Synthesis of Food Waste Compositional Data 2014 & 2015 · 2019-05-09 · Synthesis of food waste...

Final Report

Synthesis of Food Waste

Compositional Data 2014 & 2015

This report describes analysis of waste compositional data and WasteDataFlow

information to produce estimates of food in local-authority collected waste streams

from UK homes in 2014 and 2015.

Project code: CFP302-001

Research date: March 2015 – August 2016 Date: December 2016

Synthesis of Food Waste Compositional Data 2014 & 2015 2

WRAP’s vision is a world in which

resources are used sustainably.

Our mission is to accelerate the move to a

sustainable resource-efficient economy

through re-inventing how we design,

produce and sell products; re-thinking

how we use and consume products; and

re-defining what is possible through re-

use and recycling.

Find out more at www.wrap.org.uk

CSC103-001

WRAP, 2016, Quantification of food surplus, waste and related materials in the grocery

supply chain

Written by: Eric Bridgwater (Resource Futures) and Tom Quested (WRAP)

Front cover photography: Examples of food waste (WRAP)

While we have tried to make sure this report is accurate, WRAP does not accept liability for any loss, damage, cost or expense incurred or

arising from reliance on this report. Readers are responsible for assessing the accuracy and conclusions of the content of this report.

Quotations and case studies have been drawn from the public domain, with permissions sought where practicable. This report does not

represent endorsement of the examples used and has not been endorsed by the organisations and individuals featured within it. This material

is subject to copyright. You can copy it free of charge and may use excerpts from it provided they are not used in a misleading context and you

must identify the source of the material and acknowledge WRAP’s copyright. You must not use this report or material from it to endorse or

suggest WRAP has endorsed a commercial product or service. For more details please see WRAP’s terms and conditions on our website at

www.wrap.org.uk

http://www.wrap.org.uk/

http://www.wrap.org.uk/

Synthesis of food waste compositional data 2014 & 2015 3

Executive summary

Background and aim

A key objective for WRAP is to support and enable efforts to reduce the quantity and

environmental impact of household food waste across the UK, working in partnership

with a wide range of organisations including local authorities (LAs), grocery retailers,

food manufacturers and community groups. It is therefore necessary to quantify

household food waste in order to track progress and to understand the impact of work

to date. WRAP most recently published estimates for household food waste in 2013,

which related to waste produced in 20121.

The aim of the current project was to produce estimates of food waste2 collected by (or

on behalf of) LAs from UK homes in 2014 and 2015 using the most recent data available.

Food waste collected from households by LAs makes up around two-thirds of the total

food thrown away by households, with the remaining one-third going down the drain,

home composted or fed to animals. For estimates of these other disposal routes, see

Household Food Waste in the UK, 20153.

These results have been compared to estimates from previous years. Estimates were

also produced for England, Scotland and Wales, as well as for London. However, it was

not possible to produce estimates for Northern Ireland due to lack of waste

compositional data. In addition, the proportion of total food waste captured by

collections targeting this material was also calculated.

Method

Data from waste composition studies was collated and analysed alongside the most

recently available data about Local Authority household waste collections from

WasteDataFlow, in order to arrive at estimates of LA-collected household food waste in

the UK.

For the purposes of this project, LA-collected household food waste was assumed to

include food in the following household waste streams:

● Kerbside residual: (i.e. ‘general’ waste)

● Kerbside collections targeting food waste: collections of either separate food

waste or mixed garden and food waste

● Kerbside dry recycling: food waste contamination of kerbside dry recycling

collections from households

● Household Waste Recycling Centre (HWRC) residual waste

1 http://www.wrap.org.uk/content/household-food-and-drink-waste-uk-2012

2 Within this report, ‘food’ is used as a short hand for ‘food and drink’.

3 WRAP (2016) http://www.wrap.org.uk/hhfw2015

http://www.wrap.org.uk/content/household-food-and-drink-waste-uk-2012

http://www.wrap.org.uk/hhfw2015


When assessing how the amount of household food waste collected by LAs had changed

over time, the main comparison was between 2015 estimates (the most recent

estimates produced in this study) and estimates produced in previous studies. However,

estimates for 2014 were also produced during this study and these are also included in

this report for completeness.

Data on food waste arisings in kerbside residual – the most important waste stream to

consider for this project – was obtained for 116 LAs across the UK (out of a total of

approximately 400) for the 2015 estimate. The LAs with compositional studies were

reasonably representative of the UK and the sample was stratified by presence of

collections targeting food waste to account for differences in collection systems between

the sample and the population of LAs.

Results

The estimated amount of household food waste collected by LAs in the UK in 2015 was

4.90 million tonnes. This is equivalent to 178 kg per household or 75 kg per person

(Table ES1).

