Improving Household Consumption and ...

Still waiting for Godot? Improving Household Consumption and Expenditures Surveys (HCES) to enable more evidence-based nutrition policies

John L. Fiedler, Calogero Carletto, and Olivier Dupriez Abstract Background. The constrained evidence base of food and nutrition policy-making compromises nutrition programs. Nutrition policy-making must do better than relying exclusively on Food and Agriculture Organization Food Balance Sheets. The strategy of relying on observedweighed food record or 24-hour recall surveys has not proven practical either; they remain few in number, generally not nationally representative, and of dubious external validity. Although Household Consumption and Expenditures Surveys (HCES) have shortcomings, they are increasingly being used to address this information gap. Objective. To promote dialog within the nutrition community, and between it and the greater community of HCES stakeholders, in order to identify their shared agenda and develop a strategy to improve HCES for analyzing food and nutrition issues. Methods. The diverse origins and objectives of HCES are described, the evolution of their use in addressing food and nutrition issues is traced, and their shortcomings are identified. Results. The causes, relative importance, some potential solutions, and the strategic implications of three distinct categories of shortcomings are discussed. Elements of a possible approach and process for strengthening the surveys are outlined, including identifying best practices, developing guidelines and more rigorously analyzing the tradeoffs involved in common, key survey design and implementation decisions. John L. Fiedler is affiliated with HarvestPlus, the Biofortification Research Project implemented by the International Center for Tropical Agriculture (CIAT) and the International Food Policy Research Institute (IFPRI), Washington, DC; Calogero Carletto is affiliated with the Development Research Group of the World Bank, Washington, DC; Olivier Dupriez is affiliated with the Development Data Group of the World Bank, Washington, DC. Please direct queries to the corresponding author: John L. Fiedler, HarvestPlus, IFPRI, 2033 K Street NW, Washington, DC 20006; e-mail: [email protected].

S242

Conclusions. To date, the nutrition community’s role in most HCES has been as a passive user of secondary data. The nutrition community must become more involved in the design, implementation, and analysis of HCES by identifying criteria for prioritizing countries, establishing assessment criteria, applying the criteria in retrospective assessments, identifying key shortcomings, and recommending alternatives to ameliorate the shortcomings. Several trends suggest that this is a propitious time for improving the relevance and reliability of HCES.

Key words: Evidence-based policy, fortification, HCES,

Household Consumption and Expenditures Surveys, household surveys, nutrition, nutrition policy

Introduction Food and nutrition policy-making has long been constrained by its evidence base. For the past several decades, the Food and Agriculture Organization (FAO) Food Balance Sheets (FBS) have been the principal data source for monitoring food security at the global level and a primary advocacy tool for improved food and nutrition policy-making. FBS, however, only contain data about national food availability; they do not provide information on access to available food or about how it is distributed within the country [1]. Nor do they provide information about how much of the available food is consumed or by whom; i.e., they do not provide information on household or individual food consumption patterns. Despite these well-recognized and considerable shortcomings, FBS play a fundamental role in shaping food and nutrition policies due to the lack, or perceived lack, of better data from other data sources on these quintessentially important measures. Few countries have food consumption data obtained from what nutritionists generally regard as the gold standard method—viz., observed-weighed food records (OWFR)—because they are expensive and difficult to conduct [2]. In fact, very few countries have

Food and Nutrition Bulletin, vol. 33, no. 3 (supplement) © 2012, The United Nations University. Delivered by Publishing Technology to: The World Bank IP: 192.86.100.77 on: Tue, 27 Aug 2013 22:08:53 Copyright (c) Nevin Scrimshaw International Nutrition Foundation. All rights reserved.

S243

Elements of a strategy and process for improving HCES

even what nutritionists would regard as the “next best” approach, 24-hour recall surveys, because they too are costly and complex [3]. Basing food and nutrition policy on estimates of national food availability, rather than data on household or individual food consumption, leaves enormous room for error and misinterpretation. Many nutrition programs are probably suboptimally designed and implemented due to these data shortcomings and to the lack of alternative data sources. Moreover, in many countries, the absence of food consumption data has created doubts and uncertainties about food and nutrition issues and has discouraged the development of programs to combat malnutrition. Nutrition policymaking must do better than relying only or primarily on FBS, but insisting that the only alternatives are OWFR or 24-hour recall surveys and waiting for them to be implemented on a large scale has been costly in terms of malnutrition-related morbidity and mortality. In terms of unfulfilled expectations, it is reminiscent of Samuel Beckett’s Waiting for Godot, a play from the theater of the absurd in which the two protagonists wait expectantly and incessantly, but in vain, for the arrival of Godot, a savior who will “take care of everything.” At the same time, however, one must acknowledge that poor data are not necessarily better than no data. Over the past decade, a growing body of work has demonstrated that data on food consumption obtained from a family of specialized and multipurpose surveys—which we will refer to as Household Consumption and Expenditures Surveys (HCES)— can contribute to addressing the food consumption information gap and to making nutrition policy more evidence based [4–22]. How well these surveys measure food consumption, however, what their key limitations are, and the extent to which their shortcomings can be ameliorated or eliminated are critically important issues. The nutrition community has become a new HCES stakeholder and brings a new set of concerns and issues about these data, as well as new methods and applications for using them, new criteria for assessing their quality and relevance, and a growing, still evolving, set of suggestions for improving them. This paper is intended to promote the dialog both within the large and diverse nutrition community, and between the nutrition community and the greater community of HCES stakeholders, by articulating how the nutrition community may want to see HCES instruments “improved,” without creating false expectations. The development of a common understanding of the nutrition community’s “needs” is the sine qua non for crafting a strategy for moving the household survey reform process “forward”—as judged from a nutrition perspective. The organization of the paper is as follows: the next section reviews the growing use of HCES in food and nutrition analyses. A general discussion of the diversity

of HCES and the relative strengths and weaknesses of these surveys follows. The fourth section discusses some of the priority reform agenda items and proposes a general two-tiered process for refining the agenda and implementing it.

