The objective of this systematic literature review was to compile and assess food losses and waste estimates, from developed countries, across the food supply chain. The methodology involved systematically identifying studies and extracting, compiling and analysing their estimates of food losses and waste. Of the 55 estimates extracted, from these studies, the most (43·6%) were from the consumption (average 114·3 (kg/capita)/year) part of the food supply chain. On average, total food losses and waste were 198·9 (kg/capita)/year. While this review revealed a high degree of variability of estimates and inconsistent trends for the independent variables: scope of food waste, geography and study methodologies; food waste generation, at the consumption part of the food supply chain, was significantly higher for North American compared with European estimates (p = 0·003); and significantly higher (p = 0·030) for indirect than direct estimates. Similarly, total food waste generation indirect estimates were significantly higher (p = 0·035) than directly measured estimates. To improve the accuracy and precision of food losses and waste estimates, additional research is required to develop and implement a bespoke, weight-based and statistically sound methodology for its direct measurement.
World food production has increased substantially in the past century, as has calorie intake per capita (Nellemann et al., 2009). Nevertheless, food insecurity persists: according to the Food and Agriculture Organization (FAO), 795 million people are undernourished globally, including 15 million in developed regions (FAO et al., 2015).
The quantification of food losses and waste (FLW) is being used to draw attention to the poor use of food resources. According to Gustavsson et al. (2011), developed countries generate more FLW than developing countries. Its reduction presents opportunities to reduce its economic (e.g. wasting money), environmental (e.g. greenhouse gas (GHG) generation) and social (e.g. food security) impacts. To develop effective FLW reduction interventions and measure their impact, it is essential to have a more precise understanding of its generation. Since a variety of methods have been used to collect FLW data, precise estimates have been elusive. The objective of this systematic literature review is to compile and critically assess current annual per capita weight-based estimates of FLW along the various parts of the food supply chain (FSC) in developed countries.
Figure 1 depicts the various parts of the FSC that consist of ‘agricultural production, postharvest handling and storage’, processing and packaging, ‘distribution (i.e. retail sale)’ and ‘consumption’. The authors’ conceptualisation incorporates system boundaries adapted from Gustavsson et al. (2011), Nahman and de Lange (2013) and Parfitt et al. (2010). It highlights the progression of food from farmers to consumers. Each stage of the FSC is a FLW generation and intervention point.
Schneider (2013) summarises a number of definitions that have been applied to FLW. In this paper, the front part of the FSC (Figure 1) encompasses ‘agricultural production, postharvest handling/storage and processing’, with food that becomes unavailable for human consumption referred to as food losses (Gustavsson et al., 2011; Kummu et al., 2012; Parfitt et al., 2010). The back part of the FSC encompasses ‘distribution’ and ‘consumption’, with food that becomes unavailable for human consumption referred to as food waste (Parfitt et al., 2010). FLW are either deemed edible or inedible, which are referred by some as avoidable or unavoidable food waste, respectively (Beretta et al., 2013; WRAP, 2009). Edible FLW is food that was at one point edible. Inedible FLW is food that was never edible (e.g. vegetable peels, egg shells, bones).
The total production of edible food has been estimated at about 900 (kg/capita)/year in North America and Europe (Gustavsson et al., 2011). Meanwhile, estimates of total food available for consumption vary considerably: total food consumption in the developed world has been estimated at 1006 (kg/capita)/year (Alexandratos and Bruinsma, 2012). Per country food consumption estimates include 779 (kg/capita)/year in Canada (Abdulla et al., 2013), 500 (kg/capita)/year in the USA (Kantor et al., 1997), 500–600 (kg/capita)/year of food purchased for consumption in Finland (Tike, 2010; Viinisalo et al., 2008) and 687 (kg/capita)/year food consumption, at the retail level in Switzerland (Beretta et al., 2013).