Table ES1: Estimates of household food waste collected by local authorities in the UK,

2015, tonnes

Waste Stream

Food waste arisings in 2015

Thousand

tonnes

kg per

household kg per person

Kerbside residual 4,117 150.0 63.2

Kerbside collections targeting

food waste 639 23.3 9.8

Kerbside dry recycling

(contamination) 80 2.9

1.2

Household Waste Recycling

Centre residual 60 2.2

0.9

TOTAL 4,896 178.4 75.2

95% confidence interval ±134 ±4.9 ±2.1

The vast majority of this food waste (4.12 million tonnes; 84% of the total) was found in

the kerbside residual waste stream, with most of the rest (0.64 million tonnes; 13% of

the total) in kerbside collections targeting food waste.

For the UK, per capita food waste arisings were much lower in 2015 compared to 2007

(the first year with estimates that are sufficiently similar to allow comparison). As shown

in Figure ES1, the majority of the reduction occurred between 2007 and 2010. Since

2012, there has been no significant evidence of further reductions.


Figure ES1: Estimated arising of local authority collected household food waste in the

UK 2007 to 2015, kg per person (±95% confidence interval)

Results cover food waste in kerbside residual and collections targeting food waste for treatment, but

exclude minor waste streams (HWRC residual and contamination of dry recycling).

The 2015 estimate for the average amount of household food waste per person for

England was not significantly different to that for the UK as a whole. Similarly, the

amount in London was not significantly different to the rest of England.

In Wales, there was a reduction of 12% in the total amount of food waste per person

generated from 2009 to 2015. In 2015, the amount wasted per person was significantly

lower than for England by around 9%. However, methodological issues may have

influenced these results to a small degree (see Appendix A). Potential reasons for this

difference between Wales and England are discussed in Household Food Waste in the UK,

20154.

The most recent estimate for household food waste in Scotland (2014) was not

significantly different from that for the rest of the UK for the same year. The average

amount of HHFW per person was 6% lower in 2014 compared to 2009; however, this

difference was not statistically significant.

The capture rate – the proportion of LA-collected household food waste found in

collections targeting food waste – increased between 2007 and 2015 from around 2% to

13% (Figure ES2). However, this means that the proportion of food waste in collections

targeting food waste in 2015 was still relatively low, with the vast majority – 87% – being

found in the residual waste streams (or contamination of dry recycling).


50

55

60

65

70

75

80

85

90

95

100

2007 2008 2009 2010 2011 2012 2013 2014 2015

Food waste (kg per person per year)



Figure ES2: Trend in UK capture rate of household food waste (±95% confidence

interval)

In LAs that had a wide coverage of separate food waste collections, the capture rate was

around one-third (35%) with two-thirds still going into the residual waste. This suggests

that there is the potential to increase the use of these collections where they are

present.

The capture rate in Wales in 2015 was 48%, much higher than the rest of the UK. This is

due to very high proportion of households with a collection targeting food waste and

greater use of these collections compared with the rest of the UK.

The regression analysis in this report suggests that there are differences in results from

the various methodologies and approaches used by different contractors undertaking

waste compositional analysis, although it was not possible to identify the exact cause.

Therefore, it is important that contractors employ methodologies that are as consistent

as possible when carrying out waste compositional studies. Zero Waste Scotland has

published some useful guidance aimed at achieving a higher level of consistency in

waste composition methodologies5. It is worth the relevant actors in the rest of the UK

considering whether a standardised approach would be practical, as such a change

would benefit many studies (such as this one) that use waste compositional data.

5 Guidance on the Methodology for Waste Composition Analysis, Zero Waste Scotland, 2015


Contents

1.0 Introduction ........................................................................................................... 10

1.1 Background ........................................................................................................... 10

1.2 Definition of LA collected household food waste ........................................... 11

2.0 Methodology .......................................................................................................... 13

2.1 Introduction .......................................................................................................... 13

2.2 Compositional datasets included in the new pooled estimates ................... 13

2.2.1 Collation of compositional datasets ...................................................... 13

2.2.2 Selection criteria ....................................................................................... 14

2.2.3 Inclusion of single phase waste compositional analyses ................... 15

2.2.4 Use of individual waste compositional analyses phases ................... 15

2.2.5 Coverage of compositional datasets in the current study ................. 16

2.2.6 Food packaging in waste compositional analysis data ....................... 16

2.2.7 Checking data points ............................................................................... 17

2.3 WasteDataFlow (WDF) time period used for analysis ..................................... 17

2.4 Standard method for calculating household food waste .............................. 18

2.4.1 Standard method: kerbside residual ..................................................... 18

2.4.2 Standard method: kerbside collections targeting food waste .......... 19

2.4.3 Standard method: kerbside dry recycling ............................................ 20

2.4.4 Standard method: HWRC residual ......................................................... 20

2.5 Household and population data ........................................................................ 20

2.6 Confidence intervals ............................................................................................ 21

2.7 Interpretation of trends in food waste arisings ............................................... 22

3.0 Results .................................................................................................................... 24

