LSMS GUIDEBOOK July 2017 The Use of Non-Standard Units for the Collection of Food Quantity A Guidebook for Improving the Measurement of Food Consumption and Agricultural Production in Living Standards Surveys Gbemisola Oseni, Josefine Durazo, and Kevin McGee The Use of Non-Standard Units for the Collection of Food Quantity A Guidebook for Improving the Measurement of Food Consumption and Agricultural Production in Living Standards Surveys Gbemisola Oseni, Josefine Durazo, and Kevin McGee World Bank ABOUT LSMS The Living Standards Measurement Study (LSMS), a survey program housed within the World Bank’s Development Data Group, provides technical assistance to national statistical offices in the design and implementation of multi-topic household surveys. Since its inception in the early 1980s, the LSMS program has worked with dozens of statistical offices around the world, generating high-quality data, developing innovative technologies and improved survey methodologies, and building technical capacity. The LSMS team also provides technical support across the World Bank in the design and implementation of household surveys and in the measurement and monitoring of poverty. ABOUT THIS SERIES The LSMS Guidebook series offers information on best practices related to survey design and implementation. While the Guidebooks differ in scope, length, and style, they share a common objective: to provide statistical agencies, researchers, and practitioners with rigorous yet practical guidance on a range of issues related to designing and fielding high-quality household surveys. The Series aims to achieve this goal by drawing on the experience accumulated from decades of LSMS survey implementation, the expertise of LSMS staff and other surveys experts, and new research using LSMS data and meth- odological validation studies. Copyright © 2017 The World Bank. Rights and Permissions This work is available under the Creative Commons Attribution 3.0 IGO license (CC BY 3.0 IGO) http://creativecommons.org/licenses/by/3.0/igo. Under the Creative Commons Attribution license, you are free to copy, distribute, transmit, and adapt this work, including for commercial purposes, under the following condition: Attribution—Please cite the work as follows: Oseni, G., Durazo, J., & McGee, K. 2017. The Use of Non-Standard Units for the Collection of Food Quantity: A Guidebook for Improving the Measurement of Food Consumption and Agricultural Production in Living Standards Surveys. Washington DC: World Bank. Disclaimer The findings, interpretations, and conclusions expressed in this Guidebook are entirely those of the authors. They do not necessarily represent the views of the International Bank for Reconstruction and Development/World Bank and its affiliated organizations, or those of the Executive Directors of the World Bank or the governments they represent. Living Standards Measurement Study (LSMS) World Bank Development Data Group (DECDG) lsms@worldbank.org www.worldbank.org/lsms data.worldbank.org Cover images: J. Durazo/World Bank; Ra Puji Wulandari Tungga Dewi/World Bank; Kevin McGee/World Bank Cover design and layout: Deirdre Launt TABLE OF CONTENTS ACKNOWLEDGMENTS..................................................................................................................................................................................... v EXECUTIVE SUMMARY..................................................................................................................................................................................... vi 1. INTRODUCTION............................................................................................................................................................................................ 1 1.1 Standard vs. non-standard units................................................................................................................................................................................................1 1.2 The market survey..........................................................................................................................................................................................................................3 1.3 The main survey...............................................................................................................................................................................................................................4 1.4 This Guidebook................................................................................................................................................................................................................................4 2. METHODOLOGIES FOR REPORTING CONSUMPTION AND PRODUCTION QUANTITIES........................................................5 2.1 Collecting data on food consumption.....................................................................................................................................................................................5 2.2 Collecting data on agricultural production............................................................................................................................................................................6 2.3 Impor tance of non-standard units in household surveys................................................................................................................................................6 2.4 Non-standard units in household surveys.............................................................................................................................................................................7 Benefits of allowing repor ting in non-standard units.......................................................................................................................................................7 Challenges of allowing repor ting in non-standard units.................................................................................................................................................8 How common are non-standard units in surveys?............................................................................................................................................................9 3. GUIDELINES AND PROCEDURES FOR CAPTURING AND USING NON-STANDARD UNITS....................................................10 3.1 Market survey planning and preparation..........................................................................................................................................................................10 Timing of market survey............................................................................................................................................................................................................10 Selection of markets to visit....................................................................................................................................................................................................11 Preparation of survey materials.............................................................................................................................................................................................11 3.2 Constructing the list of non-standard units......................................................................................................................................................................13 Market survey before/independent of main survey......................................................................................................................................................13 Market survey after main survey...........................................................................................................................................................................................14 3.3 Collecting weights for conversion factors.........................................................................................................................................................................14 3.4 Collecting reference photos....................................................................................................................................................................................................16 Which item-units require photos?........................................................................................................................................................................................17 Guidelines for reference photos............................................................................................................................................................................................18 Creating and using the photo reference album..............................................................................................................................................................23 3.5 How to use the non-standard units libraries...................................................................................................................................................................23 4. BENEFITS OF USING COMPUTER ASSISTED PERSONAL INTERVIEWING (CAPI)........................................................................25 4.1 CAPI for market surveys..........................................................................................................................................................................................................25 4.2 CAPI for food consumption and agricultural production survey............................................................................................................................26 5. CONCLUSION...............................................................................................................................................................................................27 REFERENCES......................................................................................................................................................................................................28 ANNEX 1: SURVEY INSTRUMENTS ...............................................................................................................................................................29 NSU market survey: questionnaire (Nigeria) NSU market survey: manual (Nigeria) Household survey: reference photo album (Ethiopia) Household survey: consumption module (with NSUs) Household survey: training manual (excerpt) * Additional examples available online .............................. 61 ANNEX 11: LIBRARY OF NONSTANDARD UNIT CONVERSION FACTORS AND REFERENCE PHOTOS...... Ethiopia: documentation and reference photographs Malawi: documentation and reference photographs Nigeria: documentation and reference photographs Uganda: documentation * All documents available online at www.worldbank.org/lsms iv ACKNOWLEDGMENTS This Guidebook was made possible by generous funding from UK Aid, through the grant “Improving Productivity, Gender and Innovation Data in Low Income Countries.” The authors would like to thank Alberto Zezza, Olivier Dupriez, Kristen Hime- lein, Heather Moylan, Talip Kilic, and Diane Steele for their inputs during the preparation and review of this Guidebook. The supporting documentation in the Annexes was prepared as part of data-collection activities conducted in partnership with the Central Statistical Agency of Ethiopia, the National Bureau of Statistics in Nigeria, the National Statistics Office of Malawi, and the Uganda Bureau of Statistics. Our greatest appreciation goes to these partners for their dedication to the projects. v EXECUTIVE SUMMARY This Guidebook is a reference for survey practitioners, providing advice on how to incorporate non-standard units (NSUs) of measurement into household surveys for the collection of food consumption and production quantities. Food consump- tion and agricultural production are two critical components for monitoring poverty and household well-being in low- and middle-income countries. Accurate measurement of both provides better contextual understanding and contributes to more effective policy design. At present, there is no standard methodology for collecting food quantities. In many household surveys, respondents are forced to estimate quantities in standard or metric units, typically kilograms or liters. This method requires respondents to convert from whatever unit they actually consumed (e.g., a bowl of rice) to a standard unit. This conversion process is often an unfamiliar or difficult task for respondents and can introduce measurement error. We argue that allowing respondents to report quantities directly in NSUs places less of a burden on respondents and will ultimately improve the accuracy of the information they provide. Despite these benefits, there are some challenges with this approach. First, these NSU quantities must still be converted into standard units for aggregation and analysis. Often, conversion factors are not readily available and must be created, a process that involves its own data-collection effort. A second challenge is that NSUs are by their nature not necessarily stan- dardized across respondents. One person’s “bunch” of bananas could be half the size of another person’s “bunch.” Showing reference photos of “bunches” to respondents can ensure that the unit “bunch” is further standardized when reported. This requires that a photo reference album is also prepared. This Guidebook explains how to properly incorporate NSUs into data-collection activities—from establishing the list of allowable NSUs to incorporating all components into household surveys. A NSU-focused market survey is a critical part of preparing the conversion factors required for effectively using NSU data in analysis work. As such, the bulk of