Parfitt et al. (2010) reported that there is no consensus on the amount of global food production that is lost, with ranges of 10–40% and up to 50%. Gustavsson et al. (2011) estimated that a third of the food produced for human consumption, or about 1·3 billion tons/year is lost or wasted annually, but because many assumptions had to be made to develop these estimates, they note that the results must be interpreted with great caution. It was estimated that developed regions (Europe and North America) generate 95–115 (kg/capita)/year FLW, which is considerably higher than for developing regions (Sub-Saharan Africa and South/Southeast Asia), which generate 6–11 (kg/capita)/year. Abdulla et al. (2013), using Statistics Canada and World Bank data (1961–2009), estimated that the amount of FLW in Canada averaged 40% of food available for consumption and that in 2009 ∼7·3 Mt was wasted in Canada. Agriculture and Agrifood Canada (AAFC, 2015) reported there were ∼6 Mt/year of FLW from retail and household consumption.
There is some agreement among researchers about the state of FLW estimates. According to Parfitt et al. (2010), there is no consensus on the amount of FLW due to data gaps and uncertainties. Furthermore, many existing estimates link back to the same limited primary datasets, with much of the published data originating from fieldwork undertaken in the 1970s and 1980s. Langley et al. (2009) concluded that calculating and estimating the amount of food waste is a difficult issue due to a lack of real and meaningful data. Indeed, a number of researchers have identified that there are major data gaps in the knowledge of global FLW, that necessitates using secondary (i.e. indirect) rather than primary datasets, and that further research is urgently required to improve FLW estimates (Abdulla et al., 2013; Gustavsson et al., 2011). It is clear that there are some challenges with the available FLW estimates, including how these estimates are gathered and their precision.
Koester (2013) questions how FLW quantities have been calculated and suggests that current estimates are inflated. In questioning the results of Buzby and Hyman (2012), he suggests that rather than summing food losses, calorific values should be presented, although even this approach would result in an over-estimation because in some cases ‘food loss could have been economically rational’ (Koester (2013): p. 64). Elaborating on these assertions, Koester (2014) posits that the current definition of FLW is inadequate and not suitable for developing policies that contribute to food security, or improve efficiency of resource use and contribute to a sustainable environment. The foremost need is to develop appropriate measures, perhaps using multiple methods, for aggregating FLW across the FSC. Koester's (2013) arguments are echoed by Buzby et al. (2014), who note that FLW is becoming an increasingly important topic both domestically and internationally. Better estimates of the amount and value of FLW could help serve as quantitative baselines to develop interventions to reduce FLW generation.
A systematic literature review was performed by adapting methods described in Petticrew and Roberts (2006) and PRISMA guidelines (Liberati et al., 2009). This method, which is widely used in medical and social science fields, has been applied to this review to facilitate a systematic retrieval of relevant research papers. The purpose is to impart additional rigour to the literature review process. This was accomplished by collaboratively developing all search terms, identifying databases to be used and identifying inclusion/exclusion criteria in advance of starting the review.
Studies that examined the amount of FLW generated along the FSC were identified through searching the following databases: Scopus, Geobase and Web of Science.
The search included articles published between 1 January 1985 and 15 October 2015. The following search terms were used: ‘Food’ AND ‘Waste’ AND Quant*; ‘Food Waste’ AND Quant*; ‘Food Waste’ AND ‘Characterization’; ‘Food Waste’ AND ‘Cost’; ‘Food Loss’; ‘Food Losses’; ‘Food Waste’ AND ‘Composition’; ‘Food Waste’ AND Measure*; ‘Food Waste’ AND Agri* AND Quant*; ‘Food Waste’ AND Household* AND Quant*; ‘Food Waste’ AND Food Process* AND Quant*; ‘Food Waste’ AND ‘Supply Chain’ AND Quant*; ‘Waste’ AND ‘Characterization’ AND ‘Food’; ‘Waste Characterization’ AND Method*; ‘Waste Characterization’ AND ‘Food’; ‘Waste Audit’ AND Method*; and ‘Waste Audit’ AND ‘Food’.
The inclusion/exclusion criteria for studies included: (a) detailing research between 1985 and 2015, (b) English language, (c) quantitative and qualitative studies, (d) results of food waste quantification by weight and (e) research conducted in developed countries.