3.1 UK results .............................................................................................................. 24

3.1.1 UK arisings in 2014 ................................................................................... 24

3.1.2 UK arisings in 2015 ................................................................................... 25

3.1.3 UK trends in LA collected household food waste ................................ 27

3.2 Arisings in England, Wales, Scotland and London .......................................... 31

3.2.1 Arisings in England ................................................................................... 31

3.2.2 Arisings in Wales ....................................................................................... 32

3.2.3 Arisings in Scotland .................................................................................. 33

3.2.4 Arisings in London .................................................................................... 35


3.3 Capture rates for food waste ............................................................................. 35

4.0 Conclusions ............................................................................................................ 38

Appendix 1: Regression analysis of influences on food waste arisings .................... 39

A1.1 Regression results: percentage of kerbside residual which is food waste .. 40

A1.2 Regression results: food waste at the kerbside per household per year .... 47

Appendix 2: Coverage Assessment ................................................................................ 51

A2.1 Coverage by levels of deprivation ..................................................................... 51

A2.2 Coverage by nation .............................................................................................. 54

A2.3 Coverage by population density ........................................................................ 55

A2.4 Coverage by collection system ........................................................................... 56

A2.5 Coverage by period and season ........................................................................ 58

A2.6 Number of samples included in each set of estimates .................................. 60

Appendix 3: Detailed method for estimating food waste in collections targeting

food waste ........................................................................................................................ 61

Appendix 4: Scaling results from sample to national level – detailed method ....... 69

A4.1 Accounting for differences between sample and population in the standard

method .............................................................................................................................. 69

A4.1.1 Stratification .............................................................................................. 69

A4.1.2 Adjustment for differences in yield ....................................................... 70

A4.2 Methodological details for the ‘standard’ method .......................................... 71

A4.3 ‘Alternative’ method for calculating UK estimates .......................................... 72

A4.4 Results and discussion of alternative and standard methods ...................... 73

A4.5 Current and historic methodologies (comparison of estimates presented in

this report) ........................................................................................................................ 75

A4.6 Current and historic methodologies (comparison of previously published

estimates) .......................................................................................................................... 76

Appendix 5: Single-year estimates................................................................................. 78

Appendix 6: Uncertainty ................................................................................................. 81

Appendix 7: Peer review report ..................................................................................... 83


Glossary

● Capture rate – the amount of household food waste captured in kerbside collections

targeting food waste as a percentage of all household food waste (see §3.3)

● Defra – Department for Environment, Food and Rural Affairs

● HWRC – Household Waste Recycling Centre (also known as a civic amenity site)

● LA – Local authority

● LA collected household food waste – food waste found in the following waste

streams: Kerbside residual: (i.e. ‘general’ waste collected from the households);

Kerbside collections targeting food waste, which includes collections from

households of either separate food waste or mixed garden and food waste; Kerbside

dry recycling (food waste contamination of kerbside dry recycling collections from

households); HWRC residual waste

● HH - household

● ‘Pooled’ estimates – estimates of food waste that combine waste-compositional

information from a 24-month period around the ‘target’ year. The main advantage of

the pooled estimates is that they use a relatively large number of studies, reducing

the uncertainty around each estimate c.f. ‘single-year’ estimates and see §2.2

● ‘Single-year’ estimates – estimates of food waste that only use waste-compositional

information from the ‘target’ year. These are used as a check for the ‘pooled’

estimates (see above and Appendix 5 for more details).

● WDF – WasteDataFlow, a reporting system for waste collected by local authorities in

the UK (http://www.wastedataflow.org/)

● WRAP – Waste and Resources Action Programme

Shorthand used in this report for previous related reports:

● The 2012 study: Synthesis of Food Waste Compositional Data 2012

http://www.wrap.org.uk/sites/files/wrap/hhfdw-synthesis-food-waste-composition-data.pdf

● The 2010 study: Synthesis of Food Waste Compositional Data 2010

www.wrap.org.uk/sites/files/wrap/Synthesis%20of%20Food%20Waste%20Compositional%20Da

ta%202010%20FINAL.pdf

● Defra WR0119: Municipal Waste Composition: Review of Municipal Waste Component

Analyses

http://randd.defra.gov.uk/Default.aspx?Module=More&Location=None&ProjectID=15133

● Defra EV0801: Defra EV0801 National compositional estimates for local authority

collected waste and recycling in England, 2010/11


Acknowledgements

We would like to thank all of the organisations and companies that provided waste

compositional data to assist with this project. We are also grateful to the peer reviewer

Andrew Davey (WRc plc), as well as colleagues in WRAP, Defra, Zero Waste Scotland and