this Guidebook focuses on implementing the market survey and on calcu- lating conversion factors to ensure the highest-quality data when using NSUs. Practical guidance on non-standard units, conversion factors, and reference photos Although existing data must first be taken into consideration, establishing a baseline of properly documented NSUs will most often require conducting a market survey, whereby survey teams seek out item-unit combinations in the market to weigh and photograph. Both market outputs then become inputs to the main household survey: the reference photos are shown to respondents during interviews and the weights are used to create conversion factors that are applied to the reported NSU quantities, facilitating their use in data quality assurance and data analysis. Collectively, these components comprise what is referred to herein as the NSU library. There are several important steps to follow in preparing the library: 1) Preparation—Plan the timing (relative to the main survey) and the locations of the market survey, prepare the necessary market-survey materials (instruments and manuals), and construct a list of item-unit combinations that will be allowed in the main survey; 2) Market survey implementation— Collect weights and reference photos, taking into account any sub-national variation; and 3) Data documentation for the main survey—Using the market data, create conversion factors for the NSUs and draft clear user protocols for enumera- tors (in terms of reference photos) and data users (in terms of conversion factors). vi Procedures for properly implementing these steps are summarized here, and are then covered in detail throughout the Guidebook. a. In terms of planning and preparation, a list of valid item-unit combinations should first be constructed by reviewing, updat- ing, and supplementing as necessary any existing sources that contain information on common NSUs. Next, when planning the market survey, it is especially important to consider its timing relative to the main survey where consumption and agricultural production data will be collected. Ideally, the market survey should be conducted prior to the main survey in order to use the reference photos during the main survey. If necessary, a much smaller-scale market survey can be conducted after the main survey to collect missing weights for any unanticipated conversion factors. Finally, markets should be selected to ensure adequate coverage of NSUs in the relevant context. This is particularly important if NSUs differ across regions. b. Following these preparatory steps and the detailed market-survey implementation guidelines herein will ensure that as many item-unit combinations are collected as possible, the weights collected are comparable and accurate, and the reference photos clearly demonstrate the actual size of the NSUs. Annex I contains sample survey instru- ments. c. After the market survey, the information collected should be prepared for use with the main survey. A library of NSU materials should be compiled, starting with the calculation of conversion factors that can be applied to NSU consump- tion and production quantities collected during the main survey. These conversion factors are used to flag unreasonable quantities for further verification; when surveys are conducted using computer-assisted personal interviewing (CAPI), this can be done during the course of fieldwork. When the main survey is complete, the conversion factors can be used to calculate total consumption, analyze poverty, etc. The library should also include an album of refer- ence photos compiled from the photos collected in the market survey. This album should be used by the enumerators conducting the main survey to provide a reference size for NSUs. Finally, the library must include documentation of how the materials were prepared and how to properly use them during the main survey. We highly recommend that the library be made publicly available for use in other surveys in order to further standardize NSU reporting across data-collection efforts. The library can be continually updated as more information is collected. Annex II to this Guidebook contains libraries for Ethiopia, Nigeria, Malawi, and Uganda, and is available online. Although they are targeted for use with LSMS-ISA surveys, the libraries are intended to be used by any researchers conducting simi- lar survey activities in these countries. The libraries should be considered living documents, to be revised and updated with each new data-collection effort given that available foods and commonly used units and quantities may vary over time. Even so, making NSU libraries publicly available for more countries will make it easier to implement surveys that allow NSUs and will therefore result in improved data-collection for quantities of food consumption and agricultural production. vii 1. Introduction Measuring poverty often depends on measuring food—food that is both purchased and harvested from the field. In low- and middle-income countries especially, food consumption still constitutes the largest share of total household consumption. As such, constructing a food poverty line and using it to estimate the total poverty line is the preferred methodology for measuring the share of households that are poor, which in turn is one of the most common welfare-analysis indicators for developing economies. Another important element of welfare analysis is the productivity of income-generating activities. In many low- and middle-income countries, agriculture is a major source of livelihood, and measuring agricultural productivity requires adequately measuring the quantity of agricultural output. Data on food quantity is also important for the computation of unit values for food items and crops, which in turn is critically important for monitoring and analyzing prices. Despite the importance of this information, accurately measuring both the quantity of food consumed and the quantity of agricultural output can be very challenging. 1.1 STANDARD VS. NON-STANDARD of cognitive task. Recent studies show that asking respon- dents to combine memory recall with cognitive tasks, such as UNITS abstracting consumption to a “typical week or month,” leads One important aspect of collecting information on food con- to less accurate self-reporting (Beegle et al., 2010). sumption and agricultural production is the choice of units in which respondents can report quantities. Many surveys The forced conversion from non-standard to standard require quantities to be reported only in “standard” units units requires respondents to undergo the process depicted such as kilograms, pounds, liters, etc. In these cases, “local” in Figure 1. Respondents 1) must have a good understanding or “non-standard” units are disallowed. Forcing respondents of what a standard unit of a food item is (e.g., how much is a to report only in standard units simplifies the use of the data kilogram of rice), 2) must estimate how many standard units (since aggregation/analysis of food-item consumption often correspond to the NSU they know (e.g., how many kilograms requires a common unit of measure) but it can impose a sig- fit into a cup of rice), and finally 3) using the conversion from nificant cognitive burden on the respondent, which in turn 2, must calculate the quantity consumed in standard units can reduce the accuracy of the resulting data. (e.g., 1 cup of rice is about 0.5 kg, I consumed 1.5 cups of rice, so I consumed about 0.75 kg of rice). All three stages place a Many respondents in low- and middle-income countries are cognitive burden on the respondent and can lead to sizable more comfortable reporting their food consumption and pro- measurement error. Allowing respondents to directly report duction using familiar “local” or “non-standard” units instead consumption in NSUs would ease the burden on the respon- of standard units. Forcing respondents to convert from these dents and will ultimately result in more accurate reporting familiar units into standard units during an interview is a type of their consumption. 1  1. Introduction 2 Figure 1 — Forcing NSU Conversion vs. Allowing NSUs FORCING STANDARD UNITS: More burden on the respondent, less consistency in conversion factors We consumed 1.5 bowls of rice. How much rice is a kg? I am not really sure. How many kilograms How many kgs of rice are in a bowl? I guess about 0.5 kg. of rice did you consume So then I guess we consumed 1.5 bowls of rice X 0.5 kg in a bowl=0.75 kg of rice! in the past 7 days? We consumed 0.75 kg. ALLOWING NSUs: Simpli es respondent’s role, conversion factors are consistent Was the bowl How much rice similar to this We consumed did you consume size, or to this 1.5 bowls of in the past 7 days? size? rice. Apply collected conversion factor of 0.689 kg of rice per bowl to get 1.0335 kg of rice consumed. This size. Source: World Bank, LSMS Team. 3  THE USE OF NON-STANDARD UNITS FOR THE COLLECTION OF FOOD QUANTITY While allowing NSU reporting will eliminate some burdens Figure 2 — Vaguely Defined NSUs for the respondent, it does not mean the issues of NSU con- version disappear. Instead, it falls to the survey or research team to acquire the necessary information to take NSU quantities from respondents and convert them to common standard units (i.e., undertake steps 1, 2, and 3 mentioned above). The most critical information required to make the conversions is a list of standard-unit conversion factors for each NSU as well as for each food item. These item-unit con- version factors can be applied to NSU quantities reported by How much? respondents to convert them into standard units (typically kilograms and liters). In an ideal world, a list of such conver- One small pile. sion factors would already exist for the relevant country or context. However, the current reality is that conversion fac- tors are not readily available in many low- and middle-income countries. When they are available, they are often limited in scope or poorly documented, making their applicability and reliability hard to determine. When reliable conversion fac- Source: World Bank, LSMS Team. tors are not available, it is up to the national statistics agency or research team to collect the weights and calculate the such as household and individual welfare as well as agricul- conversion factors required to convert non-standard units tural productivity. into clearly and widely measurable standard units such as kilograms or liters. 1.2 THE MARKET SURVEY Going about collecting the necessary information to prop - There are a few methods for calculating conversion factors. erly incorporate NSUs into a survey and make NSU quan- One such method suggested by Capéau (1995) and Capéau tities usable for analysis is not a trivial process. The varied and Dercon (2006) is to compare unit prices using econo - nature of NSUs introduces significant challenges to any sur- metric techniques to estimate conversion factors. While this vey or research team undertaking this task. For example, method is fairly simple to implement, it suffers from some similarly named NSUs can vary significantly within countries drawbacks. Primary among these is that unit prices can vary or subnational regions. Even within the same locality, NSUs because of factors unrelated to the actual mass or volume of often come in more than one size (e.g. small, medium, large). an item. For example, unit prices can vary because of quality The challenges are particularly significant for vaguely defined differences (Deaton 1997) or because of price discounts on NSUs such as pieces, heaps, bunches, etc. A “heap” of toma- larger units (Attanasio & Frayre, 2006).1 In addition, unlike toes can vary dramatically in size, making it difficult to con- conversion factors, prices can be subject to significant volatil- vert each respondent’s “heap” in a consistent and accurate ity due to market forces. These sources of variability in unit manner. While these challenges are significant, there are no prices unrelated to mass or volume can result in distorted comprehensive guidelines on how to properly collect this or imprecisely estimated conversion factors. information. The main alternative method is to conduct a market sur- This Guidebook is meant to fill this gap by highlighting vey where non-standard units are sought out and directly the necessary steps and best practices for collecting this weighed. This is a more intensive process than calculating information. Establishing a systematic, well-documented, conversion factors from unit prices, but will likely result in and more precise set of conversion factors for non-stan- more accurate conversion factors. When conducting a market dard units—and using it to both inform survey design and to convert reported measurements—will go a long way toward increasing the accuracy of crop-output estimates 1  Under this methodology, a numeraire unit price (usually kilograms or liters) is used to compare with other units. For larger units, there may be discount in the and household consumption. This in turn will allow for more price per kilogram and thus applying the numeraire unit price would underestimate informed policymaking on important development issues the conversion factor for larger units. The reverse is also true for smaller units. 