Relevant studies were identified first through title screening and then abstract reviews of titles that passed first screening. Studies remaining after abstract screening were subject to full-text screening and a final decision on relevance for inclusion in the review. Paper relevance was determined through the application of the inclusion/exclusion criteria. Studies from which weight-based FLW quantities could be extracted and normalised on a per capita basis were selected. This was confirmed by both authors.
Data points were extracted from studies and grouped by their respective part(s) of the FSC. Data points were normalised, where necessary, to (kg/capita)/year by dividing the annual weight of food waste generated by the appropriate population. The number of data points per part of the FSC were counted and averaged. The following independent variables were identified: (a) scope of FLW (inedible/edible or edible), (b) geography (Europe or North America) and (c) study methodologies (direct or indirect measurement). Statistical analysis of the data was undertaken by establishing null hypotheses that each independent variable had no impact on the amount of FLW generated for the dependent variables: ‘distribution’, ‘consumption’ and ‘total’ FLW, which were chosen because they had the most FLW data points. This was assessed using an independent sample's t-test.
Figure 2 depicts the results of the systematic review. After the database search and initial title screening, the authors screened the remaining abstracts of all articles in the reference list, using the inclusion/exclusion criteria. This resulted in a final reference list of 135 papers for full-text review.
The multi-staged search strategy with full-text review yielded 30 papers that met the inclusion criteria for final consideration in this review. The 30 papers included 17 papers that were found directly through the full-text review process and another 13 papers that were identified from a title review of the reference lists of these papers, as well as a scoping review of the grey literature.
Table 1 presents summaries of each of the selected studies. Essentially all the studies were North American (USA and Canada) and European (EU and Scandinavian countries). Almost all studies were undertaken and published after 2005 and include both edible and inedible FLW in their estimates, rather than segmenting out just the edible fraction of FLW. The studies included: regional estimates (e.g. Europe, North America and Oceania – Gustavsson et al., 2011); countrywide estimates (e.g. Canada – Abdulla et al., 2013; USA – Buzby and Hyman, 2012; Buzby et al., 2014); state-wide estimates (e.g. Hawaii – Okazaki et al., 2008), county estimates (Griffin et al., 2009); and neighbourhood estimates (e.g. Malmo, Sweden – Bernstad et al., 2012, 2013). Few studies provided estimates of FLW across each part of the FSC. Quantitative research included mostly uncontrolled studies that combined and extrapolated existing datasets to develop estimates, or the studies of weight-based waste composition measurements. Research studies also included surveys and diaries.
|Author||Year||Country/region||State of food||Method||Study methodology||Food supply chain: (kg/capita)/year||Total|
|Agricultural production||Postharvest handling and storage||Processing and packaging||Distribution||Consumption|
|Abdulla et al. (2013)||2009||Canada||Edible and inedible||Used secondary Statistics Canada data, which in turn used USDA ‘waste factors’, and World Bank data to develop estimates||Indirect||308·2|
|AAFC (2015)||2010||Canada||Edible and inedible||Used Statistics Canada and USDA data to develop estimate||Indirect||54·5||121·5|
|Bernstad et al. (2012)||2008||Sweden||Edible and inedible||Weighing all waste leaving the study area and repeated waste composition analyses of large samples from all fractions||Direct||95·7|
|Bernstad et al. (2013)||2009||Sweden||Edible and inedible||Weighing of all waste fractions, repeated waste composition analyses of disposed waste and a questionnaire||Direct||68·8|
|Brautigam et al. (2014)||2006||EU||Edible and inedible||Calculations are based on a FAO mass flow methodology (based on Food Balance Sheets, described in Gustavsson et al., 2011, 2013)||Indirect||95·2||30·6||33·5||14·4||114·8||288·5|