Welsh Government who reviewed the report. Their constructive comments led to many

improvements.

http://www.wastedataflow.org/


http://www.wrap.org.uk/sites/files/wrap/Synthesis%20of%20Food%20Waste%20Compositional%20Data%202010%20FINAL.pdf




WRAP – 10

1.0 Introduction

1.1 Background

A key objective for WRAP is to support and enable efforts to reduce the quantity and

environmental impact of household food waste across the UK, working in partnership

with a wide range of organisations including local authorities (LAs), grocery retailers and

food manufacturers and community groups. It is therefore necessary to quantify

household food waste in order to track progress and to understand the impact of work

to date. Prior to the current report, the most recent estimates for household food waste

published by WRAP were for 2012, published in 20136.

The aim of this project was to produce estimates of food waste7 collected by (or on

behalf of) LAs from UK homes using the most recent data available. These results were

compared to estimates from previous years. Estimates were also produced for England,

Wales and Scotland, as well as for London. However it was not possible to produce

estimates for Northern Ireland due to lack of waste compositional data. In addition, the

capture rate (the proportion of food waste captured by collections that target this

material) was also calculated.

Alongside this report, a further report (Household Food Waste in the UK, 20158) has been

published, which includes estimates of the disposal routes for food waste in the home

not covered in the current report (sewer, home composting and fed to animals). In

addition, it provides background information relating to food waste to help understand

the trends seen in this report.

The main body of the current report describes the analysis and results relating to

estimates for 2014 and 2015. However it is important not to directly compare the 2014

and 2015 estimates, as there was significant overlap in the waste compositional studies

used to produce each set of estimates. Valid comparisons can be made either between

the 2015 and earlier estimates (e.g. for 2012 and 2010, discussed below) or between the

2014 estimates and earlier estimates.

The results are based on data from WasteDataFlow (WDF) – the UK’s repository for data

relating to waste collected by LAs – and waste compositional analyses conducted by

individual LAs, waste authorities and other commissioning bodies. The definition of food

waste used in this study is in §1.2, the methodology is found in §2.0, and the results in

§3.0.

This project is similar to the study Synthesis of Food Waste Compositional Data 20129

(hereafter referred to as ‘The 2012 study’), which produced estimates of food waste

collected by LAs from UK households for 2012. That project was based on the Synthesis

6 http://www.wrap.org.uk/content/household-food-and-drink-waste-uk-2012

7 Within this report, ‘food’ is used as a short hand for ‘food and drink’.

8 WRAP (2016): http://www.wrap.org.uk/hhfw2015

9 http://www.wrap.org.uk/sites/files/wrap/hhfdw-synthesis-food-waste-composition-data.pdf





of Food Waste Compositional Data 201010 (hereafter referred to as ‘The 2010 study’), which

performed the same exercise for 2010.

The 2014 and 2015 estimates – and estimates for 2012 and 2010 – combine information

from a range of years around a ‘reference’ year: these are referred to in the report as

‘pooled’ estimates. The main advantage of the pooled estimates is that they each use a

relatively large number of studies, thus reducing the uncertainty in each estimate. In

addition to these pooled estimates, estimates of food waste arisings from 2006 to 2015

based solely on information from each year are also calculated; these are referred to as

‘single-year’ estimates (Appendix 5). The single-year estimates were generated as a

check of the pooled estimates, specifically to determine whether any spurious effects

resulted from using data from a range of years.

1.2 Definition of LA collected household food waste

The focus of this study is household food waste collected by local authorities (LAs). For

the purposes of this report, this was defined as food waste which is likely to have been

generated from within the household: i.e. from food that was purchased (or otherwise

taken into the home) or home grown and then either some or all of it disposed of in a LA

collection. It includes food found in kerbside waste streams and those associated with

household waste recycling centres (HWRCs).

Given the above, LA collected household food waste was classified, for the purposes

of this study, as that found within the following streams:

● Kerbside residual: (i.e. ‘general’ waste collected from the households).

● Kerbside collections targeting food waste: this includes collections from

households of either separate food waste or mixed garden and food waste. This

collected material is diverted from either landfill or energy from waste. The

treatment process usually consists of in-vessel composting, although a growing

fraction of this material is sent to anaerobic digestion and other treatment methods.

For the purposes of brevity, this waste stream is hereafter referred to as ‘collections

targeting food waste’.

● Kerbside dry recycling: food waste contamination of kerbside dry recycling

collections from households.

● Household Waste Recycling Centre (HWRC) residual waste.

Negligible quantities of food waste collected for treatment were reported for HWRCs in

the UK (2,898 tonnes reported in WDF 2014/15) and therefore food waste collected for

treatment from HWRCs has not been considered in this study.