1. Introduction 4 survey, there are certain protocols that must be followed to benefits that are derived from utilizing CAPI survey meth- ensure the collected weights are accurate and usable for cre- ods to both collect and use information contained in a NSU ating conversion factors. For example, conversion factors for library. Section 5 offers concluding remarks. Annex I pro - vaguely defined units (especially non-container units such as vides a set of sample instruments for collecting and then pieces, heaps, bunches, etc.) are most reliable when accom- using NSUs. Annex II is available online and provides librar- panied by reference photos. These photos can be shown to ies of NSU conversion factors from four countries (Ethiopia, the respondents to provide standardized reference sizes for Malawi, Nigeria, and Uganda). a “small heap” of onions, for example. Without the photos, the “small heap” reported by the respondent could be con- siderably different from the “small heap” used to establish the conversion factors (see Figure 2). These reference pho- tos must therefore be taken and collected along with the weights. THE LIVING STANDARDS M E A S U R E M E N T S T U DY– 1.3 THE MAIN SURVEY INTEGRATED SURVEYS ON Once all the requisite information is collected for proper AGRICULTURE (LSMS–ISA) is a implementation of NSUs in a survey, the interview process household survey project to foster becomes much less taxing on the respondents, without innovation and efficiency in statisti- additional burden on the enumerators. The bottom panel cal research on the links between of Figure 1 depicts the revised process. The respondent is agriculture and poverty reduction only required to think about consumption or production in in the region. Recognizing that the unit with which she is most familiar. The enumerator existing agricultural data in Sub-Sa- simply confirms this unit using the reference photo and then haran Africa suffers from inconsis- records the amount in NSUs. Afterward, conversion factors tent investment, institutional and are applied to the reported NSUs to arrive at the correct sectoral isolation, and methodolog- standard weight. ical weakness, the LSMS-ISA proj- ect collaborates with the national 1.4 THIS GUIDEBOOK statistics offices of its eight partner This Guidebook serves as a reference for preparing and countries to design and implement using non-standard units: establishing a list of valid NSUs, household surveys with a strong collecting standard weights and reference aides for NSUs focus on agriculture. In each part- (usually via a market survey), calculating conversion factors ner country, the LSMS-ISA sup- from these weights, and incorporating NSUs into household ports multiple rounds of a nationally and agriculture surveys. It also provides a library of local representative panel survey with a units, conversion factors, and photographic aids for selected multi-topic approach designed to countries. The Guidebook is structured as follows. Section improve the understanding of the 2 provides background on non-standards unit (NSUs), dis- links between agriculture, socio - cusses the importance of properly quantifying household economic status, and non-farm consumption and production, and offers various methods income activities. The frequency of of collecting this data. The section also details the bene - data collection is determined on a fits and challenges of NSUs and documents their use in the country-by-country basis, depend- Living Standards Measurement Study-Integrated Surveys on ing on data demand and the avail- Agriculture (LSMS-ISA) implemented by the LSMS team of ability of complementary funding. the World Bank. Section 3 outlines the components of a high-quality NSU library as well as the necessary procedures for conducting a market survey to collect the components of the library. Section 4 discusses some of the important 2. Methodologies for Reporting Consumption and Production Quantities Food consumption and agricultural production are two of the most important measurements in living standards surveys as well as in many other household surveys in low- and middle-income countries. Food consumption is the primary component for many measures of poverty, nutrition, and food security. Information on agricultural production provides important insights on agricultural performance as well as farm household income and own-food consumption. The critical importance of consumption and agricultural production quantities has led to the development of several different measurement methodologies. Each method has its merits and drawbacks and is not necessarily applicable in all situations. However, we argue that the use of non-standard units to measure quantity is widely applicable and strikes a fair compromise in terms of cost. 2.1 COLLECTING DATA ON FOOD 1. Metric (i.e., standard) units: Respondents report quantities in metric units such as kilograms or CONSUMPTION grams. While this method is low cost and relatively The various methods of collecting food-consumption data easy to record, it can result in inaccurate esti- are the subject of a large and well-established body of liter- mates in some circumstances (see below). ature. In general, consumption information is typically col- 2. Monetary value: Respondents estimate the mone- lected via respondent recall interviews or a consumption tary value of the quantity consumed or produced. diary. For recall, respondents are asked to estimate their This method requires additional collection of met- consumption of an item over a specified period, typically ric prices in order to estimate quantities, which seven days. Under the diary method, respondents are asked can also be subject to significant errors. to keep a daily diary of their consumption. Both methods require respondents to report quantities of food consumed. 3. Local (i.e., non-standard) units: Any unit of mea- Though there is a broad range of survey-design issues (see surement that does not have an objective, univer- Beegle et al., 2010 for a review of these issues), the discussion sal metric or standard weight. This includes items here will focus on collecting quantities, as this is the specific such as “pail,” “basket,” or “pieces;” the latter is design focus of this Guidebook. discussed in #4 below. 4. As in #3, respondents report quantities in terms There are many methods used to collect information on of non-standard units with which they may be the quantity of food consumed in a household survey. Smith more familiar. These methods ease the burden on & Subandoro (2007) discuss seven primary methods, sum- the respondent (in terms of memory recall and marized as follows: conversion calculations), but can increase the cost of survey implementation. 5  2. Methodologies for Reporting Consumption and Production Quantities  6 5. Volumetric equivalents: Respondents demonstrate costly to implement on a large scale. While there are numer- how much space the food they consumed would ous issues associated with all three of these methods, they take up. Conversion factors would need to be ap- are largely beyond the scope of this Guidebook (see Fermont plied to convert to metric units. & Benson, 2011 and Sud et al., 2016 for a review). The focus 6. Linear dimensions: Respondents provide here is on the collection of harvest quantities. linear measurements (length and width or In principle, many of the consumption-quantity collection circumference) of the amount of food consumed. methods outlined by Smith & Subandoro (2007) are also As Smith & Subandoro (2007) point out, this applicable to the collection of agricultural-production quan- method likely takes more time to complete as tities. However, there is one additional issue that is specific it requires physical measurement rather than a to the measurement of crop harvests. The condition of the simple vocal response. crop—threshed, shelled, fresh, dried, etc.—can have a large 7. Food models: Respondents choose a two- or impact on reported harvest quantities (Fermont & Benson, three-dimensional depiction of a food item that 2011; Diskin, 1999; Murphy et al., 1991). The weight difference best corresponds to their consumption amount. is either due to discarding a portion of the crop via thresh- This method can provide very accurate estimates, ing, shelling, or peeling, or is the result of a change in mois- but it can be costly to prepare the models and cal- ture content through drying. These processes are particularly culate their weights. important for cereals and legumes, which are quite often processed before being used or sold. It is therefore import- While one method may be optimal for certain items, it may ant to ensure that when a harvested quantity is reported, not be feasible or appropriate for others. Smith & Subandoro the condition of the crop to which the quantity refers is also (2007) advocate using a combination of these methods. This specified. When quantities are reported for various condi- Guidebook and the accompanying library will focus on four tions, additional condition-specific conversion factors can be of these methods, which are complementary and compre - applied to render the quantities comparable. hensive: metric units, local units, including a count of pieces, with some two-dimensional depictions (i.e., reference pho- 2.3 IMPORTANCE OF NON- tos). Joint use of these four methods will minimize the burden on the respondent as well as on the enumerator, although STANDARD UNITS IN HOUSEHOLD it may require additional costs beyond the main survey visit. SURVEYS The methodological issue that is the focus of this Guidebook 2.2 COLLECTING DATA ON is the use of non-standard units in the collection of con- AGRICULTURAL PRODUCTION sumption and production quantities. But what exactly are “non-standard” versus “standard/metric” units? Both stan- There are three prevailing methods for measuring agricul- dard and non-standard units are commonly used in markets tural production: farmer recall, whole plot harvest, and crop or by households in many countries. Standard units are uni- cutting (Sud et al., 2016). Under the recall method (as with versally constant, referring to a clearly defined weight and/ recall for food consumption), farmers estimate how much of or volume. A kilogram in Uganda is the same as a kilogram in a particular crop they have harvested since a certain date. France. Likewise, a kilogram of maize is the same weight as Both whole plot harvest and crop cutting are much more a kilogram of wheat. For the most part, “standard” encom- labor-intensive processes that attempt to eliminate the sub- passes metric units, imperial measurements, and other inter- jective bias or error inherent in farmer estimates. Under the nationally standardized units that are easily converted into crop-cutting method, a portion of a farmer’s crop is cut and metric units. For example, the conversion between kilograms measured by enumerators at the time of harvest. However, and pounds is constant regardless of region or item. there are some potential sources of bias that arise in the crop-cutting method (Fermont & Benson, 2011). Whole plot In contrast to standard units, non-standard units (NSUs) harvest is similar to the crop cutting method, but the output often vary greatly from item to item, region to region, and of the entire plot is cut and measured. This is considered even village to village. Table 1 presents examples of some the most accurate yield measurement, but is also extremely common standard units and NSUs. 7  THE USE OF NON-STANDARD UNITS FOR THE COLLECTION OF FOOD QUANTITY Table 1 — Examples of Standard and in the second wave of the General Household Survey Panel Non-Standard Units (GHS-Panel) for each of the six geopolitical zones. Only one Standard Non-standard unit (milk cup) is common to all six zones. Most units are only found in two to three zones and are rarely or never observed Common Local (Uganda) in others. These complexities associated with NSUs create Kilograms Sack Jerrican some challenges relative to standard units. However, there Grams Bunch Kimbo/Blueband tin are additional factors that contribute to the relative merits Liters Heap Nido tin of the methods, detailed further below. Centiliters Piece/number Cup/mug Pounds Bucket Nice cup 2.4 NON-STANDARD UNITS IN   Crate HOUSEHOLD SURVEYS   Plastic basin Source: World Bank, LSMS Team. BENEFITS OF ALLOWING REPORTING IN NON-STANDARD UNITS Some NSUs are common across many locations. For example, throughout the world, bananas and other items There are trade-offs involved in deciding whether to allow are often measured in bunches. While this NSU is common, respondents to report quantities in NSUs or to restrict it is not standardized. The number and size of the bananas respondents to reporting in only standard units. Although that are in a bunch are not standard; one bunch of bananas NSUs are subject to significant variation, there are tangible could be three times the size of another bunch. Likewise, a benefits to allowing respondents to report in NSUs. The bunch of bananas is not equivalent to a bunch of herbs. The most important and overriding benefit is that respondents same is true for pieces, heaps, and other units. In addition will likely be better able to estimate quantities using NSUs. to these common NSUs, there are also NSUs that are spe - In rural areas especially, standard units may not be commonly cific to a country or region. Table 1 includes several NSUs used in markets and respondents may not regularly use stan- commonly used in Uganda (far right column). Many of these dard units in their daily activities. Even though respondents units are locally familiar containers of a standard volume; may know exactly what a kilogram of sugar looks and feels however, the weight of the contents will vary depending on like (a very common sales unit for sugar), they may not know the item. For example, a Nido tin of rice does not weigh the this for cassava, maize, or other items that are not typically same as a Nido tin of groundnuts. The use of local units can traded in kilograms at the household level. Likewise, many vary significantly within a country. As an example, Table 2 items are not generally consumed in standard units. Often, presents the distribution of local units in Nigeria observed fruit is sold by the piece instead of by weight; herbs are Table 2 — Regional Variation of NSUs in Nigeria % of all NSU Observations in Zone   North Central North East North West South East South South South West Mudu 56.7 62.4 17.4 0.0 0.7 0.0 Olodo 0.0 0.0 0.0 0.0 14.0 0.0 Congo 7.5 0.0 0.0 0.0 0.0 47.9 Paint