|Buzby and Hyman (2012)||2008||USA||Edible||Compiled estimates of food loss using the USDA ERS's (2014) Loss-Adjusted Food Availability (LAFA) data||Indirect||188·0|
|Buzby et al. (2014)||2010||USA||Edible||Compiled estimates of food loss using the USDA ERS's (2014) LAFA data||Indirect||63·3||195·0|
|Defra (2010)||2007||England||Edible and inedible||The results from various curbside and recycling depot waste audits were summarised to develop an estimate of waste quantity and composition, including food waste||Direct||83·2|
|EC (2010)||2006||EU||Edible and inedible||Used 2006 Eurostat data and various national sources to compile total food waste per country||Indirect||70·0||33·9||75·9||179·8|
|Edjabou et al. (2015)||2013||Denmark||Edible and inedible||A waste auditing methodology was developed and used to estimate the composition, including food waste, of curbside single family and multi residential waste||Direct||86·6|
|Edjabou et al. (2015)||2013||Denmark||Edible and inedible||A waste auditing methodology was developed and used to estimate the composition, including food waste, of curbside single family and multi residential waste||Direct||75·3|
|Garrone et al. (2014)||Not provided||Italy||Edible and inedible||Case studies were used to develop an extrapolated estimate of the amount of food waste generated from food processing and distribution/retail sales in Italy||Indirect||3·0||13·0|
|Griffin et al. (2009)||1999||USA||Edible and inedible||A case study was conducted in a single upper New York State County to estimate food waste generation along the FSC. Information was gathered through interviews and from published county, state and national sources||Indirect||19·3||1·2||17·6||57·6||95·6|
|Gustavsson et al. (2011)||2007||North America and Oceania||Edible||Used FAO datasets and assumptions to address data gaps to develop estimates of FLW along the FSC||Indirect||115·0||300·0|
|Gustavsson et al. (2011)||2007||Europe||Edible||Used FAO datasets and assumptions to address data gaps to develop estimates of FLW along the FSC||Indirect||95·0||280·0|
|Hodges et al. (2010)||2008||USA||Edible||Used 2008 USDA ERS, data to develop an estimate of food waste||Indirect||64·0||123·8|
|Jones (2006)||Not provided||USA||Edible and inedible||No explanation of methods||213·6|
|Jones (2005)||Not provided||USA||Edible and inedible||No explanation of methods||106·4||296|
|Kantor et al. (1997)||1995||USA||Edible and inedible||The USDA ERS (2014) developed preliminary estimates of food waste generation by retail, food service and consumer sectors. Many of the studies on which these estimates are based date from the mid-1970s or before||Indirect||9·3||155·0||266·3|
|Katajajuuri et al. (2014) and Silvennoinen et al. (2014)||2010||Finland||Edible||Household food waste data was collected through a diary study. Food service sector data was collected by way of on site (i.e. weighing of food waste along various parts of food service chain – for example, cooking, serving, customer leftovers). Retail sector and industry data was collected by way of interviews of various parties in the supply chain||Direct||14·0–26·0||26·0–30·0||23·0|
|Langley et al. (2010)||Not provided||UK||Edible and inedible||Tested a method of quantifying (self-weighing) and estimating the composition (through use of a diary) of food waste disposed by households||Direct||72·6|
|Lebersorger and Schneider (2011)||2009||Austria||Edible and inedible||Calculated the number (n) of urban and rural waste samples to collect and sort on the basis of a 95% confidence level, using mean and standard deviation (SD) data from previous waste analyses||Direct||49·6|
|Lebersorger and Schneider (2011)||2009||Austria||Edible and inedible||Calculated the number (n) of urban and rural waste samples to collect and sort on the basis of a 95% confidence level, using mean and SD data from previous waste analyses||Direct||18·8|
|Matsuto and Ham (1990)||1989||USA||Edible and inedible||Waste samples collected from curb in Madison, Wisconsin and Sapporo, Japan and manually sorted into a number of categories, including food waste||Direct||65·3|
|Matsuto and Ham (1990)||1988||Japan||Edible and inedible||Waste samples collected from curb in Madison, Wisconsin and Sapporo, Japan and manually sorted into a number of categories, including food waste||Direct||123·0|