The 2012 study made a separate estimate of food waste arisings in street sweepings

(91,000 tonnes in the UK in 2012)11. Due to lack of recent waste compositional data for

street sweepings, an updated estimate has not been produced for this study. Moreover,

estimates for food waste in street sweepings were excluded from the 2012 estimate of

10 www.wrap.org.uk/sites/files/wrap/Synthesis%20of%20Food%20Waste%20Compositional%20Data%202010%20FINAL.pdf

11 Synthesis of Food Waste Compositional Data 2012, WRAP 2013, Section 4.4.



household food waste, as Defra has released guidance indicating that these are not

classified as “waste from households” (although they are defined as ‘household waste’)12.

Therefore, they were excluded from the estimates presented in this report as it was

assumed that the food waste found in these streams was less likely to have come from

food that had entered a household.

Some previous studies of LA-collected food waste included an estimate of food waste

arising in fines (small particulate material) in residual waste streams. An estimate of

food waste in fines was omitted from this study due to a lack of robust data. To allow

robust comparisons over time, retrospective adjustment has been made of the historic

data to remove the estimate relating to fines.

The estimates in this report quantify food waste found in the above waste streams. This

includes both edible material (e.g. bread, the flesh of fruit, meat) and inedible material

associated with food (e.g. egg shells, citrus peel, meat bones). This approach is

consistent with the recent definitional framework for food waste developed by the

European-funded FUSIONS project13 and the Food Loss and Waste Standard developed by

the World Resources Institute14. This is also consistent with previous estimates of food

waste produced by WRAP, which included ‘avoidable’, ‘possibly avoidable’ and

‘unavoidable’ food waste. An estimate for the split between these fractions is provided in

Household Food Waste in the UK, 2015.

12 2013/14 National Statistics on Local Authority Collected Waste Management in England, Methodological Summary, Defra,

https://www.gov.uk/government/uploads/system/uploads/attachment_data/file/388983/mwb_201314_stats__Methodology_summ

ary.pdf

13 http://www.eu-fusions.org/

14 http://flwprotocol.org/

https://www.gov.uk/government/uploads/system/uploads/attachment_data/file/388983/mwb_201314_stats__Methodology_summary.pdf

https://www.gov.uk/government/uploads/system/uploads/attachment_data/file/388983/mwb_201314_stats__Methodology_summary.pdf

http://www.eu-fusions.org/

http://flwprotocol.org/


2.0 Methodology

2.1 Introduction

This section describes the methodology used in this report to quantify food waste

collected by (or on behalf of) local authorities (LAs) from UK homes in 2014 and 2015.

This methodology is similar to that used in previous synthesis studies (e.g. the 2010 and

2012 studies) and differences between the current and historic methodologies are

discussed in §2.7.

The method involves analysing waste compositional data alongside information on the

amount of material in various waste streams from WasteDataFlow (WDF, which records

the amount of material collected in various waste streams directly from LAs).

Many LAs in the UK commission studies to examine the waste they collect or that is

collected on their behalf by waste contractors. These compositional studies classify the

waste into different materials, usually between 40 and 70 categories depending on the

detail required and the amount of waste to be sorted. Food waste is generally one of

these categories and often this is further subdivided: e.g. home compostable / non-

home-compostable; packaged / non-packaged. These studies typically collect and sort

the waste from 135-250 households.

In this study, data was collated from a range of compositional analyses commissioned

by different LAs. This data was analysed alongside information from WDF, which

provides the total weight of material collected within a given waste stream. LAs provide

this information to the organisations overseeing WDF and, once checked, the data is

published and can be used in research (such as these ‘syntheses’). This can negate the

need for all waste streams to be sampled in waste compositional analyses. For instance,

LAs with separate food waste collections will usually record the amount in the correct

category in WDF and therefore do not need further measurement to determine the

quantity of food in this waste stream. (For comments on contamination in food waste

collected for treatment, refer to §2.4.2.)

However, the type of materials present in the residual waste streams are not recorded

in WDF. To obtain an estimate of the amount of food in each residual stream, the

percentage of food waste in a stream (as measured by the LA compositional analysis

studies) needs to be applied to the total weight of that residual waste stream. This

section and the related appendices describe how these two sources of data were

combined and the reasons for the methodological choices made.

2.2 Compositional datasets included in the new pooled estimates

As discussed in §1.1, the new estimates presented here are pooled estimates for 2014

and 2015. This section discusses the compositional datasets included for these

estimates. Other estimates were calculated as a check for these pooled estimates and

these are presented in Appendices 4.3 and 5.

2.2.1 Collation of compositional datasets

During the course of this project, organisations and companies commissioning and

undertaking waste compositional analyses were contacted to obtain data. Data was

provided by Resource Futures, M·E·L, Zero Waste Scotland, as well as directly from a

number of other contractors and LAs. Of the waste compositional analyses known to be


undertaken from 2013 to 2016, data from the vast majority was obtained for this study.