rubber 3.9 0.2 0.6 12.1 0.0 2.7 Derica 0.4 1.3 0.2 2.5 16.1 13.8 Milk cup 27.3 8.9 21.4 21.6 42.6 28.8 Cigarette cup 0.5 0.0 0.1 60.7 22.3 0.0 Tiya 0.0 26.7 58.5 0.0 0.0 0.0 Kobiowu 2.8 0.0 0.0 0.0 0.0 0.0 Note: Shaded cells = Units rarely/never observed in that zone. Source: World Bank, LSMS Team. 2. Methodologies for Reporting Consumption and Production Quantities  8 sold by the bunch, regardless of weight variation; and home- One of the first challenges that survey designers face is grown fruits or vegetables are harvested and eaten without identifying the units to be included in the survey. When being weighed. When respondents are more familiar with respondents are limited to reporting in standard units, com- NSUs for specific items, it may be too burdensome to expect piling the code list is straightforward. However, compiling a them to know that item in terms of standard units. list with non-standard units requires additional information about the common NSUs in the relevant country and/or Forcing respondents to report quantities in standard units regions. When such information is limited or not available, often combines two self-reporting styles, each with its own it will need to be collected via a market survey. potential for error: memory recall and cognitive reasoning. Household consumption modules typically ask respondents In addition to identifying the NSUs to include in the main to recall a litany of food items eaten by numerous household survey, survey designers also need to ensure the clarity of the members over a given period. Farming households are asked unit definitions. Some of the most common units that fall into to recall and report on a variety of different crops harvested this category are pieces, bunches, or heaps. For example, a over a given period. The latter case is further complicated by piece of sweet potato could weigh 0.5 kilograms or 1.5 kilo - the fact that key crops throughout the region (e.g., cassava, grams. Figure 3 below illustrates this problem: the pictured maize, and plantains) are typically harvested in small quanti- containers vary significantly in size but are all called dengu in ties on a continual basis. Malawi. In order to obtain the most accurate estimates for these units, respondents should be provided with a refer- Both memory recall and ad-hoc unit conversions also ence frame for the quantity. One way to do this is to pro- require mathematical calculations that, while not necessar- vide respondents with reference photos for these items. This ily complicated, are prone to errors when done in the field resource can also be produced as part of the market survey and on the fly—even more so when considering respondent and requires additional enumerator training (detailed below). and enumerator fatigue. Combining memory recall with unit conversion increases the number of calculations required of the respondent for each value, which further increases the Figure 3 — Wide Variety of Dengus of potential for error (as shown in Figure 1). In general, allow- Tomatoes in Malawi ing respondents to report in the units they can most easily quantify simplifies memory recall and will yield estimates that are more accurate. CHALLENGES OF ALLOWING REPORTING IN NON-STANDARD UNITS Although there is a strong case for allowing respondents to report in NSUs, many surveys of consumption and agri- cultural production still restrict respondents to report- ing amounts in standard units. This is primarily due to the additional cost and challenges associated with properly implementing and operationalizing NSUs in a survey. The Source: World Bank, LSMS Team. complexity of NSUs as well as the additional steps required for their use can increase the financial and temporal bur- den of conducting a survey. The challenges associated with The final and most significant challenge in using NSUs is using NSUs broadly fall into two categories: (1) those asso - that they must be accompanied by standard-unit conversion ciated with the preparation and implementation of the sur- factors. In their raw form, quantities in NSUs are not compa- vey with NSUs, and (2) ensuring NSU measurements can be rable across units. To directly compare and aggregate quanti- converted into comparable standard units. ties, the data user must convert all quantities into a common standard unit such as kilograms. Converting between 9  THE USE OF NON-STANDARD UNITS FOR THE COLLECTION OF FOOD QUANTITY standard units is relatively easy since the conversions are Figure 4 — Recommended Steps for Using NSUs constant and well known. However, for NSUs, the conver- sion is different for each unit and often for each item. Com- plicating matters further, the standard weight for the same Prepare to use NSUs item-unit combination can vary, even within a country. For 1. Establish valid list of NSUs: create new or update existing list example, one study in Nigeria found that an average bundle 2. Plan Market Survey of sorghum weighed between 26 and 49 kilograms depending on the area (Casley & Kumar, 1988). When this is the case, region-specific conversion factors should be acquired. HOW COMMON ARE NON- STANDARD UNITS IN SURVEYS? Conduct Market Survey In low- and middle-income countries, especially in Africa, Collect national/regional weights and NSUs are used quite regularly for the most important items. photos. Skip this step ONLY IF At markets in these countries, consumers encounter a wide existing data is available and is current. variety of NSUs for their purchases. In addition, when given a choice between reporting quantities in standard or non-stan- dard units, respondents often choose to report in NSUs. For example, in the second wave of the Ethiopia Socioeco- nomic Survey from 2013/2014 when NSUs were allowed, nearly 50 percent of farmers chose to report their harvests Create Tools for Main Survey in NSUs. In the Malawi National Panel Survey, respondents 1. Generate conversion factors chose NSUs about 73 percent of the time. This provides a (preload into CAPI if using) strong indication that many respondents are more comfort- 2. Compile photo reference guide able reporting quantities in NSUs. 3. Write user protocols While the challenges associated with including NSUs in a consumption or agricultural production survey can be sig- nificant, this Guidebook provides detailed instructions to help survey designers incorporate NSUs into their surveys. Conduct Main Survey See Annexed examples for including NSUs in main survey. Finalize Data Documentation 1. Clearly document all steps and NSU/CF data sources. 2. If CFs are missing, conduct a brief follow-up survey and update final NSU and CF. Source: World Bank, LSMS Team. 3. Guidelines and Procedures for Capturing and Using Non-Standard Units This section lays out the necessary steps and procedures required to collect the information necessary to implement and use NSUs in household surveys. Before NSUs can be used, a resource library for NSUs will need to be prepared. This library should include (1) a list of common/ allowable NSUs; (2) national or regional conversion factors for all item-unit combinations; (3) a photo reference album based on an index of NSUs; and (4) clear protocols for using conversion factors and reference photos in agricultural and household surveys. The best way to collect the information for the library is to conduct a market survey to capture reference photos and the item-unit weights used to calculate conversion factors. In countries where surveys already allow reporting in NSUs, existing data (once updated, if need be) can complement the library. Taken together, these components will help researchers to adopt the use of NSU reporting; for countries with libraries provided in this Guidebook, the cost/burden of adoption is significantly eased. These libraries should not be treated as fixed, but should instead be continually updated with new NSUs and conversion factors. There are several important steps to follow when collect- 3.1 MARKET SURVEY PLANNING AND ing the components for an NSU conversion-factor library PREPARATION (see Figure 4): 1) Preparation—Plan the timing (relative to the main survey) and the locations of the market survey, prepare the necessary market-survey materials (instru- TIMING OF MARKET SURVEY ments and manuals), and construct a list of item-unit com- The data-collection schedule for the market survey should binations that will be allowed in the main survey; 2) Market take into consideration the seasonal availability of items and survey implementation—Collect weights and reference existing data-collection schedules in each country. In some photos, taking into account any sub-national variation; cases, market-survey data collection should be planned for and 3) Data documentation for the main survey—Using two separate periods to ensure more complete coverage the market data, create conversion factors for the NSUs of seasonally available items. In general, the greatest variety and draft clear user protocols for enumerators (in terms of items and the greatest variety of crop conditions will be of reference photos) and data users (in terms of conver- available during the harvest season, though some items may sion factors). Each of these steps is covered in detail here. only be available during the lean season or after a secondary harvest season. 10 11  THE USE OF NON-STANDARD UNITS FOR THE COLLECTION OF FOOD QUANTITY The timing of the market survey relative to the main sur- and their respective weights; 2) the relative timing of market vey is important. Performing the market survey before the surveys (as explained above); and 3) the types of markets fre- main survey (ex-ante or independently) has several advantages. quented by sample households. First, reference photos can be taken and then used during the Markets should be selected to ensure adequate coverage of main survey, and conversion factors can be used to validate regional units and items. For market surveys conducted after reported quantities during fieldwork, both helping to ensure the main survey, coverage can be assessed using the item- more accurate NSUs estimates. Second, conducting the mar- units observed in each stratum. The strata where the widest ket survey ex-ante allows for the identification of additional diversity of regional units is observed are prime candidates NSUs that may be missing from any existing list of answer for the market survey. If the item-units were reported in the options. Identified in advance, these units can be incorporated stratum, it is likely that measurements for that item-unit can into the unit list for the main survey. be obtained from a market in that area. If the market survey However, there are also some drawbacks to conducting is conducted independent of or prior to the main survey, ade- the market survey before the main survey. The primary one quate coverage must be assessed using external information is that new units not included in the market survey could be on regional variations in units as well as information from the reported in the main survey. When the market survey is per- pilot survey (if conducted). formed after the main survey (ex-post), the unit list for the In many countries, households may patronize various types market survey can be constructed to include all item-unit of markets, including local outdoor markets and small shops, (and item-unit-conditions) observed in the main survey, thus supermarkets, wet markets, and wholesale markets. Markets limiting any conversion-factor gaps. selected for the survey should cover the full range of markets Given these considerations, the ideal plan is to conduct two commonly used by households or farmers in the sample area. market surveys—one before and one after the main survey. In general, the number and dispersion of markets selected Both market surveys need not be equally rigorous; one will for the survey is highly dependent upon the context. For a likely be more comprehensive than the other. For example, nationwide survey in a large and diverse country like Ethio- the ex-ante survey could be limited, aiming to collect reference pia, it would likely be necessary to visit many markets across photos and weights (for conversion-factor calculations) for the country to ensure adequate coverage of NSUs, espe- the most common NSUs, while the ex-post survey could com- cially if the market survey is conducted during or after the prehensively collect weights for all additional NSUs reported main survey. However, for a market survey limited to a single during the main survey. Or the ex-ante survey could be inten- region/state or community, visiting only a few markets may sive, aiming to collect as many conversion-factor weights as be sufficient. possible—especially when conducted independently—while the ex-post survey could be limited to collecting only those PREPARATION OF SURVEY MATERIALS unanticipated item-unit combinations. In general, the intensive version of the market survey should coincide with the season Once markets are selected, the survey materials (survey when the most items will be available in the markets. instruments and supporting manuals) can be prepared. While these materials should be designed according to the local con- Many household surveys already conduct market surveys text, Annex I includes examples from NSU-focused market as part of their fieldwork to collect current pricing infor- surveys that were conducted to create the LSMS conver- mation on commonly consumed items. When surveys allow sion-factor libraries for Ethiopia, Malawi, and Nigeria. Figure 5 NSU reporting, the market surveys could also collect actual depicts a snapshot of the market survey questionnaire for weights of allowable NSU combinations that can be used to Ethiopia. calculate standard-unit conversion factors. However, in such cases where the market survey is conducted in parallel with Though these survey instruments are specific to these the main survey, reference photos would