Pacific Food Consumption Database Overview

Pacific Food Consumption Database (PFCD) Tuesday, 8 October 2024, Port Vila, Vanuatu Nathalie Troubat, Michael Sharp, Helani Kottage, and Olivier Menaouer, Pacific Community The PFCD was developed with financial support from the Australian Centre for International Agricultural Research (FIS 2018/155) and the World Bank (P169122)

Contents Timeline of HIES data harmonisation Input data Database structure Non-food data harmonisation Food data processing method: UNSD endorsed guidelines Food data processing method: Food Systems cleaning method Target users and example analysis Next steps and discussion

Food data harmonisation timeline International WB, FAO & IFPRI global (n=100) review of HIES food consumption modules UNSC endorsed guidelines for the processing of food data in HIES UNSC endorsed guidelines for the collection of food data in HIES Pacific HIES TWG recommended standard HIES (14-day diary) be adopted in the Pacific PSMB Development of the Pacific Nutrient Database Marshall Islands HIES Experiment to test UNSC guidelines recommendation to PICTs to adopt UNSC guidelines Pacific 2010 2011 2012 2013 2014 2015 2016 2017 2018 2019 2020 2021 2022 2023 2024

PFCD construct timeline 14-day diary 7-day recall Non-harmonised Experiment 10 5 0 2000 2002 2004 2005 2006 2008 2009 2010 2012 2013 2014 2015 2016 2018 2019 2021 2022 2023 Year 2019 2020 2021 2022 2023 2024 7-day recall food data routinely processed VUT, KIR, RMI, WFT WSM, KIR, PLW, NRU (nil) TON TUV 14-day diary food data reprocessed TON (2015), WSM (2018) TUV (2015), COK (2015) TOK (2015), PLW (2014) FSM (2013), NIU (2015) NRU (2015) PFCD Harmonisation of non-food variables anonymised, curated, presented to PSMB Non-food data harmonisation (nil) (nil) (nil) (nil)

Two distinct input data to PFCD Old HIES design (14-day food acquisition diary) Paper (PAPI) Expenditure recall over 7, 30, 90 or 365 day reference period New HIES design (7-day food consumption recall) Digital tablet (CAPI) Expenditure recall over 7, 30, 90 or 365 day reference period Self-reported, but adjusted (predicted rents using various methods) during data processing Individual-level 7-day consumption recall (amount only), by meal event (breakfast, lunch, dinner, snacks, drinks) Mode of data collection Method to collect non-food non- durable data Method to collect rents data Self-reported Method to collect food away from home consumption data Household level 30 day recall question on expenditure at hotels, restaurants, and the like Method to collect in-house food consumption data List of food items Household level, 14-day diary Household level, 7-day food consumption recall Open -Food acquisition via cash purchases -Consumption of home produced and gifted food Closed list of 80 to 120 foods, with other option) -Food consumption via cash purchases, home production, and gifts Closed list with country-specific units (with other option) 12-month of enumeration VUT, KIR, RMI, WFT, TON (2021) Objective of food module Unit of measurement Open with examples given Seasonality Datasets of PFCD 12-months of enumeration NRU, SLB, FSM, PLW, TKL, COK, TON (2015), WSM

Input data Final sample (HHs) Sample fraction (% of HHs) 14.9% In-house food acquisition Alcohol, tobacco, and similar Food consumption away from home Total PICT Cook Islands Micronesia, Federated States of Kiribati Marshall Islands Nauru Niue Palau Samoa Solomon Islands Tokelau Tonga-1 Tonga-2 Tuvalu Vanuatu Wallis & Futuna REGION Melanesia Micronesia Polynesia TOTAL Code COK FSM KIR MHL NRU NIU PLW WSM SLB TKL TON-1 TON-2 TUV VUT WLF population 14,684 102,544 118,480 54,388 11,627 Sub-region Polynesia Micronesia Micronesia Micronesia Micronesia Polynesia Micronesia Polynesia Melanesia Polynesia Polynesia Polynesia Polynesia Melanesia Polynesia Count of PICTs 5 7 10 22 Year Food data collection method Diary Diary Recall Recall Diary Diary Diary Diary Diary (with stock) Diary Diary Recall Diary Recall Diary (with stock) + indiv. FAFH 2016 2014 2019 2019 2012 2015 2014 2018 2012 2015 2015 2021 2015 2019 2019 680 1,610 2,182 870 450 156 838 2,996 4,364 118 1,803 2,128 701 4,549 995 13,812 32,846 34,483 18,058 10,712 3,404 19,811 85,124 121,216 2,488 40,952 62,310 9,517 114,280 19,624 367 378 3,998 793 454 472 4,777 2,672 372 9.8% 11.1% 5.8% 26.4% 30.4% 14.6% 10.3% 4.1% 47.0% 10.0% 11.3% 37.7% 7.2% 34.0% Data accessed under license between the Pacific Data Hub Microdata Library and the data producer PFCD only currently accessible by SPC staff and long-term consultants (Nathalie and Helani) This exercise is thus far an academic one and no results have been published 42 117 209 372 4,657 1,611 17,562 199,332 615,948 1,163 101,786 100,187 11,534 295,495 11,558 2024 total 11,760,000 553,900 680,800 12,994,700 86 433 1,495 4,575 95 262 2,980 123 4,476 908 28 827 5,554 247 1,979 1,693 MEL MIC POL PAC 2023 2023 2023 2023 8,913 5,950 9,577 24,440 8.1% 3.5% 0.5% 1.2% NA NA NA NA 22,112 151,781 400,932 588,637 1,003 5,322 13,085 19,777 1,721 9,402 14,087 25,664