A small number of LAs did not give permission for their data to be used in this study,

and one contractor undertaking waste compositional analyses did not respond to

requests to share data.

2.2.2 Selection criteria

Selection criteria – 2014 pooled estimates

The target period for the 2014 pooled estimates was for the calendar year 2014. The

datasets collated for the current study were assessed on two selection criteria to

determine whether to include them:

● The date of the fieldwork. For the pooled estimate, studies undertaken between

April 2013 and March 2015 were included. There is a trade-off between the number

of studies included and getting studies as close to the target timespan (calendar year

2014) as possible. By selecting these two financial years, it meant that there was no

overlap with studies carried out for the pooled estimate of 2012, whilst minimising

the number of studies collated but not used.

● Whether samples were reasonably representative of the socio-demographic

profile of the relevant LA. Only one waste compositional analysis was not used in

this study. The excluded study was the second phase of an analysis from a LA for

which the sample was socio-demographically representative of the LA, but the

second phase was carried out after a service change had taken place, towards the

very end of the target timespan (2014). Therefore, it was considered that the first

phase of data (carried out before the service change was implemented) would be

more representative of the LA during the target period for this study, and only this

first phase was used.

Selection criteria – 2015 pooled estimates

The target period for the 2015 pooled estimates was for the calendar year 2015. The

datasets collated for the current study were assessed on two selection criteria to

determine whether to include them:

● The date of the fieldwork. For the pooled estimate, studies undertaken between

April 2014 and March 2016 were included. As discussed above, there is a trade-off

between the number of studies included and getting studies as close to the target

timespan as possible. The precedent of using two financial years was set when

producing the 2014 pooled estimates.

● Whether samples were reasonably representative of the socio-demographic

profile of the relevant LA. Waste compositional analyses that included some form

of socio-demographic stratification15 were included. This criterion was met by all the

datasets collated for the 2015 pooled estimates.

Given the selection criteria above, some of the datasets collated for the 2014 pooled

estimates were also used in calculating the 2015 pooled estimates. Due to this overlap,

15 Examples included MOSAIC run by Experian and the Output Area Classification run by the UK Office for National Statistics

(ONS). The latter classifies 41 census variables into a 3 tier classification of 7, 21 and 52 groups.


estimates for 2014 and 2015 should not be compared with each other. However, valid

comparisons can be made either between the 2015 and earlier estimates (e.g. 2012 and

2010) or between the 2014 estimates and earlier estimates.

2.2.3 Inclusion of single phase waste compositional analyses

The current study used data from single-phase (fieldwork conducted over a single time

period) and multi-phase waste compositional analyses where studies were conducted

over two or more different time period to account for possible seasonal variation. This is

a similar approach to both the 2010 and 2012 studies. This is in contrast to the

compositional estimates for 2007 produced from the Defra WR0119 study, which only

included multi-phase waste compositional analyses for kerbside compositional datasets,

due to so many datasets being available for this study. The inclusion of only multi-phase

studies for the Defra WR0119 study was intended to control for seasonal variation in

arisings of various components in the kerbside stream.

The 2010 and 2012 studies both took the decision to include single-phase studies, due

to a drop off in the number of compositional studies that had been commissioned since

the Defra WR0119 study. For the 2010 study, an assessment was made of seasonal

variations in food waste through analysing separate phase data from multi-phase

studies used in the Defra WR0119 study. This assessment concluded that although some

seasonal variation was apparent in food waste arisings in kerbside waste, this was

relatively minor. For the current study, an updated analysis of seasonal variation in food

waste arisings was been carried out using more recent data (Appendices 1 & 2).

2.2.4 Use of individual waste compositional analyses phases

For the current study, multi-phase waste compositional analyses were split into their

individual phases. The most important advantage of this approach was that it increased

the number of data points for producing UK estimates, thus increasing the effective

sample size in the analysis.

The published 2010 study did not split multi-phase data into constituent phases in this

manner, and did not give greater weighting to the more comprehensive multi-phase

waste compositional analyses, which would have provided a better reflection of the

quality of information contributed by such waste compositional analyses to the sampled

dataset, in comparison to single-phase studies16. The procedure of splitting multi-phase

studies into constituent single phases for the current study overcame this problem and

effectively provided greater weighting to multi-phase studies. However, this approach

meant that the dataset for the current study included studies that sampled households

in the same areas (i.e. from the same collection rounds, although not necessarily exactly

the same households). Given that some households may have been sampled more than

once, this could have led to a small degree of non-independence between phases,

resulting in estimates of precision that are slightly optimistic (e.g. the width of the

calculated confidence intervals could be a slight underestimate).