most likely not be countries, they also serve as examples of how to prepare available for use during interviews. instruments for any country/project. Each survey can collect the following key types of data: SELECTION OF MARKETS TO VISIT • Market identification details: Name, location, GPS infor- Three main factors will influence the selection of markets mation, type of market, etc. for the survey: 1) the degree of regional diversity of units 3. Guidelines and Procedures for Capturing and Using Non-Standard Units  12 Figure 5 — Excerpt from a Market Survey Questionnaire MODULE B: ITEM-UNIT MEASUREMENT - NONCONTAINERS 1 2 3 4 5 6 7 8 9 ITEM ITEM UNIT SIZE Was item Why was item not measured? Item Sample #1 Item Sample #2 Which type of scale NAME CODE NAME measured? Weight Price Weight Price was used? NOT FOUND IN MARKET AT THIS TIME........................................................1 (KGs) (Birr) (KGs) (Birr) CROP NOT COMMONLY FOUND YES...1 IN THIS MARKET.................................2 PERSONAL DIGITAL SCALE..................................1 (7) UNIT NOT COMMONLY FOUND IN THIS MARKET.................................3 MARKET SCALE WITH NO..2 SIZE NOT COMMONLY FOUND GOVERNMENT CERTIFI- IN THIS MARKET.................................4 CATION.............................2 OTHER, SPECIFY..................................5 MARKET SCALE WITH- OUT GOVERNMENT ALL RESPONSES ( NEXT ITEM) CERTIFICATION..............3 CEREALS AND GRAINS BARLEY 1 ESIR Small 1 ESIR Medium 1 ESIR Large 1 CHINET Small 1 CHINET Medium 1 CHINET Large 1 SHEKIM Small 1 SHEKIM Medium 1 SHEKIM Large MAIZE 2 PIECES Small 2 PIECES Medium 2 PIECES Large 2 ESIR Small 2 ESIR Medium 2 ESIR Large 2 CHINET Small 2 CHINET Medium 2 CHINET Large MILLET 3 ESIR Small 3 ESIR Medium Source: World Bank, LSMS Team. • Survey management information: Date, time, duration of and thus more readily able to report, in these quantities. surveys; codes for enumerator, supervisors, and (when The Nigeria survey is split into two sections, allowing enu- applicable) data-entry staff. merators to more easily divide and share the data-collection • Data on pre-identified NSUs: Weights, prices, and basic work during each market visit. Detailed instructions on how metadata for common item-unit combinations that have to collect the data and complete the questionnaires are in the been previously identified. training manuals (also included in Annex 1). • Data on unexpected NSUs: Teams can collect the same 3.2 CONSTRUCTING THE LIST OF type of data listed above for item-unit combinations that are not pre-defined, but that are present in the market. If NON-STANDARD UNITS such NSUs are commonly used in a regional market, it is The first step in preparing the NSU library is to establish the likely that household survey respondents are purchasing, list of common NSUs that will be used in the consumption and/ 13  THE USE OF NON-STANDARD UNITS FOR THE COLLECTION OF FOOD QUANTITY or production modules of the main survey. A list of common/ annexed in this guidebook are intended to be such a source, allowable item-unit (and when applicable, item-unit-condi- but there may be additional reviews available, such as Kor- tion) combinations for NSUs should include a comprehen- mawa & Ogundapo (2004) in Nigeria. sive set of valid NSU combinations for each crop and each When comprehensive reviews are not available, the next food item. Where applicable, crop/food condition (e.g., corn preferred source for common NSUs will be other surveys in husk or not, peanuts shelled or unshelled, fresh vs. dried already conducted in the country of interest that have allowed cassava) should be considered, especially for reporting har- quantities to be reported in NSUs. These could be either vested quantities, as the condition significantly impacts the household-level surveys with consumption or agricultural weight-volume ratio. While it may be impossible to prede- components, or market surveys. For these outside sources, termine all the possible combinations, the library should the survey designer must consider the comprehensiveness of endeavor to include combinations that represent the vast the NSU list. Some surveys may include only a few of the most majority of options (preferably higher than 90 percent). Even common NSUs and exclude less common though important when the same crops are grown and the same foods are con- ones. Likewise, the geographic coverage of the survey also sumed in different countries, it cannot be assumed that the needs to be taken into account. For example, surveys that same NSUs will be used in both places.2 only cover a small area may not contain NSUs that are com- The best practice for compiling this list would be for mon in other areas of the country. Unless the selection of national statistics agencies to establish a conversion-factor NSUs is clearly documented and comprehensive, the survey library independent of any specific household survey, which designer should seek additional information. can then be made available for use with any new survey within If resources and time permit, existing NSU lists can be val- the country. Unfortunately, many low- and middle-income idated with a small pilot survey to ensure the list is compre- countries have no such source for NSU conversion factors; hensive and current. The pilot survey can either be at the when they do, documentation and other supporting materials household or market level. Performing a household-level pilot are often limited or lacking. When a conversion-factor library has the advantage of capturing consumption units used by is available and well documented, household survey teams may households, which may differ from the units used in market choose to optimize timeline and budget constraints by using transactions. However, conducting even a limited market-level this existing resource. When such a library does not exist, or pilot survey will allow for the collection of a wide array of if existing conversion-factor data are limited or outdated, an item-units in a single market, whereas it may take several NSU-focused market survey must accompany the household households to acquire a comprehensive list. The pilot survey survey. Instructions for the market survey are discussed in should be largely open ended, allowing respondents (either detail in the next section. When implementing a market sur- household members or market vendors) to report in the units vey to fill these data gaps, the procedure for identifying the with which they are most comfortable or in the units that are NSUs to include in the market survey will vary depending on most commonly available. the stage at which this step is performed. Many units may be available in different sizes, such as the MARKET SURVEY BEFORE/ array of dengus shown in Figure 2. In this case, simply listing INDEPENDENT OF MAIN SURVEY dengu in the selected unit list would not sufficiently help stan- dardize this NSU. When there is variation, the unit list should When the market survey is implemented before or indepen- include the possibility of different sizes (e.g. small, medium, dent of the main survey, the first step is to seek any infor- large) and the weights and reference photos for each size mation on common NSUs within the country. Identifying should be collected. This is particularly important for units common NSUs and the items they apply to can be quite chal- such as pieces or heaps, which are subject to within-unit lenging, depending on the quality of information available to weight variation. When NSUs are coupled with reference help guide selection. A best first source would be a com- photos depicting multiple measured sizes, it provides greater prehensive review of NSUs within a country. The libraries comparability across reported NSUs by standardizing the respondents’ reference points. For example, if tomatoes are 2  Cross-country comparisons may be used to check the consistency of allowable scarce in only one region, what is considered a “large tomato” item-condition combinations and to reconcile food-weight densities for common may be equivalent to a small one elsewhere; by providing crops. This cross-country harmonization will be the focus of future work in this series. standardized photo references, the respondent can point to 3. Guidelines and Procedures for Capturing and Using Non-Standard Units  14 “their” tomato, thus ensuring its weight is converted in a photos of item-unit combinations for respondent interviews. standardized way, regardless of local variation. Although the procedure for collecting both these items may seem straightforward, strict protocols should be followed Note that a key benefit of conducting an ex-ante market to obtain the most accurate conversion factors and produce survey and using photo references in the main survey is that useful reference photos. much of the regional variation can be eliminated, which in turn limits the scope and burden of the market surveys to be A dataset of national or regional conversion factors for conducted. Without the ex-ante collection (which allows for all allowable combinations will be the main analysis compo- greater standardization with fewer measurements), you will nent of the library. The listed conversion factors should be need to collect and compare NSU data from markets in all provided at the lowest feasible (and representative) level of regions to avoid under or over-reporting consumption across regional disaggregation. The general procedure for collecting regions with different concepts of reference sizes. weight measurements involves (1) finding vendors who have the necessary non-standard item-unit combination, (2) prop- MARKET SURVEY AFTER MAIN SURVEY erly weighing the item-unit, and (3) recording the weight of the item-unit. For market surveys conducted after the main survey, the NSU list can be constructed based on the units observed Step 1: Finding item-units to weigh in the data. Constructing the list ex-post can shorten the list Armed with the list of item-unit combinations to weigh, enu- of weights needed to exactly those necessary to make use merators should seek out each of the combinations from of the data while ensuring there are no gaps in the eventual vendors in the market. Each item-unit measurement should conversion-factor data. be taken from multiple vendors to account for any varia- When constructing the list of item-unit combinations, the tion in vendors’ subjective assessment of what constitutes data should be examined for combinations that are commonly a unit amount as well as for possible enumerator error in observed. Invalid combinations should be excluded. Ideally, the measurement itself. For each item-unit pair we recom- every valid item-unit as well as crop condition observed in mend collecting measurements from three different vendors the data should be included in the list of units for the market within each market if time, personnel, and budget constraints survey. However, if the list of observed item-units is exten- permit. sive and/or resources for conducting the market survey are Survey teams need not limit measurements to the prede- limited, then the item-unit list can be shortened. The most termined list of item-units. If additional item-unit pairs are obvious method is to eliminate the least commonly observed found at the market, record these as well. This will be partic- item-units. This will depend greatly on the survey, but in gen- ularly beneficial when the market survey is conducted before eral only very infrequently observed combinations should be the main survey as it will allow the new units to be incorpo- dropped. rated into the main survey. The extent of regional variation in reported NSUs should If an item is available in the market, every effort should be also be assessed to determine at what level the market survey made to collect all the listed unit options for that item. The should be conducted. This can be done by comparing com- greatest challenge at this stage will likely be that some items mon units at various geographic levels. If at most geographic or units are not found due to seasonal availability of the item levels the item-units are similar, then a national list can be or limited use of a unit for sales. One solution is to search for constructed. However, if there is significant variation across the item-unit at vendors nearby who are outside the formal regions, it may be more appropriate and feasible to disaggre- market. If the item-unit is found there, the alternative location gate the item-unit lists to the regional level. should be noted by the enumerator. 