Database structure

Non-food data harmonisation and anonymisation Variables Design Grouping Dependent Independent Food Household ID COICOP Country acquisition quantity Enumeration area Food edible quantity GIFT Sub-region Pacific Area (urban- rural) Strata Calories Healthy Living Source of food acquisition Sampling weight Macronutrien ts Geographic typology Food data collection method Micronutrient s Household size Wealth tercile

Name hhid iso Label Unique anonymised Unit ID ISO 3166 Appha-3 code Values 24,440 unique values Moderate 1 per PICT 1. Melanesia 2. Micronesia 3. Polynesia 1. Atoll 2. Mixed 3. High 1. Urban 2. Rural 3. No definition 15 surveys 2012 to 2021 2012 to 2021 January to December Low Diary or Recall 1. Low 2. Low-Medium 3. Medium 4. Medium-High 5. High PICT-unique PICT-unique EA-unique 1. 1 to 3 persons 2. 4 to 6 persons 3. 7+ persons 1 to 27 Disclosure risk Low subreg Pacific sub-region Low typo Gegraphic typology Low rururb Area of residence Low Low Low Low country survey_year enumeration_year Year enumeration started month Month enumeration started type Food data collection method Country and year of the survey Year survey started Low quality Quality of food data Non-food data harmonisation and anonymisation Low Moderate High Low strata ea weight Sample strata Primary sampling unit Sampling weight Name hhid iso Label Unique anonymised Unit ID ISO 3166 Appha-3 code Values 24,440 unique values Moderate 1 per PICT 1. Melanesia 2. Micronesia 3. Polynesia 1. Atoll 2. Mixed 3. High 1. Urban 2. Rural 3. No definition 15 surveys 2012 to 2021 2012 to 2021 January to December Low Diary or Recall 1. Low 2. Low-Medium 3. Medium 4. Medium-High 5. High PICT-unique PICT-unique EA-unique 1. 1 to 3 persons 2. 4 to 6 persons 3. 7+ persons 1 to 27 Disclosure risk hhsize_group Household size Low Low Moderate Moderate Low partakers numchild adulteq Partakers in in-house food consumption Number of children age <15 in household 0 to 25 Adult equivalent (0.5 if age <15) subreg Pacific sub-region Low 0.5 or 1 1. Yes 2. No 1. Yes 2. No 1. Yes 2. No 1. Yes 2. No 1. Yes 2. No 1. Renting 2. Living rent free 3. Owner occupier 1 to 3 typo Gegraphic typology agric Agricultural household Low Low fishing Fishing household Low rururb Area of residence Low Low Low Low livestock Livestock rearing household Low country survey_year enumeration_year Year enumeration started month Month enumeration started type Food data collection method Country and year of the survey Year survey started hcraft Handicraft-producing household Low drinking Drinking water improved Low Low tenure Tenure status of household Moderate Low Low Low Low quality Quality of food data tercile_exp pcconsexp tercile_inc pcincome Expenditure tercile (per capita) Annual consumption expenditure per capitaIn LCU Income tercile (per capita) Annual income per capita Low Moderate High Low 1 to 3 In LCU strata ea weight Sample strata Primary sampling unit Sampling weight hhsize_group Household size Low Moderate Moderate Low partakers numchild adulteq Partakers in in-house food consumption Number of children age <15 in household 0 to 25 Adult equivalent (0.5 if age <15) 0.5 or 1 1. Yes 2. No 1. Yes 2. No 1. Yes 2. No 1. Yes 2. No 1. Yes 2. No 1. Renting 2. Living rent free 3. Owner occupier 1 to 3 agric Agricultural household Low fishing Fishing household Low livestock Livestock rearing household Low hcraft Handicraft-producing household Low drinking Drinking water improved Low tenure Tenure status of household Moderate Low Low Low Low tercile_exp pcconsexp tercile_inc pcincome Expenditure tercile (per capita) Annual consumption expenditure per capitaIn LCU Income tercile (per capita) Annual income per capita 1 to 3 In LCU