|Okazaki et al. (2008)||2005||Hawaii, USA||Edible and inedible||Information collected on food establishments and their practices came from the State of Hawaii Department of Health list of permitted food establishments and a food waste recycling survey was developed and distributed to all permit holders||Direct||240·9|
|Parizeau et al. (2015)||2015||Canada||Edible and inedible||Data from green bin only and did not include household garbage destined for landfill. Also allowable green bin items include other items besides food waste||Direct||217·4|
|USDA ERS (2014)||2010||USA||Edible and inedible||Compiled estimates of food loss using the USDA ERS's (2014) LAFA data||Indirect||194·9|
|USEPA (2009)||2008||USA||Edible and inedible||USEPA collects and reports on waste generation and disposal data in the USA and has done so for the last 30 years. Municipal solid waste (MSW) includes household, commercial and institutional waste||Indirect||104·5|
|USEPA (2014)||2012||USA||Edible and inedible||USEPA collects and reports on waste generation and disposal data in the USA and has done so for the last 30 years. MSW includes household, commercial and institutional waste||Indirect||116·0|
|WRAP (2009)||2007||UK||Edible and inedible||Estimates of solid and liquid food waste were derived from annual waste generation totals obtained from municipalities, waste composition studies and diary research||Direct||134·3|
|WRAP (2013b)||2011||UK||Edible and inedible||Gathered production and retail data and applied estimates of food waste to develop estimates. Developed a mass balance to estimate food waste types and how are they are managed||Direct||61·7||7·1|
|WRAP (2013a)||2012||UK||Edible and inedible||The estimates for household food and drink waste were derived from a summary of waste audits and kitchen diary research||Direct||130·3|
Figure 3 depicts all of the FLW data points by position on the FSC. It is clear that most of the research in these studies was focused on the ‘consumption’ part of the FSC, followed by studies that provided a ‘total’ estimate or ‘distribution’ estimate.
Table 2 presents an overview of the FLW weight data points across the FSC, along with the results of a descriptive statistical analysis. There is a high degree of variability in the estimates across all parts of the FSC.
|Agricultural production||Postharvest handling and storage||Processing and packaging||Distribution||Consumption||Total|
There were three studies that developed estimates for these two parts of the FSC. Brautigam et al. (2014) adopted multipliers developed by Gustavsson et al. (2011) to estimate FLW and used them to develop a EU-wide estimate of 125·8 (kg/capita)/year for these two parts of the FSC, which represented 44% of their overall estimate of EU FLW.
The lower FLW estimates (1·2–3·0 (kg/capita)/year) (both generated using case studies) were from a single upstate New York State county (Griffin et al., 2009) and from a single country (Italy) (Garrone et al., 2014). The highest estimate (70·0 (kg/capita)/year) (based on Eurostat data and data from national sources) was from the EU (EC, 2010) and was similar to a UK-wide estimate (61·7 (kg/capita)/year), which used food production data to develop an estimate of FLW (WRAP, 2013b).
The lower estimates of FLW were from the UK (WRAP, 2013b) (7·1 (kg/capita)/year). Estimates for the USA (Kantor et al., 1997) were calculated using US Department of Agriculture Economic Research Service (USDA ERS, 2014) data and previous studies that estimated FLW generation. The higher estimates identified in the review were from the US state of Hawaii (Okazaki et al., 2008) (240·9 (kg/capita)/year) (survey). The ‘agricultural production’ to ‘distribution’ parts of the FSC represent pre-consumer FLW estimates. Estimates ranged from 173·7 to 185·0 (kg/capita)/year (Brautigam et al., 2014; Gustavsson et al., 2011).
In this review, the most data points and highest average annual per capita FLW generation (114·3 (kg/capita)/year) are found within the ‘consumption’ part of the FSC. The lower estimates came from a rural area (Lebersorger and Schneider, 2011) (18·8 (kg/capita)/year) (based on waste audit of curb-side waste samples) and Finland (Katajajuuri et al., 2014; Silvennoinen et al., 2014) (23 (kg/capita)/year) (based on diary study that measured edible food waste only). The highest estimate came from Canada (308 kg/capita) (Abdulla et al., 2013) (using Statistics Canada and World Bank data and applying USDA waste factors).