3.3 COLLECTING WEIGHTS FOR The day of the market visit could also be an important CONVERSION FACTORS determinant of NSU availability. In many communities, there are specific days designated as “market days.” On market The two main purposes of the market survey are to col- days, a wide array of traders and farmers will participate in lect weights in order to calculate conversion factors for the market and thus, a greater selection of items and units consumption and production NSUs and to take reference will likely be available. However, there will also be more 15  THE USE OF NON-STANDARD UNITS FOR THE COLLECTION OF FOOD QUANTITY activity on these days, making it harder to perform the mea- that fall out of a heaped container and onto the scale should surements. Vendors may also be less willing to participate in be cleared before weighing. the survey on a market day since they will be busier. Given Enumerators should not be responsible for determining this trade-off, we recommend survey teams visit the market the amount of the item with which a particular unit is filled. first on a non-market day to acquire all the measurements They should only weigh what vendors provide. This includes available, and then again on a market day if any items or units typical heaping practices for containers. If the local practice are missing. is to heap as much of an item into a container as possible, Since the focus of the survey is on consumption or produc- then that is what should be weighed; if leveling is common, tion units and not specifically on market units, some contain- then leveled containers should be measured. When container er-based units may not be found at the markets. This could quantities are available heaped and level, both should be mea- especially be the case for some agricultural production units sured and noted. used by farmers but not typically sold by vendors. Enumera- While most item-units will require physical weighing of tors can ask vendors about the units in which they themselves the unit, in some cases no weighing will be required. This is purchase items from farmers, asking them to demonstrate true for item-units that are commonly purchased prepack- the appropriate quantity of the item-unit pair. Alternately, aged, with the weight printed on the container. Some com- survey teams may be able to acquire such containers directly mon examples are bags of rice, tinned or canned foods (e.g., from the source, (i.e., nearby households or farms). Contain- tomato sauce), snack items, etc. Note that although these ers may be purchased or borrowed and then brought to the item-units need not be weighed, reference photos must still market for filling and weighing. be taken since respondents may not remember the weight Larger units, especially those used for measuring harvest of the package but can identify which size/shape package they quantities, may not be available from market vendors, but consumed. can be collected at the market scale station (further details For most consumption units, collecting the weights will be below) or from wholesale traders in the market. fairly straightforward. However, for larger units— especially Locating a crop at the market in its various conditions will those used for production—there may be additional chal- likely be difficult. Many crops will only be available in their lenges. Heavier item-unit pairs are often beyond the maxi- final condition before consumption: cereals will likely be mum range of the portable scales used for the survey. When threshed; legumes will likely be shelled. In such cases, addi- this is the case, there are two potential solutions: tional weights may need to be acquired by conducting some measurements at the farm-household level. A limited number • Break up the unit into a series of smaller groups that can of condition-specific weights can be used to create conver- be weighed separately. Once all the groups have been sion factors across item-unit-condition pairs. weighed, they can be added together to acquire the total weight of the item-unit. Depending on the size of the Step 2: Weighing the item-units item-unit and the maximum range of the scale used, this Once an item-unit is located, it must be properly weighed. can be a laborious and time-consuming process. Further- When weighing a container unit, the empty container’s weight more, vendors may be unwilling to open larger units (if must be excluded from the measurement. Many modern sealed) and have them handled by enumerators. scales can automatically subtract the weight of a container • Make use of other scales that have a higher maximum (the “tare” weight) from the total weight. If the scale being weight. These can be either additional scales that enumer- used does not include the option to zero out the tare weight, ators bring with them to the market or higher-capacity then the subtraction must be done manually. scales found in the market. In many markets, there will Enumerators must be properly trained in the use of scales, be bulk traders or aggregators that purchase items from including how to identify appropriate (even) surfaces on which farmers for resale to market vendors. Since these traders to use the scales. Scales should be calibrated regularly during deal in large quantities, they will likely have a scale that fieldwork to ensure consistency across measurements. It is can measure these heavier weights. Making use of these important that the scale be kept clear of any other objects, market scales may be easier than breaking up the item- including any spillage from containers. For example, any grains unit into multiple groups, but it does require an additional 3. Guidelines and Procedures for Capturing and Using Non-Standard Units  16 step: calibrating the market scale. In general, market scales comparison of the reference photos and reassignment of size. may not be as advanced or accurate as the main scales However, this can be burdensome if there are many measure- used for the survey. Any error in the market scale mea- ments. An alternative method is to classify measurements surement must be estimated and corrected. That can be based on their position in the distribution of measurements done by selecting an item that weighs close to the maxi- for that particular item-unit pair. The most basic approach is mum of the survey scale. This same item should then be to classify observations that fall below the 33rd percentile as weighed using the survey and the market scales and both small, between the 33rd and 66th percentile as medium, and measurements should be recorded. Comparing these two above the 66th percentile as large. However, the number of measurements will allow for error correction in the mar- sizes must be considered before applying this method. Some ket scale’s measurement during the data-review process. units may only be found in two relatively uniform sizes, in Only one calibration is needed for every market scale. which case only small and large size should be assigned. If Since it may be impossible or at least impolite to adjust possible, a review of the photos is arguably a more compre- or even scrutinize the market’s scale, consider doing this hensive approach, or at least a verification step, to solving after all unit measurements are collected. This procedure this problem. requires that the measurement tool also be noted: survey For some items, the additional component of condition scales or a market scale. These scales will typically not be will also need to be taken into consideration when calculat- as precise as the smaller-capacity scales, but they are suf- ing conversion factors. In most of these cases, conversion ficiently precise for larger units. If higher-capacity market factors should be applied not only for converting to kilogram scales are not available or common in the market, then amounts, but also to render the quantities comparable to larger-capacity scales may be acquired for use by enumer- each other. For example, maize/corn can be harvested on the ators for larger units. cob (usually fresh) or without the cob in grain form (usually Another potential challenge for production units is the dry). The kilogram conversion reported for fresh, on the cob adjustment of weights by bulk traders or aggregators. In some maize is not directly comparable with the kilogram results of cases, traders will purchase an item-unit from a farmer and dry maize grains. To compare all reported maize conditions adjust the weight before distributing it to market vendors. For with each other, the conversion between maize on the cob example, a farmer may bring a sack with 115 kg of wheat, but and maize dry kernels/grains is also needed.3 after purchasing it, the trader might adjust the weight of the Once cleaned, the measurements must be aggregated to sack to an even 100 kg before selling it to market vendors. an appropriate level. The mean or median measurement for The purpose of the market survey is to acquire conversion each container unit can be used. For non-containers, the con- factors for units reported by farmers, so every effort should version factor will be item specific and should correspond to be made to weigh the item-unit the farmer brings to the the reference photo included in the library. If there is signifi- market (e.g., the 115-kg sack of wheat) before it is adjusted cant regional variation, then regional-level conversion factors by the trader. should be given. Otherwise, national conversion factors are Step 3: Calculating the conversion factors adequate. The conversion-factor database should be orga- nized so that there is a single conversion factor for each item- Calculating conversion factors can be a complicated process. unit at the appropriate geographic level, though there may be Results from the market survey should be cleaned and outli- some item-units not found in a particular region. Therefore, ers scrutinized. If there are relatively few measurements for we recommend that national-level conversion factors also be each item-unit, outliers can distort conversion factors sub- provided even if there is significant regional variation. Figure 6 stantially. If different sizes for a unit were allowed, the mea- presents a subset of the conversion-factor library for Nigeria. surement data may require further processing. A problem In the figure, conversion factors are provided for the six geo- may arise where classifications of a small, medium, and large political zones as well as the national average. versions of a unit could vary considerably. For example, the small size of a unit found in market X may be larger than the large version collected in market Y. These must be reconciled so that there is a standard classification of small, medium, and large within the relevant level of geographic aggregation (e.g., 3  The appropriate adjustment factors for this exercise are not part of the original region, state). This can be done manually through review and set of libraries found in Annex 11, but could be considered in future conversion libraries and methodological research.  17  THE USE OF NON-STANDARD UNITS FOR THE COLLECTION OF FOOD QUANTITY Figure 6 — Excerpt from a Conversion Factor Library for Nigeria NIGERIA GHS-PANEL WAVE 3 CONVERSION FACTORS ITEM ITEM UNIT UNIT UNIT CONVERSION FACTOR (KG) CODE NAME CODE DESCRIPTION SIZE NATIONAL BY ZONE AVERAGE NORTH NORTH NORTH SOUTH SOUTH SOUTH CENTRAL EAST WEST EAST SOUTH WEST GRAINS AND FLOURS 10 GUINEA 11 Paint rubber 3.612 3.758 3.612 3.612 3.768 3.832 2.828 CORN/ 12 Milk cup 0.161 0.205 0.125 0.163 0.180 0.161 0.159 SORGHUM 13 Cigarette cup 0.205 0.205 0.205 0.205 0.215 0.198 0.205 14 Tin 14.738 15.510 14.738 14.738 13.965 14.738 14.738 20 Congo Small 1.000 1.280 1.000 1.000 1.000 1.000 .720 21 Congo Large 1.978 1.978 1.978 1.978 1.978 1.978 1.978 30 Mudu Small 1.073 .978 1.103 1.145 1.073 1.060 1.073 31 Mudu Large 1.353 1.368 1.248 1.445 1.353 1.353 1.353 40 Derica Small 0.238 0.238 0.238 0.238 0.238 0.138 0.338 41 Derica Medium 0.639 0.639 0.612 0.639 0.639 0.639 0.745 42 Derica Large 1.587 1.587 1.587 1.587 1.813 1.587 1.361 43 Derica Very large 1.889 1.889 1.889 1.880 1.890 1.870 1.925 51 Tiya Medium 1.825 1.825 1.825 1.825 1.825 1.825 1.825 52 Tiya Large 2.650 2.650 2.650 2.650 2.650 2.650 2.650 60 Kobiowu Small 0.595 0.595 0.595 0.595 0.595 0.595 0.595 61 Kobiowu Medium 1.110 1.110 1.110 1.110 1.110 1.110 1.110 62 Kobiowu Large 1.210 1.210 1.210 1.210 1.210 1.210 1.210 11 MILLET 11 Paint rubber 3.765 3.672 3.765 3.765 3.767 3.805 3.840 12 Milk cup 0.153 0.153 0.145 0.165 0.150 0.153 0.155 13 Cigarette cup 0.210 0.210 0.210 0.210 0.215 0.205 0.210 14 Tin 15.060 15.685 15.060 15.060 14.435 15.060 15.060 20 Congo Small 0.924 1.160 0.924 0.924 0.924 0.924 0.688 21 Congo Large 1.437 1.437 1.437 1.437 1.437 1.437 1.437 30 Mudu Small 0,988 0,893 1,058 1,135 0,988 0,988 0,988 31 Mudu Large 1,260 1,260 1,210 1,323 1,260 1,170 1,260 40 Derica Small 0,243 0,243 0,243 0,243 0,243 0,145 0,340 Source: World Bank, LSMS Team. 3.4 COLLECTING REFERENCE PHOTOS factors as described above, and something in the naming scheme of the photos should make this connection clear. The second element that must be collected during the mar- This index of photos will be used to prepare the photo ref- ket survey is a set of non-standard unit reference photos. erence album. After the market survey, all photos should be compiled and included in the library. This index should contain photos of WHICH ITEM-UNITS REQUIRE all allowable item-unit combinations, with each one directly linked to the measurements used in the conversion-factor PHOTOS? list. For example, the pieces of yam in a photo should be Ideally, each item-unit included in the survey (including pre- exactly the same pieces used to calculate the conversion packed foods) will have a reference photo. Practically, surveys 3. Guidelines and Procedures for Capturing and Using Non-Standard Units  18 may wish to limit the photo book to items that represent the reference album that will be shared with respondents. a significant portion of the total food consumed or total When enumerators are instructed to take photos of all mea- food expenditures. It is essential that separate item-unit pho- surements, the research team will have multiple pictures to tos be taken for each non-container unit such as pieces or choose from when compiling the photo reference guide. heaps. However, for container units (pails, plates, etc.), a For the reference photos to be useful, they must depict single photo for each container may be sufficient since the the referenced quantities in a way that can be easily under- volume of the container does not vary with the item it holds. stood and interpreted by survey respondents. Regardless Item-specific photos of containers are useful if the fill level of the enumerators’ general familiarity with taking photos, (heaped/level) varies significantly across each item. If units ample time should be allotted for training enumerators on are expected to differ by region (e.g., only the North uses the photo requirements for this exercise. Effective and eas- baskets, or the object called a pail in the West is different ily interpreted reference photos should adhere to these from the pail used in the East) then different photos must guidelines: be taken in each region as well. However, for units that are relatively uniform across the survey area, only one photo • Photos should be well lit so that respondents can easily need be taken. see the items and differentiate between the item and its GUIDELINES FOR REFERENCE PHOTOS shadow or background. The primary purpose of these photos is to compile a ref- • When possible, a plain background should be used for erence album for use during household survey data collec- each photo. This could be a piece of paper, a sheet, or tion. With this tool, respondents can estimate quantities in some other material. The plain background will serve to relation to the related reference size. For example, when better highlight the item, especially when its color con- shown a reference photo of a potato during the household trasts with the item color. survey, a respondent can say she ate three potatoes of the • Each photo should contain only one food item or one size shown, or consumed one potato that was about half the food unit. For example, a photo of shelled groundnuts size of the reference potato. Additionally, the photos serve should not include unshelled groundnuts or maize; a an internal purpose in the creation of the conversion factors, picture of pails (a unit used for various items) should not as they can be used for verification/validation of weight mea- include bunches or piles of a particular food. surements collected by the market team. The photo quality, while important for both applications, is far more critical for • For units that come in various sizes (e.g., small, medium, large), all sizes of the item-unit must be present in the Figure 7 — Correctly Photographed Sahins of Rapeseed The near horizontal side angle shows the containers are filled in a "heaped" style, allowing for better understanding of the volume.  