Food data processing method: UNSD Information from the HCES on food data, household and household characteristics Information on prices and weights in grams of one unit of product Pacific Nutrient Database Step 3. Check and correct inconsistencies between quantities and monetary values for each combination product/unit of measurement Step 2. Code editing, harmonization and merging of all food datasets Step 4. Conversion of quantities into gram Step 1. Gathering all input files Step 5. Identify and correct potential outliers in the distribution of quantities in grams per capita and/or food monetary values Step 8. Aggregate and derive the per capita dietary energy consumption and food monetary values distributions Step 6. Convert into nutrient values all quantities corresponding to in house consumption Step 7. Convert into nutrient values the expenditures on meals consumed away from home

Food data processing method: Food Systems cleaning method

Target users and uses Pacific Statistics Officers (and health, agriculture, fisheries, etc.) SPC-UOW Food Systems 2 project has plans for regional / sub-regional / national trainings on use of PFCD (and the Pacific Food Trade Database) Governments of the Pacific region in policy, and for further use (e.g., PPPs, FPI, SDGs) Universities and Pacific-researchers in fields of agriculture, public health, economics, statistics, etc. International and regional organisations, including SPC (SDD, PHD, LRD, FAME, Food Systems), World Bank (PPPs), FAO/WHO GIFT Country-and regional-level analysis: Source of food Composition of food consumption Economic and nutrition classification Pacific policy relevance: Non-communicable disease Agriculture and rural development Welfare and nutrition

Country Country- -level analysis: Sugar consumption level analysis: Sugar consumption 0.25 0.20 Proportion of DEC 0.15 0.10 0.05 0.00 COK 2015 FSM 2013 KIR 2019 NIE 2015 NRU 2012 PLW 2014 Country and HIES year RMI 2019 SAM 2018 SLB 2012 TKL 2015 TON 2021 TVU 2015 VUT 2019 W&F 2019 Proportion WHO recommendation 1 WHO recommendation 2

Regional-level analysis: Fruit and vegetable consumption 450 Mean consumption (g/capita/day) 400 350 Independent variable Hypothesis testing result 300 250 Agriculture Agric HH Non-agric HH 200 Area Urban Rural 150 Urban No definition 100 Rural No definition 50 Sub region Melanesia Micronesia 0 No definition Agric HH Mixed Urban Atoll Rural Non-agric HH Melanesia Micronesia High Polynesia Melanesia Polynesia Micronesia Polynesia Typology Atoll Mixed Atoll High Agriculture Area Sub region Typology Food to choose Food to limit/avoid WHO recommendation

Two-part regression for cereal_wheat Binary logistic Coefficient Linear regression Coefficient Regional Regional- -level analysis: level analysis: Mean wheat consumption Mean wheat consumption P-value P-value rururb Urban Rural -0.01 -0.85 0.95 0.00 -0.65 0.05 0.00 0.32 Group_wheat Mean StdErr LowerBound UpperBound typo Atoll High Flour, wheat, white, plain, unfortified -0.90 0.06 0.01 0.82 0.62 0.23 0.00 0.00 18.71 0.59 17.56 19.87 Cereals & their products (wheat) 49.01 1.17 46.72 51.30 subreg Micronesia Polynesia 0.00 1.01 1.00 0.00 1.18 1.40 0.00 0.00 Composite dishes with wheat 13.48 0.37 12.75 14.22 Savoury snacks made from wheat 9.87 0.23 9.42 10.31 tercile_exp 2 3 1.06 1.81 0.00 0.00 0.62 1.08 0.00 0.00 Sweets made from wheat 24.80 0.70 23.43 26.16 subreg#tercile_exp Micronesia#2 Micronesia#3 Polynesia#2 Polynesia#3 ????= 1 ?? ??????_? ??? > 0 0 ?? ??????_? ??? = 0 -0.12 -0.53 -0.32 -1.08 0.51 0.00 0.11 0.00 -0.21 -0.31 -0.25 -0.47 0.01 0.00 0.00 0.00 hhsize constant 0.12 1.13 0.00 0.00 -0.03 2.62 0.00 0.00

Next steps and discussion Discussion Recommend the Pacific Food Consumption Database be made publicly available as a Public Use File or Licensed File in the Pacific Data Hub Microdata Library How to best promote the use of this asset? No independent variables on the reference person due to inconsistent definitions of the household head Add more independent variables? Which ones? Only be HIES derived? Amount in Local Currency Units (LCU) at time of reporting Any methodological or other concerns? Next steps Document and curate the database with the Pacific Data Hub Microdata Library Seek approval from data producers to make PFCD a publicly available database (Public Use File, or Licensed Dataset) Collaborate with Pacific NSOs and data users in the use of PFCD Write standard code that considers the complex sample design of each underlying survey, and the complex structure of PFCD to support use Add more datasets as they become available Thank you tumas!