16 In contrast to the previously published results for 2010, the 2010 results presented in this report do have multi-phase waste

compositional analyses split into their constituent phases. For a discussion of the differences in methodology, see §2.7.


However, it was considered that the advantages of using a greater number of data

points through splitting multi-phase datasets outweighed the disadvantages. The

splitting of multi-phase waste compositional analyses into single-phase data also

provided more flexibility in selecting waste compositional analysis data within the time

frame of the study. Furthermore, splitting multi-phase datasets had only a small effect

on the 2012 results.

2.2.5 Coverage of compositional datasets in the current study

The coverage of the waste compositional analyses collated was assessed against a

number of factors, listed below:

● Nation within the UK

● Region within England

● Socio-economic factors (e.g. deprivation level)

● Rurality factors (e.g. population density)

● Presence and type of collection targeting food waste

● Frequency of residual collection

● Season / month fieldwork waste undertaken.

For each of these factors, coverage was assessed in terms of how representative the LAs

with usable waste-compositional-analysis data (i.e. the sample) was compared to all LAs

in the UK (i.e. the population). If there was a substantial mismatch between the sample

and the population, further analysis was conducted to assess whether this mismatch

was materially important – i.e. if there was evidence that the amount of food waste

varies according to that factor. If there was, then weighting of the sample was

undertaken to adjust for this mismatch and reduce the bias in the results.

For example, in the previous work (the 2010 and 2012 studies), the presence of

collections targeting food waste was associated with a lower percentage of kerbside

residual waste stream that is food waste – food waste being diverted from the residual

waste stream to the collections targeting food waste. Therefore, it was important to

adjust for any mismatch in the presence of collections targeting food waste between the

sample and the population. In contrast, there was little evidence that the average

percentage of kerbside residual waste stream that is food waste varied with, for

example, deprivation, and therefore a mismatch in the deprivation coverage between

the population and the sample did not need to be adjusted for via weighting.

The results of this coverage analysis are presented in Appendix 2 and are informed by

the regression analyses presented in Appendix 1. The conclusion was that – similar to

previous studies – stratification by presence of collections targeting food waste was

important. No other weighting or stratification was required.

2.2.6 Food packaging in waste compositional analysis data

During this study, it was found that different contractors undertaking waste

compositional analyses were handling food found in packaging in different ways. For

example, in some situations ‘packaged food’ was a separate sub-category found within

the ‘food’ category, the result of which was that a small amount of packaging was being

recorded as ‘food’ in these studies. In other situations, this type of packaged food was


separated out into food (recorded under the ‘food’ category) and packaging (recorded

under the appropriate material).

Previous work by WRAP17 has shown that the minority of food waste is found in

packaging – in 2012, 810,000 tonnes out of a total of 4.7 million tonnes collected by LAs;

i.e. 17%. Furthermore, the weight of this packaging is usually much less than the food it

contains. Therefore, the potential degree to which packaging recorded as food

influences the results is relatively small. Nevertheless, it is important to try and minimise

any bias introduced into the results. For this reason, the sorting protocols of contractors

were obtained to understand which datasets may include a bias. Appendix 1 presents

analysis relating to this issue and discusses the issues around adjusting for this effect.

The effects relating to including a sorting category of packaged food were not significant

in the regression models and consequently no changes were made to the methodology

for this study. However, this issue should be monitored closely in future studies of this

nature.

2.2.7 Checking data points

Waste compositional analyses with high or low levels of food waste – i.e. towards the top

and bottom end of the distributions as assessed by kg / household / week or by

percentage of the residual stream – were checked to determine whether there had been

an error at some point in the measurement or data processing. No such errors were

found for either the 2015 of 2014 pooled estimates and therefore no data points were

excluded.

2.3 WasteDataFlow (WDF) time period used for analysis

For the current study, the most recent 12 months (available at the time of the analyses)

of data from WasteDataFlow was used. These differ by nation of the UK and are

described below:

WDF period used – 2014 pooled estimates

The 2014 pooled estimates used WDF data as close to the calendar year 2014 as

possible. For England and Wales, WDF data was not available for all four quarters of

2014 at the time of analysis, and the time period October 2013 – September 2014 was

used. Northern Ireland data for the calendar year 2014 was available. The most recently

available data for Scotland at the time the analysis was carried out was for the calendar

year 2013.

WDF period used – 2015 pooled estimates

The 2015 pooled estimates used WDF data as close to the calendar year 2015 as

possible. For England and Wales, WDF data was not available for all four quarters of

2015 at the time of analysis, and the time period October 2014 – September 2015 was

used. The most recently available data for Northern Ireland data was for April 2014 to

March 2015, and for Scotland was for the calendar year 2014.

17 Household Food and Drink Waste in the UK 2012, WRAP 2013: http://www.wrap.org.uk/content/household-food-and-drink-

waste-uk-2012




The WasteDataFlow periods used in previous studies is documented in §2.7.