Source: World Bank, LSMS Team. 19  THE USE OF NON-STANDARD UNITS FOR THE COLLECTION OF FOOD QUANTITY Figure 8 — Correctly Photographed Tasas of Sunflower Seeds Plain white background contrasts nicely with the items pictured.   Reference item is included. It is placed next to item for easy com- parison, and is an appropriate size given the size of the item-unit. Source: World Bank, LSMS Team. same photo to help respondents differentiate between and include the exact same reference item (positioned sizes. The items should always be in the same size order the same way relative to each item-unit). (i.e., left to right, ordered from small to large) in the • A size reference item must be included in the picture picture. However, some units may be too large to include to illustrate the relative size of the main objects. The the size variations in a single photo. For such units, spe- item should be something that generally comes in one cial care must be taken to ensure that the photos of the standard size, is easily identifiable to respondents, and different sizes are directly comparable – this means that could be brought to interviews by enumerators. Exam- they are taken from the same angle and same distance Figure 9 — Correctly Photographed Heaps (Medebs) of Papaya Photo taken from the side angle shows the  items stacked underneath, helps in under- standing the volume of the heap. Source: World Bank, LSMS Team. 3. Guidelines and Procedures for Capturing and Using Non-Standard Units  20 Figure 10 — Correctly Photographed Empty Pails (to be used as unit reference for multiple items) Photo includes only one reference unit to avoid confusion for respondent.  To reduce enumerator error in recording  responses, items are pictured in order (small, medium, large), which is done consistently across all photos. Source: World Bank, LSMS Team. ples include a water or soda bottle, a writing pen, a box bucket, but it is impossible to tell from the photo. The item of matches, etc. This is a critical component of the photo. is also not photographed in its original container, which Without it, respondents may not be able to accurately makes it more difficult to understand the volume. judge the size of the item-unit in the photo. Figure 12 features three different sizes, but the direct over- • The dimension or volume of the item-unit must be clear. head angle may be misleading for piles of vegetables. Does Usually this means taking the picture from a side angle, the large pile have only the five pieces shown, or are there either directly horizontal to the item, or slightly above more stacked underneath? How many pieces are really in the horizontal. For some non-container units such as pieces, medium pile? There is also no reference item, so it is impossi- aerial photos (taken from directly above) may be accept- ble to tell if the small items are the size of golf balls or tennis able or sometimes preferred. The key is to ensure that balls. Finally, the items are in reverse order (large to small); the volume of the item is conveyed in the photo. assuming the other photos and the questionnaire list/label units from small to large (as is most commonly done), then Several example photos are shown in this section. Figures 7 photos that do not follow this pattern will increase the likeli- through 10 (above) are examples of photos that follow these hood of enumerators incorrectly recording (transposing) the guidelines. Each photo has a reference object (a soda bottle in unit size of the item shown during data collection. this case), a plain background, sizes shown in the appropriate order, and all taken from an angle that allows respondents to In Figure 13, all three sizes are included, as is a reference accurately gauge the size/volume of the unit. item. However, the background adds a lot of unnecessary dis- traction. And the inclusion of onions in the photo may confuse Figures 11 through 14 are examples of photos that were respondents. not taken correctly and will be difficult for respondents to interpret. Figure 11 shows a direct overhead view, whereby In Figure 14, the items are also in reverse order. More prob- the volume of the container cannot be accurately gauged. lematic, though, is that the small basket (on the right) was pho- The photo could be of a shallow plate or a very deep tographed separately using different backgrounds, angles, and 21  THE USE OF NON-STANDARD UNITS FOR THE COLLECTION OF FOOD QUANTITY Figure 11 — Incorrectly Photographed Mudu of Gari  Shallow pan or deep bucket? Overhead view makes it impossible to related volume. Photographed in a different container. Since the mudu is not in the picture, it loses context for respondents.  White Gari One Mudu 1.25kg Source: World Bank, LSMS Team. Figure 12 — Incorrectly Photographed Heaps of Sweet Potato Photo is large-to-small; if all the others are small-to-large, this may lead to mix-  ing up small and large codes in responses.  The background color is not ideal—the item blends into the background. A reference item is missing; it is unclear if this sweet potato is longer or shorter than a common pen, for example.  Sweet Potatoes (heap) 3.57kg 2.23kg 1.0kg Source: World Bank, LSMS Team. 3. Guidelines and Procedures for Capturing and Using Non-Standard Units  22 Figure 13 — Incorrectly Photographed Heaps of Green Peppers  The items in the background are very distracting. Photos should focus on only the item in question, to avoid confusion. Source: World Bank, LSMS Team. Figure 14 — Incorrectly Photographed Baskets Two different pictures are joined together, each taken from a different angle and distance. This makes it harder  to relate the size of objects in the different pictures.  Not clear which basket is small, which is medium, which is large. Source: World Bank, LSMS Team. 23  THE USE OF NON-STANDARD UNITS FOR THE COLLECTION OF FOOD QUANTITY distances from the camera for both the basket itself, as well as the list of allowable item-unit combinations, but should instead refer for the basket in relation to the reference item. All these details to the list to ensure that the item-unit provided by the respondent make the small basket in the photo visually similar (or greater) is indeed valid. If the unit the respondent gives is not listed, then the in appearance than the medium basket. This is not a useful ref- enumerator should use his or her judgement regarding its validity. erence for a respondent and will compromise the accuracy of After the respondent has specified the quantity in the preferred the data reported. unit, the enumerator should check to see if there is a reference photo for the item-unit. If there is, the enumerator should show the To comply with the guidelines discussed above, survey teams photo to the respondent and verify that the pictured unit is similar will often need to have access to a large staging area in which to to that referred to by the respondent. If applicable, the enumer- take photos, especially for larger units or when more than one ator should also ask which size of the unit most closely matches. size of a unit is photographed at once. In compact or crowded The respondent may need to re-estimate his or her consumption/ markets, this can be a significant challenge. Enumerators may production in terms of the reference photo. not have enough space to position the camera sufficiently far away to capture all elements in the photo. Similarly, enumera- tors may block passages in the market when taking the photos. 3.5 HOW TO USE THE NON- This can cause disruption in the market and create animosity STANDARD UNITS LIBRARIES from vendors or market patrons. If the market is crowded or Once all the necessary components are assembled into the NSU very compact, enumerators should try to find a staging area library, protocols should be drafted to provide guidance to both where they can take photos without much difficulty or distur- enumerators and data users on how use the library. Clear pro- bance. They should then attempt to collect as many measure- tocols for using NSUs and reference photos must be provided ments from vendors near the photo staging area. to field teams for the primary household and agriculture sur- veys. Likewise, clear protocols for using the conversion factors CREATING AND USING THE PHOTO must be provided to data users. Each NSU library should include REFERENCE ALBUM clear, coherent, and concise documentation so that the libraries Photos collected from the market survey should be scru - can easily be used by researchers and field teams. Incorporation tinized. The best photo should be selected for each item-unit of NSU materials into household and agriculture surveys will combination for inclusion in the bank of reference photos. require additional preparation, which can be done by the house- This reference tool will be used to establish a clear connection hold survey team or as a final step in completing part of the NSU between the respondent’s reporting and the established con- library documentation, in which case each household survey version-factor database. If the size of the reference album must team will need to evaluate the available materials in order to be limited, focus on the most commonly reported NSUs. Ref- adapt them to its needs. This preparation includes revising the erence aides should be printed in color, using a durable material consumption (and when applicable, harvest) questionnaire sec- that will withstand fieldwork (such as cardstock, or laminated tions to include NSUs and crop conditions; preparing the photo paper), or should be shown on tablets if the survey is conducted reference guides to be used by teams; and providing instructions using computer-assisted personal interview (CAPI) technology. to enumerators on how to effectively incorporate these new The list of reference photos should be organized to match the resources. Annex I contains examples of documents used to survey sequence and thus facilitate its use in the field. Forcing incorporate NSUs into the Ethiopian Socioeconomic Survey: the enumerators to flip through a multitude of pages of photos to food consumption section of the household survey, including a code find a particular item or unit will waste time and result in frustra- sheet for reporting NSUs; a photo reference guide to be printed, tion on the part of both the enumerator and the respondent. Both bound, and used by enumerators during their interviews; and a sam- CAPI and paper-based surveys can benefit from printed photo ref- ple section of an enumerator training manual that provides instruc- erence albums, which can sometimes be more easily shared with tion to enumerators on the use of the questionnaire and the photo reference guide. A snapshot of the Ethiopia consumption question- respondents during an interview. naire that incorporates NSUs is shown in Figure 15. When administering the consumption or agricultural produc- tion questionnaire, the enumerator should allow the respondent The LSMS team has created conversion-factor libraries for to report quantities in the unit with which the respondent is most Nigeria, Ethiopia, Malawi, and Uganda, with more planned for familiar. The enumerator should not provide the respondent with Tanzania, Niger, Mali, and Burkina Faso. They are provided 3. Guidelines and Procedures for Capturing and Using Non-Standard Units  24 online in Annex II as they were used to support LSMS Researchers conducting their own fieldwork can begin by includ- data-collection efforts in each country. The process of com- ing the existing lists of allowable item-unit pairs into consumption piling these libraries has made it possible to further refine the and production questionnaires, training enumerators on the proper guidelines and best practices outlined herein. use of photo reference aides, and incorporating the provided data- set of NSU conversion factors into interview and data entry checks. Although some of these libraries may not have the complete When possible, research teams should do a brief pilot test of the set of recommended items or may have some photos that do not commonly-available NSUs in their survey area, as these may change meet all the stated recommendations, they can still serve as valu- over time or vary across regions; conversion-factor data and pho- able resources. Both partial and completed libraries can be used tos would only need to be collected for any newly available combi- by researchers and fieldwork teams to help increase the accuracy nations. For research projects focused on analysis of existing data, of reported quantities in their own work, without incurring the where that data allowed for NSU reporting but conversion factors significant time-cost burden required to establish a new set of con- may not be available, the LSMS libraries can help increase the num- version factors. Even so, the libraries should be considered living ber of usable observations. Annex II provides additional information documents, to be revised and updated with each new data-collec- and user instructions on each of the available libraries. tion effort. Available foods and commonly used units and quantities may vary over time, so even complete libraries should be reviewed and piloted prior to their use on a new project. Figure 15 — Excerpt of a Household Survey Allowing for NSU Reporting SECTION 5A: FOOD LAST 7 DAYS CONSUMPTION UNITS 1. 