In some cases, estimates of ‘total’ FLW included the sum of FLW estimates from the different parts of the FSC, but in most cases this was characterised by standalone estimates of all FLW generated. On average it was estimated that 198·9 (kg/capita)/year of FLW is generated. The lower estimates came from a single upstate New York State county (Griffin et al., 2009) (95·6 (kg/capita)/year) (case study) and the USA (USEPA, 2014) (116 (kg/capita)/year). The highest estimate was based on studies of North America and Oceania (300 (kg/capita)/year) (Gustavsson et al., 2011).
Table 3 presents a summary of average FLW for the independent variables: (a) scope of FLW (inedible/edible or edible), (b) geography (Europe or North America) and (c) study methodologies (direct or indirect measurement), for the dependent variables: ‘distribution’, ‘consumption’ and ‘total’. The FLW differences of greatest magnitude were related to the variables geography and study methodologies. For geography, the average FLW estimates for ‘distribution’ were higher and ‘consumption’ significantly higher (p = 0·003) for North America than Europe although for ‘total’ they were similar. For study methodologies, direct measurement methods (n = 3) resulted in higher FLW estimates for ‘distribution’. Indirect measurements resulted in significantly higher FLW estimates for ‘consumption’ (p = 0·030) and ‘total’ (p = 0·035), although it should be noted that for there were only two studies that employed direct study methodologies to estimate ‘total’ FLW.
|Scope of FLW||Geography||Study methodologies|
The main risk of bias within the studies reviewed relates to how data were collected and analysed. For studies using indirect measurement the risk of bias relates to what is actually being measured. The general approach is to use estimates of the amount of food that is produced and apply estimates of FLW along the various parts of the FSC. These estimates are generally old (some date to the mid-1970s) and it is not clear how they were developed. The risk of bias, across studies, occurs when this same methodology or variation thereof is used by a number of researchers. Any under- or over-estimation could also be manifested in these studies.
For studies using direct measurements involving the collection and sorting of waste samples, the risk of bias relates to the representativeness of the samples (e.g. number and location of households from which waste is collected for composition analysis) and the meaningfulness of resultant extrapolations. For direct food waste measurements that involve self-reporting (e.g. using diaries) the risk of bias relates to the lack of blinding. Mindful that different interventions are not being assessed in these studies; bias can be manifest as performance and detection bias. There appeared to be little risk of bias across these studies because the methodologies did not appear to be shared.
The results showed that there is a greater tendency to measure FLW at or just before it gets to the consumer and that these yield the highest estimates; that there is considerable variability in the data; that North American estimates are generally higher that European ones; and that indirect measurements generally result in higher FLW estimates. The results are far from unequivocal and this exercise confirms the noted concerns about the current state of FLW data and methodological issues (Abdulla et al., 2013; Gustavsson et al., 2011; Langley et al., 2009; Parfitt et al., 2010).
A key methodological issue is that FLW estimates are derived both indirectly and directly, yielding results that are difficult to compare. Furthermore, current estimates do not always differentiate edible from inedible FLW or offer much detail on its composition. Indirect estimates are often used to develop global, continent or country wide estimates whereas direct measurements are used for smaller geographic units such as a city or a region (Table 1). Table 4 summarises the differences between indirect and direct FLW measurement.
|Indirect measurement||Direct measurement|
|General approach||Mass flows model used to estimate FLW at a specific FSC position(s)||Direct collection of waste samples to estimate FLW at a specific FSC position(s)|
|Steps to calculate FLW||FLW estimated using five-step process |
|FLW estimated using six-step process |
|Output||Results in general national and transnational FLW estimates||Results in specific and geographically scoped local FLW estimates|
|Using data||Identifies and estimates extent of FLW but offers little ‘empirical’ evidence on where to possibly implement interventions||Identifies and estimates extent of local FLW and offers ‘empirical’ evidence on where to possibly implement interventions|
Indirect estimates have been derived from estimates of how much food is available to be consumed and applying waste factors.
