkley, J.K., Eales, J., Jekanowski, M., 2000. The relation between dietary change and percent once FAFH is included. The impact on the poverty gap and rising US obesity. Int. J. Obes. Relat. Metab. Disord. 24 (8), 1032–1039. Borlizzi, A., Cafiero, C., 2014. National food security assessment through the analysis of severity of poverty has a similar direction and is in general larger. In household surveys food consumption data: the case of Brazil’s Pesquisa de Orçamento terms of inequality, accounting for FAFH reduces the Gini coefficient by Familiares 2008/09, Mimeo. 3.4 percent, a statistically significant reduction. Burns, C., Jackson, M., Gibbons, C., Stoney, R.M., 2002. Foods prepared outside the home: association with selected nutrients and body mass index in adult Australians. 155 G. Farfán et al. Food Policy 72 (2017) 146–156 Public Health Nutrition 5 (3), 441–448. Nago, E.S., Lachat, C.K., Huybregts, L., Roberfroid, D., Dossa, R., Kolsteren, P., 2010. Deaton, A., Grosh, M., 2000. Consumption. In: Grosh, M., Glewwe, P. (Eds.), Designing Food, energy and macronutrient contribution of out-of-home foods in school-going Household Survey Questionnaires for Developing Countries: Lessons From 15 Years adolescents in Cotonou, Benin. Br. J. Nutr. 103 (2), 281–288. of the Living Standards Measurement Study. Oxford University Press (for the World O’Dwyer, N.A., Gibney, M.J., Burke, S.J., McCarthy, S.N., 2005. The influence of eating Bank), Oxford, U.K, pp. 91–133. location on nutrient intakes in Irish adults: implications for developing food-based Gewa, C.A., Murphy, S.P., Newman, C.G., 2007. Out-of-home food intake is often omitted dietary guidelines. Public Health Nutrition 8 (3), 258–265. from mothers’ recalls of school children’s intake in rural Kenya. J. Nutr. 137, Oguntona, C.R.B., Tella, T.O., 1999. Street foods and dietary intakes of Nigerian urban 2154–2159. market women. Int. J. Food Sci. Nutr. 50 (6), 383–390. Gibson, J., Rozelle, S., Huang, J., 2003. Improving estimates of inequality and poverty Pradhan, M., 2001. Welfare analysis with a proxy consumption measure: Evidence from a from Urban China's Household Income and Expenditure Survey. Rev. Income Wealth repeated experiment in Indonesia. Tinbergen Institute Discussion Paper, Amsterdam, 49 (1), 53–68. Available at: < http://dspace.ubvu.vu.nl/bitstream/handle/1871/9494/01092.pdf? Guthrie, J.F., Lin, B.H., Frazao, E., 2002. Role of food prepared away from home in the sequence=1 > (Accessed 10/27/14). American diet, 1977–78 versus 1994–96: changes and consequences. J. Nutr. Educ. Ravallion, M., Chen, S., 2011. Weakly relative poverty. Rev. Econ. Stat. 93 (4), Behav. 34 (3), 140–150. 1251–1261. Jolliffe, D., 2001. Measuring absolute and relative poverty: the sensitivity of estimated Smith, L.C., 2013. The Great Indian Calorie Debate: Explaining Rising Undernourishment household consumption to survey design. J. Econom. Social Meas. 27, 1–23. during India’s Rapid Economic Growth. IDS Working Paper 430. Kakwani, N., 2000. On measuring growth and inequality components of poverty with Smith, L., Dupriez, O., Troubat, N., 2014. Assessment of the Reliability and Relevance of application to Thailand. J. Quant. Econ. 16, 67–68. the Food Data Collected in National Household Consumption and Expenditure Kant, A.K., Graubard, B.I., 2004. Eating out in America, 1987–2000: trends and nutri- Surveys. International Household Survey Network, Food and Agriculture tional correlates. Prev. Med. 38 (2), 243–249. Organization of the United Nations, World Bank. Kearney, J.M., Hulshof, K.F., Gibney, M.J., 2001. Eating patterns – temporal distribution, Sujatha, T., Shatrugna, V., Narasimha Rao, G.V., Ghenna Krishma Reddi, G., Padmavathi, converging and diverging foods, meals eaten inside and outside of the home – im- K.S., Vidyasagar, P., 1997. Street food: an important source of energy for the urban plications for developing FBDG. Public Health Nutrition 4 (2B), 693–698. worker. Food Nutr. Bull. 18 (4). Le Francois, P., Calamassi-Tran, G., Hebel, P., Renault, C., Lebreton, S., Volatier, J.L., United States Department of Agriculture, Economic Research Service, 2011. 1996. Food and nutrient intake outside the home of 629 French people of fifteen “International Food Consumption Patterns, 2005”, Washington, DC: USDA/ years and over. Eur. J. Clin. Nutr. 50 (12), 826–831. ERS < www.ers.usda.gov/datafiles/International_Food_Consumption_Patterns/2005_ Lin, B.-H., Guthrie, J., 2012. Nutritional quality of food prepared at home and away from Data/Table9.xls > . home, 1977–2008. Economic Information Bulletin, No. 105, USDA. Vandevijvere, S., Lachat, C., Kolsteren, P., Van Oyen, H., 2009. Eating out of home in Liu, Y., Zhai, F., Popkin, B.M., 2006. Trends in eating behaviors among Chinese children Belgium: current situation and policy implications. Br. J. Nutr. 102 (06), 921–928. (1991–1997). Asia Pacific J. Clin. Nutr. 15 (1), 72–80. Van’t Riet, H., den Hartog, A.P., van Staveren, W.A., 2002. Non-home prepared foods: Lozada, M., Sanchez-Castillo, C.P., Cabrera, G.A., Mata, I.I., Pichardo-Ontiveros, E., Villa, contribution to energy and nutrient intake of consumers living in two low-income A.R., James, W.P.T., 2008. School food in Mexican children. Public Health Nutrition areas in Nairobi. Public Health Nutrition 5 (4), 515–522. 11 (9), 924–933. 156