2. 3. UNIT SIZE UNIT CODE Over the past one week (7 days), did you or How much in total How much came Kilogram 1 others in your household consume any [ITEM]? did your household from purchases? Gram 2 consume in the INCLUDE FOOD BOTH EATEN COMMUNAL- past week? IF NONE Litres 4 F O LY IN THE HOUSEHOLD AND THAT EATEN RECORD 0 Centilitres 5 O SEPERATELY BY INDIVIDUAL HOUSEHOLD D MEMBERS Jog 8 I D YES...1 SEE UNIT SEE UNIT Melekiya 9 NO...2 NEXT ITEM CODES ABOVE CODES ABOVE Birchiko Small 31 QUAN- UNIT QUAN- UNIT TITY CODE TITY CODE Birchiko Medium 32 Birchiko Large 33 CEREALS 1 Teff Esir Small 61 2 Wheat Esir Medium 62 Esir Large 63 3 Barley 4 Maize Festal Small 71 Festal Medium 72 5 Sorghum Festal Large 73 Source: World Bank, LSMS Team. 4. Benefits of Using Computer-Assisted Personal Interviewing (CAPI) Materials collected to support the use of NSUs can be used with both paper-based and computer- based surveys. However, some aspects of the information collected for use with NSUs can be greatly enhanced when used with computer-assisted personal interviewing (CAPI). CAPI provides unique benefits when conducting a market survey, particularly with respect to the ability to directly link weight measurements with reference photos. When conducting the main food consumption or agricultural production survey, CAPI can make better use of collected reference photos as well as conversion factors (to identify outliers). Both these aspects are discussed in turn here. 4.1 CAPI FOR MARKET SURVEYS coordinates where each specific measurement is taken (or at least the more general market location) can be automatically Market surveys are ideal candidates for collection using com- captured by the CAPI device. Likewise, the date and time the puter-assisted personal interview (CAPI) technology. Per- measurement was taken can also be automatically recorded. haps the strongest advantage that CAPI collection has over paper is that photos can be directly linked to measurements. Collection using CAPI also allows for on-the-fly consis- When conducting a market survey using paper, one must tency checks. Since relatively few measurements will be taken ensure that the photos can be linked to the correct weight within a market, it is important to limit the potential for measurement observation. One way to ensure this link is to error when collecting weights in standard units. For exam- apply a rigorous naming scheme for the photos, referenc- ple, the current measurement can be compared with previ- ing the item-unit, the market in which it was taken, and the ous measurements and flagged if it is significantly different. measurement observation it refers to (if there are multi - Likewise, a predetermined reasonable range for a particular ple measurements within the same market). Renaming these item-unit can be applied. These bounds must be made flexible photos while conducting the survey can be time consuming and must only account for the most egregious mismeasure - for enumerators and can lead to mistakes. However, when ments. For example, for very small units, any measurement using CAPI software (such as Survey Solutions), photos can over X kg would be unreasonable. These kinds of checks can be taken immediately after recording the measurement and identify some common errors such as reporting weights in can be directly linked to that measurement observation. The grams instead of kilograms. photo is automatically named with a reference to that spe - However, there is at least one potential drawback to using cific case. In addition, CAPI software can provide a prompt to CAPI to conduct a market survey. In some cases, it could be enumerators to take a photo of the measured item. This can more difficult to move between item-units within the listing help ensure that there is at least one photo taken for every on a CAPI survey. While conducting a market survey, the item-unit measurement collected. enumerators will not go item-unit by item-unit. Instead, they In addition, CAPI technology also makes the collection will move within the market collecting what item-units they of additional metadata much easier. For example, GPS see, not necessarily in order. For the CAPI program to be 25  4. Benefits of Using Computer-Assisted Personal Interviewing (CAPI)  26 usable in the market setting, enumerators must be able to building them into the parameters of the survey reduces the move easily between item-units in the list. Survey Solutions number of invalid observations reported during data collec- CAPI allows for such flexibility; evaluation of other software tion. By applying conversion factors to data as they are being options should take this into consideration. collected, reporting errors can be flagged and reviewed with respondents at the time of the interview, further reducing 4.2 CAPI FOR FOOD CONSUMPTION the number of invalid observations and eliminating the need AND AGRICULTURAL PRODUCTION for costly follow-up visits. CAPI programs can include checks of each item, including confirmation that price per kilogram SURVEYS and/or total and per capita standard-unit quantities are within All the reference library resources detailed above can be reason. Some CAPI programs can also generate checks and used with both paper-based and computer-assisted per- reports compiled across multiple items entered, creating a sonal interviews (CAPI). Several CAPI-based programs have summary list of all crop harvests in kilograms, listed in order capabilities that allow photo references to be incorporated of quantities, that enumerators can review with households into the interview, so that an enumerator can share relevant for on-the-spot validations, ensuring that top-reported crops images with the respondent as an item is being discussed. In match farmer’s expectations, for example. When collect- several cases, programs connect the photo directly to the ing data on household consumption, caloric values can be item-unit combination represented, so that “selecting” the included to generate food-consumption summaries; enumer- photo automatically defines the conversion factor for the ators can review these immediately with household mem- item reported. bers, checking, for example, that the average caloric intake of household members is within reason, and that the ranking The importance of collecting data on allowable item-unit combinations and calculating their conversion factors prior of foodstuffs by caloric share of diet makes sense. to the start of fieldwork is made even more critical with As with any survey using CAPI, it is worth emphasizing CAPI. When used with CAPI, these tools can create more the importance of dedicating sufficient additional time and dynamic in situ validation checks for enumerator use. Allow- resources to ensure the CAPI program is well programmed able combinations can be programmed into CAPI, so that and that all checks and validations are incorporated before only these options can be selected for any given item. The fieldwork—and even before training and piloting—begins. full set of such combinations is usually far more than an enu- This additional up-front time will ensure that interviews run merator can be expected to recall during an interview, so more smoothly, save time, and produce less data errors. 5. Conclusion Food consumption and agricultural production are two of the most important and commonly measured quantities for welfare analysis in low- and middle-income countries. Both are critical inputs into poverty estimates for these countries and agricultural production is essential for estimating farmer productivity. Many strides have been made in improving several aspects of these estimates, but until recently the challenge of converting non-standard (NSUs) has received less attention. The usual practice has been either to limit households to on how to create a complete NSU library resource for coun- reporting in standard units or to have enumerators estimate tries where one does not currently exist. In addition, the the conversion to a standard unit on an ad-hoc basis, both of Annexes to this Guidebook include sample questionnaire which can be very problematic and lead to poor estimates. instruments as well as resource libraries from the LSMS-ISA The use of NSUs can increase the accuracy of reported project (Nigeria, Ethiopia, Malawi, and Uganda). The librar- quantities in food-consumption and agricultural-production ies can be of use when working on any surveys or with any surveys. Reliably documented conversion factors for NSUs survey data in the selected countries. The Annexes include ensure that data robustness is not reduced by the loss of the allowable item- (and condition-) unit combinations for valid observations. each of the countries and all photo references collected. The Stata files containing the conversion factors are available at The objective of this Guidebook has been to provide www.worldbank.org/lsms under Publications/Guidebooks. advice to survey practitioners on incorporating non-stan- dard units into their surveys, along with practical guidance 27  REFERENCES Attanasio, O., & Frayne, C. (2006). Do the poor pay more? Presented at: Eighth BREAD Conference on Development Economics. Ithaca, New York. Beegle, K., De Weerdt, J., Friedman, J., & Gibson, J. (2012). Methods of household consumption measurement through surveys: Experimental results from Tanzania. Journal of Development Economics, 98(1), 3-18. Capéau, B. (1995). Measurement error and functional form: a proposal to estimate prices and conversion rates from the ERHS1994. Mimeo. Capéau, B., & Dercon, S. (2006). Prices, unit values and local measurement units in rural surveys: an econometric approach with an application to poverty measurement in Ethiopia. Journal of African Economies, 15(2), 181-211. Casley, D. J. & Kumar, K. (1988). Collection, analysis and use of monitoring and evaluation data. Baltimore, MD: The John Hopkins University Press. Deaton, A., (1997). The Analysis of Household Surveys: a Microeconometric Approach to Development Policy. Washington D.C. and Baltimore: The World Bank and Johns Hopkins University Press. Deaton, A., & Dupriez, O. (2011). Spatial price differences within large countries. Manuscript, Princeton University. Diskin, P. (1997). Agricultural Productivity Indicators Measurement Guide. Food and Nutrition Technical Assistance Project. Washington, DC: US Agency for International Development. Fermont, A., & Benson, T. (2011). Estimating yield of food crops grown by smallholder farmers. IFPRI Discussion Paper. Washington DC: International Food Policy Research Institute. Fiedler, J. L., Carletto, C., & Dupriez, O. (2012). Still waiting for Godot? Improving Household Consumption and Expenditures Surveys (HCES) to enable more evidence-based nutrition policies. Food & Nutrition Bulletin, 33(Supplement 2), 242S-251S. Kormawa, P. & Ogundapo, A.T. (2004) Local weights and measures in Nigeria: A handbook of conversion factors. IITA Monograph. Ibadan, Nigeria: International Institute of Tropical Agriculture. Murphy, J., Casley, D. J. & Curry, J. J. (1991). Farmers’ Estimations as a Source of Production Data. World Bank Technical Paper 132. Washington, DC: World Bank. Smith, L. C., & Subandoro, A. (2007). Measuring food security using household expenditure surveys (Vol. 3). IFPRI Technical Guide. Washington DC: International Food Policy Research Institute. Sud, U.C., Ahmad, T., Gupta,V.K., Chandra, H., Sahoo, P.M., Aditya, K., Singh, M., & Biswas, A. (2016). Research on Improving Methods for Estimating Crop Area,Yield and Production under Mixed, Repeated and Continuous Cropping. Global Strategy: Improving Agricultural and Rural Statistics, Working Paper No. 5. Rome: Food and Agriculture Organization of the United Nations. 28 ANNEX 1 SURVEY INSTRUMENTS NSU MARKET SURVEY: QUESTIONNAIRE (NIGERIA) NSU MARKET SURVEY: MANUAL (NIGERIA) HOUSEHOLD SURVEY: REFERENCE PHOTO ALBUM (ETHIOPIA) HOUSEHOLD SURVEY: CONSUMPTION MODULE (WITH NSUs) HOUSEHOLD SURVEY: TRAINING MANUAL (EXCERPT) * Additional examples available online 29 SELECT LSMS GUIDEBOOKS Measuring the Role of Livestock in the Household Economy Alberto Zezza, Ugo Pica-Ciamarra, Harriet K. Mugera, Titus Mwisomba, and Patrick Okell November 2016 Land Area Measurement in Household Surveys Gero Carletto, Sydney Gourlay, Siobhan Murray, and Alberto Zezza August 2016 Measuring Asset Ownership from a Gender Perspective Talip Kilic and Heather Moylan April 2016 Measuring Conflict Exposure in Micro-Level Surveys Tilman Brück, Patricia Justino, Philip Verwimp, and Andrew Tedesco August 2013 Improving the Measurement and Policy Relevance of Migration Information in Multi-topic Household Surveys Alan de Brauw and Calogero Carletto May 2012 Living Standards Measurement Study www.worldbank.org/lsms data.worldbank.org