How we got here: Tracing the growth in HCES-based analyses of food and nutrition issues The pioneer in the use of HCES to analyze food and nutrition issues was the DAta Food NEtworking (DAFNE) Project of the University of Athens [16–19]. It began using national household budget surveys to monitor trends in food habits and food availability in 1987. The DAFNE-ANEMOS Project harmonized HCES data and developed software to conduct comparative analyses of food trends in its network of 28 European countries. The approach improved upon the FBS by going beyond national measures of food availability to provide information on household acquisition and apparent consumption of food.* It used food composition table data to quantify the caloric content of the foods acquired by the household and, to take into account the demographic composition of the household, used the Adult Male Consumption Equivalent (AME) to estimate average caloric intake [20]. The second benchmark in the use of HCES for food and nutrition analyses was the International Food Policy Research Institute’s Assessing Food Insecurity (AFINS) Project [21, 22]. From 2000 to 2006, AFINS analyzed the HCES of 20 developing countries in subSaharan Africa and Asia and developed guidelines for using HCES databases to more precisely measure food insecurity. Like DAFNE, it developed householdlevel indicators of caloric intake and nutrient intake inadequacy, using the AME to adjust for household composition. AFINS demonstrated the usefulness of HCES in better understanding and measuring national food security, in particular, and provided important insights about the distribution of food within a country—geographically, as well as by a variety of household characteristics (e.g., income quintile, rural vs. urban residence, region or state). AFINS was a one-time assessment of getting beyond FBS by developing a new approach to using already existing databases, and published an excellent set of practical guidelines, Measuring Food Security Using Household Expenditure Surveys [22]. But AFINS was not part of a longer-term plan to improve food security indicators, and it did not have an institutional home that would facilitate more routine and wider use of the methods it had developed. When * “Apparent consumption” assumes that food acquisitions reported during the recall are consumed during the recall period.

Delivered by Publishing Technology to: The World Bank IP: 192.86.100.77 on: Tue, 27 Aug 2013 22:08:53 Copyright (c) Nevin Scrimshaw International Nutrition Foundation. All rights reserved.

S244

J. L. Fiedler et al.

the project ended, concerted work on the application of the new methods that had been developed ended. The FAO approach to measuring food insecurity or undernourishment has long been based on FBS data in combination with HCES data. The FAO approach, however, has made only limited use of HCES data (within the framework of a two-parameter log-normal distribution model) to provide information about the distribution of calorie consumption and caloric requirements. In 2008, motivated by the need to assess food insecurity at a subnational level, FAO initiated a new approach to how it employed HCES when it began routinely analyzing these databases to estimate household-level food access, food security, and inadequate nutrient intake at a subnational level. That same year, FAO published a set of guidelines for conducting this type of analytic work, Measuring Hunger at Subnational Levels from Household Surveys Using the FAO Approach [23]. In the 4 years since the publication of the manual, FAO has analyzed the HCES of 46 countries, but it has done little to publicize its use of this new routine approach, publishing the analyses of only a handful of the countries [23–25]. At least for the time being, reliance on HCES is not intended to supplant FBS, but rather to complement them.* There have been a growing number of other analyses of HCES that incorporate a further elaboration of the basic DAFNE-AFINS approach. The hallmarks of this work are making inferences about individual household members’ consumption of foods (usually based on the assumption that food is distributed within the household in direct proportion to household members’ shares of the household’s total AMEs); and analyzing household members’ individual nutrient intakes, which in some of the studies included assessing the adequacy of nutrient intake, a proxy measure of nutrient deficiency [6–15, 26]. The event giving rise to this supplement—a workshop convened by a working group of 20 technical experts from international agencies and nongovernmental organizations to inventory and juxtapose alternative food consumption measurement methods—may be regarded as another development in the use of HCES to inform food and nutrition policy; viz., that it has achieved critical thresholds in terms of acknowledged momentum, credibility, and promise as a means to effectively overcome a portion of the * In fact, food consumption data remain an indispensable part for the estimation of at least one of the parameters of the distribution underlying FAO’s parametric approach and may serve a key role in validating the one parameter derived from FBSs; viz., the mean of the distribution. Thus—abstracting for a moment from issues of “ideological” disagreements on the use of a parametric approach and the feasibility of improving the data underlying the FBS—improving the quality and availability of food consumption data from HCES may be seen as being consistent, and not in conflict, with the FBSbased approach.