In their widely cited paper, Gustavsson et al. (2011) present global and regional (Europe, North America and Oceania, Industrialised Asia, Sub-Saharan Africa, North Africa, west and central Asia, south and southeast Asia and Latin America) FLW estimates on behalf of the FAO. A mass flows model was used to estimate FLW along each part of the FSC. The production volumes for all commodities were collected from the 2009 FAO Statistical Yearbook (FAOSTAT, 2010a) and the 2007 FAO Food Balance Datasheets (FAOSTAT, 2010b). Allocation and conversion factors were applied to determine food available for human consumption. The authors made assumptions and estimates based on FLW in similar regions and other factors where there were data gaps. There is insufficient data presented on how estimated/assumed FLW percentages across the FSC of each region were derived.
Countrywide estimates of American FLW data were developed by the USDA ERS (2014) starting with Kantor et al. (1997) to the most recent estimates by Buzby et al. (2014). USDA ERS (2014) provides some insights into how these estimates were developed, how they have been improved, as well as their limitations. The basis of these estimates was derived from the ERS’s Loss-Adjusted Food Availability (LAFA) data series. It uses ERS's food availability data, which estimates the annual production of more than 200 foods and then adjusts for food spoilage, plate waste and other losses at different stages along the food supply and consumption chain, to more closely approximate actual consumption. Food loss coefficients were gathered from published reports and discussions with commodity experts. Loss assumptions were based on data and studies from the mid-1970s onwards.
As described in USDA ERS (2014), attempts have been made to improve the underlying assumptions used to make estimates of FLW. Estimates of primary level (i.e. farm to retail weight) FLW were updated through industry interviews and research. Some retail FLW estimates were updated by comparing supplier shipment data with point of sales data at large national supermarket chains and supplemented with qualitative information from retail contacts. Consumer-level loss estimates for cooking loss and food loss from edible food were updated through: (a) a review of the literature, (b) a small set of restaurant interviews, (c) a numerical estimation method to calculate consumer-level food loss estimates using Nielsen Homescan data (food purchase data) and (d) the dietary intake component of the National Health and Nutrition Examination Survey (food consumption data). In 2012, ERS used the ‘best estimate’ of these consumer FLW estimates, but continued to use the LAFA dataset when updated data was unavailable.
The LAFA dataset does not measure actual consumption or quantities ingested because it is not based on direct observations of individual intake. Furthermore, LAFA does not identify where, along the FSC, FLW is created. Ultimately, these estimates function as a proxy of per capita consumption and FLW generation along the FSC.
Abdulla et al. (2013) used reports published from Statistics Canada and the World Bank to calculate FLW from food available for consumption. Statistics Canada used ‘waste factors’ provided by the USDA (Statistics Canada, 2010) to estimate FLW at the consumption part of the FSC. Canada does not have the data required to empirically quantify FLW at each point in the FSC. Abdulla et al. (2013) recommends launching a replicable pilot study in an area or region to measure FLW across the FSC and then replicate elsewhere in Canada.
Direct measurements of FLW are taken where it is possible to collect and sort waste samples. To date, this has tended to occur with post-consumer waste and specifically at the ‘consumption’ part of the FSC.
The UK's Waste Reduction Action Programme (WRAP, 2009, 2013a, 2013b) has developed a number of solid and liquid ‘consumption’ estimates of FLW. This relies on waste management tonnage data collected by local authorities, the results of waste composition analysis (i.e. waste audits) and use of kitchen diaries (i.e. FLW tracking by residents). It multiplies the percentage of FLW in the waste stream with the total amount of waste generated and supplements this with waste composition data and with kitchen diary data (i.e. which also included detail on pet feeding or home composting of FLW).
‘Processing and packaging’ and ‘distribution’ data were obtained from various industry surveys (i.e. by the Environment Agency, Department for Environment, Food and Rural Affairs (Defra), Food and Drink Federation), business reports on waste (to satisfy permitting requirements) and using business register data to estimate and extrapolate waste generation. FLW estimates were developed from a variety of datasets, because individual datasets did not provide a complete set of information.