nutrition information gap. A still small, but growing, number of studies have juxtaposed 24-hour recall and HCES data and have found relatively high levels of consistency between the surveys in quantifying the proportion of households reported to be consuming most food items [27–30]; the proportion of households purchasing most food items [27–30]; the nutrient density of most food items consumed [30–32]; the nutrient density of most food items purchased [30–32]; and no statistically significant differences between estimated zinc and vitamin A intakes, but a difference in estimated iron intake [33]. This is encouraging, but clearly, the external validity of these studies is questionable and more comparative analyses are warranted before definitive conclusions can be drawn. Both of these appraisals, and the entire body of work using HCES to analyze food and nutrition that has been done to date, have consisted (with one important exception [34]) exclusively of retrospective studies, and the role of the analysts—and by extension the nutrition community—has been one of a passive user of secondary data. What if the nutrition community became a more active HCES stakeholder? How might HCES be modified to improve their reliability and precision in measuring food acquisition and consumption? How “good” could they become? Could they be used to estimate nutrient intakes with “adequate” precision or will they, at best, only constitute a source of complementary data? Answering these questions requires an understanding of the key characteristics and recent developments of the “HCES world,” to which the discussion now turns.

The large, heterogeneous family of HCES: A plethora of opportunities for improving food consumption measurement Over the past two decades there have been dramatic increases in the number, quality, and availability of HCES in developing countries. The World Bank’s 1990 World Development Report presented original cross-country analyses of HCES from 22 countries, with a single survey for each country. Today, 116 countries have more than 700 surveys, an average of 6 per country [35]. The dramatic improvement in the availability of consumption data has been accompanied by widening discrepancies in terms of instruments and methods, which have been particularly common and acute in measuring food consumption. With varying levels of success, countries measure food consumption (or proxies thereof) using different surveys, including Household Budget Surveys (HBS), Household Food Consumption Surveys, Individual Dietary Surveys, Nutritional Status Surveys, Household Income and Expenditure


S245


Surveys (HIES), Living Standards Measurement Studies (LSMS), Integrated Household Surveys (IHS), Priority Surveys (PS), Core Welfare Indicator Questionnaire (CWIQ), Comprehensive Food Security and Vulnerability Assessment (CFSVA) Surveys, and Welfare Monitoring Surveys (WMS). These surveys have diverse objectives and, as such, any modification to better serve the needs of the nutrition community must take that into account to maintain the integrity of these original objectives. HBS and HIES were initially designed primarily to provide input into the consumer price indices and national accounts. Others—including LSMS, IHS, and other welfaremonitoring surveys—are mainly intended to generate multitopic datasets to measure and understand poverty and other dimensions of welfare, including food and nutrition security. Some surveys—such as PS, CWIQ, CFSVA, and WMS—are primarily concerned with monitoring select socioeconomic indicators over time, emphasizing “breadth” over “depth,” and collect only partial consumption data. In contrast, surveys like the LSMS, HBS, HIES, and IHS are designed to collect comprehensive food (and other) consumption information, although in the case of HBS and HIES, the emphasis may be on expenditures or on particular subnational samples. The marked variability of key survey and food and nutrition-related data characteristics in 74 surveys from 69 countries is exemplified in table 2 in the paper in this Supplement by Fiedler et al. [36]. Although there are international guidelines and recommendations for the design and implementation of each of these surveys [37, 38], they are specific to each type of survey, are generally not prescriptive, and leave much flexibility to national survey statisticians. As a result, available household consumption and expenditure datasets lack coherency and reflect the diversity of their objectives, questionnaire design, data collection methods, and data-processing practices. Consequently, there are issues of data reliability and relevance for some surveys, and poor comparability of the data over time, as well as across and within countries. Which type of HCES would be useful and which would be the “best” for collecting food consumption and expenditure data depends on concomitant consideration of what are the priority food and nutrition issues in the country in question and which types of HCES are available in the country.* Box 1 shows some criteria for assessing an HCES for analyzing food consumption. The “best” HCES, of course, is * This is a simplification made for expository purposes. It assumes that the food consumption analysis will be based on a single HCES and that there will be efforts to improve the survey’s measurement of food consumption, and it does not take into account other considerations (such as the technical capacity, timing, and budgetary constraints) of either current HCES stakeholders or those who are working to modify the survey.

the one that is thought to have fewer and less serious “shortcomings.” The limitations of an HCES in analyzing one nutrition problem, however, may not be a shortcoming for analyzing another, as exemplified in box 2. Selecting a particular HCES, therefore, requires prioritizing the specific objective(s) for collecting food consumption data, while taking into account (among other things) the strengths and shortcomings of the various possible survey vehicles. Particularly in light of the cross-country variability in HCES instruments, this is a country-specific determination. Articulating the shortcomings of available HCES concomitantly identifies concerns that (from the nutrition community’s perspective) “should be” included in the HCES reform agenda. Some common limitations that could be eliminated or ameliorated with a modicum of relatively minor modifications include getting more precise information about individual household member characteristics, such as distinguishing infants less than 6 months old from those 6 to 11 months old, and identifying whether or not women are pregnant or lactating so that the very different Estimated Average Requirements (EARs) of these nutrition priority populations can be analyzed; and reviewing HCES food lists to ensure that they include all “important” foods (including popular foods that may include substantial amount of candidate fortification vehicles, such as wheat flour or maize meal) and ensuring that food item categories are defined as similar to food composition table entries as possible, so as to facilitate their being more unambiguously matched, thereby ensuring more accurate estimation of the nutrient content of foods. In countries where the HCES questionnaire contains more limited food items, consideration should be given to developing screens to selectively ask additional questions to enable the amount of a food consumed to be more precisely distinguished from the amount acquired during the recall period, especially for the more “important” foods. Although developing anything approaching a comprehensive list of HCES shortcomings is beyond the scope of this paper, some areas of concern are identified in boxes 1 and 2. The nutrition community needs to develop a more comprehensive and prioritized list. There exist three sets of guidelines for analyzing HCES food data (for different purposes) that provide useful suggestions [12, 22, 24], which, together with several detailed discussions [10, 13, 21, 23, 34, 36–40], provide a good starting point.

Going forward: Elements of a strategy and a possible process The nutrition community needs to become a more active actor in the HCES arena to advance its agenda. Recognizing that HCES have already been implemented


S246


in most countries for more than a decade and that they already have many stakeholders, the latitude for introducing changes will be constrained. The respective surveys obviously must retain their fundamental ability to fulfill their primary objectives. It should also be expected that data producers in countries where HCES have been institutionalized will resist changes that would introduce breaks in their data series. Nevertheless, there is much common ground between the nutrition community and other stakeholders’ priorities. Most of the purposes for which consumption data are collected—poverty analysis, national accounts, price indices, and others—would benefit from an improved measurement of household food consumption. Furthermore, most data producers and survey sponsors will welcome opportunities to demonstrate

the relevance and increase the use of the data, thereby justifying their budgets and investments in data collection. For its part, the nutrition community (among other things) can help to respond to the food price crisis-sparked, growing demand for more and better information about food and nutrition. It can also provide expertise needed to improve survey methods and show how some of the shortcomings of HCES can be ameliorated or eliminated, with relatively minor changes to existing methods. It is our position that modest “quick wins” are possible in the short term and that in the long run greater coherence in both methods and processes can be achieved. Improving the relevance of survey questionnaires for food and nutrition analysis may be throttled by the lack of easy access to well-documented HCES microdata.

BOX 1. Criteria for selecting Household Consumption and Expenditures Surveys (HCES) and assessing food consumption data: Potential HCES shortcomings Household versus individual food acquisition and consumption Identifying individual rather than household food acquisition and consumption requires more time and risks interviewer fatigue, and is not practically feasible in HCES. Studies have shown that the intrahousehold distribution of food may be approximated by each household member being assigned a share of the household’s total food supply that is in direct proportion to the individual’s share of the household’s total energy requirements (as measured by AMEs), but there is little evidence about the distribution of micronutrients. Food diary or recall Implementing diaries requires respondents to record daily any food items purchased or consumed over the reference period. Although this method is generally regarded as more accurate than recall, its accuracy has been shown to vary by level of literacy, size of family, and level of urbanization. Diaries are also relatively expensive, requiring more house visits by an enumerator and more field supervision. Food acquisition and/or consumption Estimating actual intake requires recording data on actual intake utilization and wastage, which is time consuming and error prone; consequently, often information on food acquisition from different sources (viz., whether purchased, from own production, or acquired through other means) is collected. It is possible to collect information on both acquisition and consumption, but it will greatly increase interview time and fatigue. Recall time period, bulk food purchases, and estimated food consumption The possibility of large purchases and differences in the frequency of acquisition can result in consumption taking place outside of the recall period, undermining the use of purchases as a proxy for consumption. Increasing the length of the recall period can ameliorate this possibility, but at the risk of increased recall error.

Survey implementation and seasonality Food consumption surveys differ in terms of the duration of data collection. While the data collection periods of most HBS and HIES span a 12-month period, with representative subsamples collected each month or quarter, others collect data for only a few months. In order to avoid issues of seasonality, surveys may be conducted during “average” months, when consumption is not unusually low or high, or they may spread the data collection activity over 12 months to control for seasonality, or collect consumption information during the “usual month.” Analytical tools using external data sources have also been used to smooth consumption in the presence of large seasonality. Food away from home Food eaten outside the home in the form of prepared or packaged meals is usually not adequately captured and often not asked about. This source of underreporting of consumption, which is already among the most important shortcomings of HCES, will be exacerbated by increasing urbanization and increasing income. Different survey instruments deal with the issue in different ways, but most solutions adopted are ad hoc and unsatisfactory. Length and composition of the food item list Inadequate lists risk underreporting consumption. It is particularly important for nutrition analysis that food groups be adequately disaggregated. However, comprehensive lists risk increasing interview time and fatigue and may bias consumption estimates upward. Some HCES food lists include both raw and processed foods, requiring the development of primary equivalents, which may require developing recipes, increasing data processing and analysis time and complexities. Not including “important” processed foods results in underestimates of the coverage of some fortification vehicles, such as wheat and maize flour.