Some studies used direct measurement to either estimate and/or test methods to estimate FLW generation (Katajajuuri et al., 2014; Langley et al., 2010; Lebersorger and Schneider, 2011; Okazaki et al., 2008; Parizeau et al., 2015; Silvennoinen et al., 2014) or to assess the impacts of FLW reduction interventions (Bernstad et al., 2012, 2013). Studies typically included a weight-based assessment of FLW and in some cases included diary studies or surveys. For other studies the focus was on estimating the composition of the overall waste stream of which FLW was a component (Defra, 2010; Edjabou et al., 2015; Matsuto and Ham, 1990). The challenge with the direct measurement of FLW is the ability to extrapolate the resultant data. WRAP (2009, 2013a, 2013b) and Defra (2010) have demonstrated a possible methodological approach.
Given the challenges described for using indirect sampling it is difficult to envision its use for developing anything more than a general picture of the current situation, but not to inform the development of interventions in any meaningful way. Direct measurement of FLW, from collected waste samples, should result in more precise estimates of FLW, at least for the geographic area in which they were completed. This data can be used to inform intervention development and importantly can subsequently be re-measured to assess the efficacy of the intervention.
Additional research is required to better understand FLW generation across all parts of the FSC. To date FLW estimates have focused on ‘consumption’ and to a lesser extent ‘distribution’ and ‘total’ estimates. Additional FLW estimates are required for ‘agricultural production’, ‘postharvest handling/storage’ and ‘processing and packaging’. Although challenging, particularly for ‘agricultural production’, where in-field or in-barn measurements would be necessary, direct measurements should be taken to develop these estimates. Furthermore, FLW estimates from ‘agricultural production’ require a more precise definition to determine when food becomes FLW. For instance, there may be in-field sorting of a crop whereby a portion of the crop is left behind in the field. While this could be construed to be FLW it could also be considered as a source of organic matter necessary to maintain soil tilth.
Additional methodological development to directly measure FLW across the FSC are required as initiated by Langley et al. (2010) and Lebersorger and Schneider (2011) with the possible enhancement of these estimates through mathematical methods (Langley et al., 2009). The basis of these methods should focus on statistically sound weight-based assessment of waste samples, through waste auditing but also provide additional detail on the various food fractions (e.g. bakery, meat) that make up FLW. As well, it should include consideration of edible against inedible FLW because this can help establish the net amount that is recoverable for human consumption. These methods should be tested and refined in small geographic areas (e.g. cities and towns). Ultimately, the results of FLW estimates from small geographic areas can be assembled and extrapolated to develop broader regional (e.g. province) or countrywide FLW estimates.
Efforts to add more rigour to FLW measurement are underway and include The Food Loss and Waste Accounting and Reporting Standard (FLW Protocol, 2016). Although largely neutral on indirect and direct methods, it presents a detailed and systematic framework on how to approach FLW measurement, so that it meets the needs to those measuring FLW and facilitating potential comparison of results.
This review only looked at weight-based FLW estimates and did not consider GHG, calorie- or dollar-based FLW estimates for two reasons. First, this was the most common FLW estimation approach, by far. Second, these other metrics are largely inferred from weight-based estimates and given the above noted FLW estimation challenges this was deemed to be of limited value.
Based on this systematic review of the literature, ‘total’ average FLW in developed countries is estimated to be 198·9 (kg/capita)/year, while average ‘consumption’ related FLW is estimated to be on average 114·3 (kg/capita)/year. There is considerable variability in the various FLW estimates and this is a function of how this data has been collected, and in particular, if the data was collected indirectly or directly. While indirect measurements can provide an overview of the current situation, direct measurements are needed to develop more accurate and precise estimates of FLW, as well as its composition. Ultimately, what is required is the development and testing of a bespoke and statistically sound methodology to directly measure FLW. This method should be developed so that it is replicable and usable in a variety of geographic contexts (e.g. city, region). While global or countrywide FLW estimates developed through indirect data collection are interesting, more scoped estimates will provide improved data from which purpose-built interventions to reduce FLW can be developed and implemented.
The authors thank Dr Jamie Seabrook, Associate Professor, School of Food and Nutritional Sciences Brescia University College, Western University for providing statistical assistance.