S247


Historically, technical, financial, legal, ethical, and sometimes political issues have raised obstacles to microdata access, and the general rule has been that these data are not publicly available or are released with considerable delays. The past few years have seen major improvements in access, which are evidenced in part by the increasing number of statistical agencies (and other data producers) that have established national data archives. This progress has been made possible by a combination of factors: the development of appropriate metadata standards for the documentation of microdata (in particular the Data Documentation Initiative [DDI]) and related software applications; an increased concern by national statistical agencies for the preservation and use of their valuable datasets; the increased support provided by international agencies and others to build capacity and to provide technical support in microdata management*; and a culture change of “data liberation” or “open data.” These trends, and their causes and promoters, are additional factors the nutrition community should consider in developing its approach: it should add its voice to those stakeholders who are promoting the growing global data access imperative.** Turning to the more technical aspects of devising a strategy to address the shortcomings of HCES in analyzing food and nutrition issues, it is useful to consider the following characterization of shortcomings, bearing in mind that there is some overlap. General HCES shortcomings. The major survey design issues identified in box 1. These shortcomings are due to the limitations or tradeoffs related to the general data collection approach adopted in the HCES, are generally characteristics of the specific type of HCES, and may be found to be issues in both the food and the nonfood sections of the surveys. Shortcomings attributable to the particular requirements of undertaking more nutrition-specific analyses. Limitations that are of lesser or no concern to those who are not nutrition analysts. Country-specific shortcomings. Limitations attributable to the particular way a specific country has asked a questionnaire, or a series of questions, or the particular way it implemented the survey, or processed and disseminated the data. The common ground of all HCES stakeholders— the nutrition community and others—is the need to address general HCES shortcomings. In particular, getting the food list and food quantities “right” is everyone’s concern. Clearly, the “devil will be in the details”: the definition of “right” will vary by stakeholder group. But there are general threats to the relevance of HCES * See www.ddialliance.org for more information on the DDI standard and www.ihsn.org for information on related software and guidelines. ** See, for example, the World Bank—PARIS21 Accelerated Data Program, www.ihsn.org/adp.

that can and should be addressed—such as the inadequate capturing of data on food consumed away from home or households’ consumption of their own food production, or the inadequate disaggregation of specific food groups. These must be among the top priorities. Other quantity-related general HCES issues include how to improve the conversion of nonstandard units of measure, the growing nonresponse rates of HCES, BOX 2. Household Consumption and Expenditures Surveys (HCES) shortcomings vary by type of nutrition analysis: Fortification versus nutrient intake Food purchases versus food quantities If the key objective is to identify a food vehicle for a fortification program, it may not be as important that a particular type of HCES collects data on only the value of food purchases, and not the quantities of foods purchased and/or consumed. The number and percentage of households consuming the food may be adequate. If, however, the intention is to analyze nutrient intake, this will be a more serious shortcoming, involving considerably more work (assembling price data that take into account subnational geographic and seasonal variation and dividing the food values by these prices to obtain estimated quantities) and introducing considerably more imprecision. Select versus comprehensive food list If the objective is to design a fortification program, a specialized survey that collects data on only a few select foods is adequate. For nutrient intake analysis, however, as complete an accounting of all food intakes as possible is necessary; this may require balancing having a long food item list with risking interviewee and interviewer fatigue. (Whether striking that balance entails too great a compromise of nutrient intake estimates is uncertain.) Food acquisition versus food consumption—1 For purposes of selecting a fortification vehicle program, it is useful to identify the source of foods to enable distinguishing food that is purchased and thus more amenable to being fortified, from that which is consumed from own production. Estimating nutrient intakes does not require differentiating food sources. Food acquisition versus food consumption—2 For purposes of setting or assessing fortification levels, it is necessary to differentiate consumers and nonconsumers and to know the average daily quantity of the food vehicle they consume. Analysis of HCES, with their mix of acquisition and consumption data, usually assumes that acquisition during the recall period is equal to consumption during the recall period. Bulk food purchases can result in some portion of the item being consumed outside the reference period and undermine the accuracy of apparent consumption estimates. For fortification purposes, it is possible to ask additional questions about the food fortification vehicle to disentangle acquisition and consumption.


S248


systematic underreporting in food consumption, how to better deal with seasonality and the identification of usual consumption or intake and the annualization of food consumption, the tradeoffs involved in opting for recall versus diary and in the choice of reference period, and asking about consumption as well as or in addition to acquisition. Some 24-hour recall surveys have incorporated HCES-like questions about food acquisition (by source) in order to improve understanding of the relative strengths and weaknesses of the two methods and to enable cross-fertilization [30, 41–43]. Addressing shortcomings attributable to the particular requirements of undertaking more nutritionspecific analyses is where the nutrition community will confront the greatest challenges. The most modest goals are probably those that would be required to improve HCES for analyzing food group diversity indicators [44]. The most ambitious goal—which is not (yet?) shared by the entire nutrition community—is to use HCES to conduct nutrient intake analysis. The major issues here are how to measure or model the intrahousehold distribution of food, together with how to improve the current common approach of simply assuming that all foods are allocated to household members based on their share of total household AME. The challenge to the nutrition community will be to figure out ways to improve on the AME assumption about the intra-household distribution without greatly increasing costs or interview time requirements. This is another obvious area in which comparative analysis of HCES and OWFR or 24-hour recall surveys is likely to provide insights. The concept of “apparent food consumption” should be analyzed to better understand the limitations inherent in using acquisition data to proxy food consumption, and alternative methods should be investigated for distinguishing consumption and acquisition. A small number of countries attempt to collect complete information on both, and analyses have been published juxtaposing the measures from Armenia, Kenya, and Cape Verde [41–43]. The results have been useful, although somewhat mixed. Additional work needs to be done to assess the tradeoffs of increased precision and additional resource requirements. An alternative approach that might be considered would be to “adjust” implausible daily consumption levels due to bulk purchasing using, for instance, data from OWFR or 24-hour recall when available, or more simply, introducing small changes in questionnaires to capture infrequent bulk purchases in order to better characterize “usual consumption” or “usual intake” or better annualized food consumption. That leaves the third category of shortcomings; those that are country-specific in nature. These (and conversely “better practices”) may be related to questions and questionnaire design, or they may be related to implementation and data processing. Given that

there is relatively little documentation on the actual implementation of HCES (both fieldwork and data processing), major improvements could be achieved by promoting stronger protocols based on state-ofthe-art practices in select countries. The idiosyncratic nature of this category prompts taking into account the process by which this work is done. Recognizing that the survey design and method have to be adapted to each country’s policy needs, objectives, and strategies and responding to each country’s technical and financial constraints, and building on a country’s own experiences and good practices, the objective should be to present options and possible solutions to statistical agencies interested in increasing or broadening the relevance of their data. The work could begin with retrospective analysis of the HCES database, questionnaires, interviewer guides, supervisor guides, and other supporting documentation. In a given country, it will be important to bring this discussion into existing National Strategies for the Development of Statistics other existing strategic frameworks for improving data collection efforts. Technical working groups and data user groups that coordinate and provide technical inputs to household survey activities already exist in most countries, but in countries where to date the nutrition community has not been involved or adequately involved in the design, implementation, or analysis of HCES, it should join existing working groups or form new ones to review HCES materials and more actively participate in this process. At the global level, a strong institutional mandate to promote both short-term, quick fixes and a longer-term agenda will be needed. In the short term, some of the low-hanging fruit can be exploited by conducting cross-country analyses to juxtapose countries’ approaches and identify lessons and better practices (see Fiedler et al. [40] for examples). In this process, the nutrition community should contribute by establishing criteria by which to assess HCES, developing methods to operationalize them, applying the criteria in the retrospective assessments of HCES, identifying key shortcomings and their causes, and recommending alternative ways to address shortcomings. If such a process were undertaken and multiple, retrospective country reviews were conducted, they could be synthesized into an across-country analysis describing and quantifying common problems and shortcomings, best practices could be distilled and prioritized, suggestions could be made for changes in questionnaires and data processing procedures, and guidelines could be developed. As of September 2011, FAO, the International Household Survey Network (IHSN), and the World Bank are working on developing a memorandum of understanding to undertake a joint assessment of HCES, with the objective of deriving recommendations and suggestions for improving the reliability and relevance of food consumption data


S249


for various key categories of stakeholders, who would be invited to participate in the project. From a longer-term perspective, an additional step in the process would consist of taking this work forward, in a proactive mode. This could consist of working with countries and across agencies to conduct rigorous methodological work on specific measurement issues. The measurement of food consumed away from home is a critical issue, which is expected to be covered under the proposed collaboration between FAO and the World Bank. Other potential issues include the assessment of the relative merits of various data collection methods; for example, the use of the recall method versus diary, short versus long food item lists, the optimal length of the recall period, and the additional requirements of improving food consumption versus nutrient intake data. The work of Beegle and colleagues in Tanzania may serve as a model [34]. The work could include addressing the issue of the external validity of their findings by analyzing some of the same, fundamental (general HCES shortcomings) issues in other countries. As Beegle and colleagues noted in their report, how well a particular survey approach works is conditioned by the characteristics of a country. This suggests that it might be useful to identify prototypes of countries classified by these characteristics in combination with nutrition status. A final step in this proactive phase of the process might involve the implementation of pilot surveys, including technical support to implement or test and refine recommendations and produce guidelines for best practices. Is this a propitious time to promote such an agenda? On the one hand, due in part to tightening budget constraints, countries and donors alike are progressively expanding the scope of individual surveys in an attempt to reduce the number of often overlapping data collection efforts, while they push for greater standardization of data collection instruments. The distinct ability of many HCES to enable multiple nutrition interventions to be analyzed—supplementation, biofortification, as well as fortification—is an important characteristic that is just starting to be appreciated and its potential harnessed. The growing significance

and concerns about the overlap of nutrition programs and the need to manage nutrition programs more proactively as a portfolio of programs can be expected to fuel the use of HCES. Together with the new data access imperative, these trends bode well for improving the content, comparability, and relevance of household consumption surveys for food and nutrition security analyses. On the other hand, the increased attention currently being given to food consumption has created a window of opportunity for more and broader changes than would otherwise likely have existed. More stakeholders are demanding more and better information about food and nutrition. Initiatives such as the current revision of the FAO methodology to measure undernourishment may provide the necessary momentum. There is mounting evidence that HCES are already useful for addressing many important nutrition issues and understanding the root causes of malnutrition and food insecurity. To make them an even more powerful tool, the nutrition community must become more involved in their design, implementation, and analysis. Making nutrition more evidence-based is essential to becoming better able to make nutrition programs more effective, while enabling nutrition programs to become more accountable. Making nutrition more evidence-based is something that needs to be done even if the FAO/ IHSN/World Bank initiative fails to materialize. But this initiative, while still being defined and negotiated, is a unique and timely opportunity to rally broader support that could finally mean the arrival of Godot.

Acknowledgments The views and opinions presented in this paper are entirely those of the authors and do not necessarily reflect those of their organizations. John Fiedler acknowledges support from the Nutrition and Economic Research Support for Grand Challenge #9 Projects Grant OPP52013 to HarvestPlus from the Bill & Melinda Gates Foundation.

References 1. Riley F, Mock N, Cogill B, Bailey L, Kenefick E. Food security indicators and framework for use in the monitoring and evaluation of food aid programs. Food and Nutrition Technical Assistance (FANTA) Project document. Washington, DC: Academy for Educational Development, 1999. Available at: www.fantaproject.org. Accessed 7 May 2012. 2. Gibson RS. Principles of nutritional assessment, 2nd ed. New York: Oxford University Press, 2005. 3. Fiedler JL, Martin-Prével Y, Moursi M. Comparing the costs of 24-hour recall Surveys and Household

Consumption & Expenditures Surveys (HCES): a nutrition analysis perspective. HarvestPlus Working Paper. Washington, DC: HarvestPlus/International Food Policy Research Institute and International Relief and Development, 2011. 4. Coates J, Colaiezzi B, Fiedler JL, Wirth J, Lividini K, Rogers B. A program needs-driven approach to selecting dietary assessment methods for decision-making in food fortification programs. Food Nutr Bull 2012;33:S146–56. 5. Stein AJ, Nestel P, Meenakshi JV, Qaim M, Sachdev HPS, Bhutta ZA. Plant breeding to control zinc deficiency


S250

6. 7. 8.

9. 10.

11.

12.

13.

14.

15.

16.

17.

18.

19.


in India: how cost-effective is biofortification? Public Health Nutr 2007;10:492–501. Imhoff-Kunsch B, Flores R, Dary O, Martorell R. Wheat flour fortification is unlikely to benefit the most needy in Guatemala. J Nutr 2007;137:1017–22. Stein AJ, Menakshi JV, Qaim M, Nestel P, Sachdev HPS, Bhutta ZA. Potential impacts of iron biofortification in India. Soc Sci Med 2008;66:1797–808. Meenakshi JV, Johnson NL, Manyong VM, DeGroote H, Javelosa J, Yanggen DR, Naher F, Gonzalez C, García J, Meng, E. How cost-effective is biofortification in combatting micronutrient malnutrition? An ex ante assessment. World Dev 2010;38:64–75. Stein AJ, Sachdev HPS, Qaim M. Genetic engineering for the poor: golden rice and public health in India. World Dev 2008;36:144–58. Fiedler JL, Smitz MF, Dupriez O, Friedman J. Household income and expenditure surveys: a tool for accelerating the development of evidence-based fortification programs. Food Nutr Bull 2008;29:319–32. Fiedler JL, Macdonald BA. A strategic approach to the unfinished fortification agenda: feasibility, costs and cost-effectiveness analysis of fortification programs in 48 countries. Food Nutr Bull 2009;30:283–316. Fiedler JL. Strengthening Household Income and Expenditure Surveys as a tool for designing and assessing food fortification programs. Working Paper No. 001 of the International Household Survey Network. October 2009. Available at: http://www.internationalsurveynetwork.org/home/index.php?q=activities/ working-papers. Accessed 7 May 2012. Fiedler JL, Helleranta M. Recommendations for improving Guatemala’s food fortification program based on household income and expenditure survey (HIES) data. Food Nutr Bull 2010;31:251–69. Fiedler JL, Afidra R. Vitamin A fortification in Uganda: comparing the feasibility, coverage and cost-effectiveness of vegetable oil and sugar. Food Nutr Bull 2010; 31:193–205. Fiedler JL, Babu S, Smitz MF, Lividini K, Bermudez O. Indian social safety net programs as platforms for introducing wheat flour fortification: a case study of Gujarat, India. Food Nutr Bull 2012;33:11–30. Trichopoulou A, Lagiou L, eds. Methodology for the exploitation of HBS food data and results on food availability for five European countries. Data Food Networking (DAFNE I), Directorate-General, Science, Research and Development, European Commission, Athens, 1997. Available at: http://www.nut.uoa.gr/ downloadsENG.html. Accessed 7 May 2012. Lagiou P, Trichopoulou A, the DAFNE Contributors. The DAFNE initiative: the methodology for assessing dietary patterns across Europe using household budget survey data. Public Health Nutr 2001;4:1135–42. National and Kapodistrian University of Athens, School of Medicine, Department of Hygiene, Epidemiology and Medical Statistics. DAFNE-DAta Food NEtworking. Available at: http://www.nut.uoa.gr/dafneENG.html. Accessed 7 May 2012. DAFNE-ANEMOS. Software package for harmonizing national household budget data from the 28 DAFNE-ANEMOS countries. Available at: http://www. hhf-greece.gr/images/dafne_anemos_soft_leaflet.pdf.

Accessed 29 May 2012. 20. Weisell R, Dop MC. The Adult Male Equivalent concept and its application to Household Consumption and Expenditures Surveys (HCES). Food Nutr Bull 2012; 33:S157–62. 21. Smith LC, Alderman H, Aduayom D. Food insecurity in Sub-Saharan Africa: New estimates from household expenditure surveys. IFPRI Research Report 146. Washington, DC: International Food Policy Research Institute, 2006. 22. Smith LC, Subandoro A. Measuring food security using Household Expenditure Surveys. Food Security in Practice Series. Washington, DC: International Food Policy Research Institute, 2007. Available at: http://www.ifpri. org/sites/default/files/publications/sp3.pdf. Accessed 7 May 2012. 23. Sibrián R, ed. Deriving food security information from National Household Budget Surveys: Experiences, achievements, challenges. Rome: Statistics Division, Food and Agriculture Organization, 2008. Available at: http://foodsecuritylink.net/index.php?option=com_rem ository&Itemid=27&func=fileinfo&id=47. Accessed 7 May 2012. 24. Sibrián R, Seevalingum R, Mernies J. Measuring hunger at sub-national levels from household surveys using the FAO approach. Statistics Division Working Paper Series, ESS 5. Rome: Food and Agriculture Organization, 2008. 25. Tanzania National Bureau of Statistics, Ministry of Finance and Economic Affairs, European Community, Food and Agriculture Organization. Trends in food insecurity in mainland Tanzania. Food Security and Nutrition Analysis of Tanzania Household Budget Surveys 2000/1 and 2007. Dar es Salaam: Government of the United Republic of Tanzania, 2010. 26. Dop MC, Pereira C, Mistura L, Martinez C, Cardoso E. Using Household Consumption and Expenditures Survey (HCES) data to assess dietary intake in relation to the nutrition transition: a case study from Cape Verde. Food Nutr Bull 2012;33:S221–7. 27. Lambe J, Kearney J, Becker W, Hulshof K, Dunne A, Gibney MJ. Predicting percentage of individuals consuming foods from percentage of households purchasing foods to improve the use of household budgets in estimating food chemical intakes. Public Health Nutr 1998;1:239–47. 28. Friel S, Nelson M, McCormack K, Kelleher C, Thriskos P. Methodological issues using household survey expenditure data for individual food availability estimation: Irish experience in the DAFNE pan-European project. Public Health Nutr 2001;4:1143–7. 29. Naska A, Vasdekis VGS, Trichopoulou A. A preliminary assessment of the use of household budget survey data for the prediction of individual food consumption. Public Health Nutr 2001;4:1158–65. 30. Rambeloson Jariseta Z, Dary O, Fiedler JL, Franklin N. Comparison of estimates of the nutrient density of the diet of women and children in Uganda by Household Consumption and Expenditures Surveys (HCES) and 24-hour recall. Food Nutr Bull 2012;33:S199–207. 31. Naska A, Oikonomou E, Trichopoulous A, Wagner K, Gedrich K. Estimations of daily energy and nutrient availability based on nationally representative household budget survey data. The Data Food Networking


S251


(DAFNE) project. Public Health Nutr 2007;10:1422–9. 32. Nelson M, Dyson P, Paul AA. Family food purchases and home food consumption: comparison of nutrient contents. Br J Nutr 1985;54:373–87. 33. Sekula W, Nelson M, Figurska K, Oltarzewski M, Weisell R, Szponar L. Comparison between household budget survey and 24-hour recall data in a nationally representative sample of Polish households. Public Health Nutr 2005;8:430–9. 34. Beegle K, DeWeerdt J, Friedman J, Gibson J. Methods of household consumption measurement through surveys: experimental results from Tanzania. World Bank Policy Research Working Paper 5501. Washington, DC: World Bank, 2010. 35. Ravallion M. Global poverty measurement: Current practice and future challenges. PowerPoint presentation. National Academy of Sciences. A sustainability challenge: Food security for all. Workshop 2: measuring food insecurity and assessing the sustainability of global food systems. February 16, 2011. Venable LLP Conference Center, Washington, DC. Available at: http://sites. nationalacademies.org/PGA/sustainability/foodsecurity/PGA_060826. Accessed 7 May 2012. 36. Fiedler JL, Lividini K, Bermudez OI, Smitz M-F. Household Consumption and Expenditures Surveys (HCES): a primer for food and nutrition analysts in low- and middle-income countries. Food Nutr Bull 2012;33:S170–84. 37. International Labor Organization. Household income and expenditure statistics. Report II of the Seventeenth International Conference of Labor Statisticians. Geneva: International Labor Organization, 2003. 38. Grosh M, Glewwe P, eds. Designing household survey questionnaires for developing countries: lessons from

39. 40.

41.

42.

43.

44.

15 years of the Living Standards Measurement Study. Volumes 1, 2, and 3. Washington, DC: World Bank, 2000. United Nations. Handbook of household surveys. New York: United Nations, 1984. Coates J, Colaiezzi B, Fiedler J, Wirth J, Lividini K, Rogers B. Food consumption measurement methods for decision-making in food fortification and other nutrition programs: a Joint GAIN/HarvestPlus/Tufts Working Paper. Geneva: Global Alliance for Improved Nutrition, 2012. Martirosova D. Food data collected using acquisition and consumption approaches with daily diaries in Armenia’s ILCS 2004. In: Sibrian R, ed. Deriving food security information from national household budget surveys: experiences, achievements, challenges. Rome: Food and Agriculture Organization, 2008:59–68. Pereira C, Troubat N, Sibrian R. Food data collected using acquisition and intake approaches in Cape Verde’s IDRF 2001/2002. In: Sibrian R, ed. Deriving food security information from national household budget surveys: experiences, achievements, challenges. Rome: Food and Agriculture Organization, 2008:81–90. Kaara J, Ramasawmy S. Food data collected using acquisition and consumption approaches with a seven-day recall method in Kenya’s KIHBS 2005/2006. In: Sibrian R, ed. Deriving food security information from national household budget surveys: experiences, achievements, challenges. Rome: Food and Agriculture Organization, 2008: 69–80. Ruel MT, Deitchler M, Arimond M. Developing simple measures of women’s diet quality in developing countries: overview. J Nutr 2010;140:2048–50.