2.4 Standard method for calculating household food waste

The 2010 and 2012 studies used two methods for calculating UK household food waste

arisings, a ‘standard’ method and an ‘alternative’ method, with the standard method

being that used to produce the final estimates. The same approach has been applied for

the 2014 and 2015 estimates in this report.

However, the alternative method was also applied, in order to understand how the

choice of calculation method impacts on the results. The alternative method is described

in detail in Appendix 4. They key difference is that the alternative method determines

the total food waste in both the residual and targeted streams for each LA before scaling

this to the population. Both produced similar results.

A similar process was followed to produce results for England, Wales and London. For

Scotland, the results for the standard and alternative method are presented for 2014.

For a comparison between 2009 and 2014, the alternative method was used. This is

because, for the 2009 estimate, there was insufficient data in one of the two strata used

in the standard method to obtain a robust estimate. However, this lack of data did not

affect the alternative method, hence why it has been used. This is discussed in more

detail in a recent Zero Waste Scotland report18.

Separate estimates for either 2014 or 2015 could not be produced for Northern Ireland

due to insufficient waste composition data.

The sum of food waste arisings across these four household waste streams (detailed

below) was calculated to arrive at an estimate of UK arisings of LA collected household

food waste.

The main aspects of the standard method are described below, organised by waste

stream, with some further details for the standard method, and the alternative method,

discussed in Appendix 4.

2.4.1 Standard method: kerbside residual

The weight of food waste in the kerbside residual waste stream was determined by

taking the percentage of kerbside residual waste that is food from each of the waste

compositional studies. From this information, an average percentage was calculated and

multiplied by the amount of kerbside residual waste collected in the UK.

Within previous studies, a number of developments to the method have been made to

improve its accuracy and these have been applied to this study:

● The stratification of the sample and population according to whether LAs have any

collections targeting food waste. Consideration was given to stratifying for other

factors, but either there was no relationship with the percentage of residual waste

that was food, the sample was representative, or the effect on the results was

minimal.

18 Household food and drink waste in Scotland 2014, Zero Waste Scotland (2016)

http://www.zerowastescotland.org.uk/sites/default/files/Household%20Food%20and%20Drink%20Waste%20Estimates%202014

%20Final.pdf

http://www.zerowastescotland.org.uk/sites/default/files/Household%20Food%20and%20Drink%20Waste%20Estimates%202014%20Final.pdf



● Adjustment to account for different yields of collections targeting food waste

between the sample and the population (described in Appendix 4.1.2); and

● Disaggregation of multi-phase studies (see §2.2.4).

To scale from the sample of LAs with waste compositional analyses to the whole of the

UK, the average percentage of food waste in each stratum was calculated from waste

compositional studies from LAs within that stratum. The two strata are:

● LAs that have kerbside collections targeting food waste (e.g. separate food-waste

collections, mixed food and garden collections)

● LAs that do not have kerbside collections targeting food waste

For each stratum, the average percentage of food waste was then multiplied by the

amount of residual waste for all LAs in the stratum (irrespective of whether they had a

waste compositional analysis). This gave a total of food waste in the kerbside residual

waste stream for each stratum. These totals were then added to obtain an estimate for

the whole of the UK.

In the past, a number of slightly different methods were used in the standard method to

scale information from the sample of LAs to the whole of the UK for the kerbside

residual stream. The different methods are discussed in Appendix 4.2, including a

rationale for why the above method has been used.

2.4.2 Standard method: kerbside collections targeting food waste

This section describes the method used to estimate the amount of food waste collected

at kerbside for each LA. A detailed flow diagram for this process is in Appendix 3.

Data sources

The process primarily used WasteDataFlow (WDF) and information collated by WRAP on

waste collections systems used by LAs. Occasionally individual LAs were contacted for

additional data where there were apparent discrepancies between the data reported in

WDF and the scheme information collected by WRAP.

Kerbside organics tonnages were taken from Question 10 of WDF. Tonnages were

presented by LA for each organic material stream. The tonnages have been reported

quarterly, except for Scotland where tonnages have been only reported annually.

The following WDF categories are relevant:

● Waste food only: this category is straightforward as the vast majority is food waste,

and it is also an important indicator of the presence of separate food waste

collections. A small amount of the material collected as food waste consists of

contamination. An analysis of 8 waste composition datasets of separate food waste

collections indicated an average contamination rate of 2.0%. Therefore, for the

current study, a reduction of 2.0% was applied for separately collected food waste to

account for this contamination. Contamination rates were not previously accounted

for in the 2010 and 2012 studies as published, but they have been subsequently

recalculated as part of the current study to allow a robust comparison.

● Mixed garden and food waste: this category is less easily dealt with as the

proportion of food waste in this material cannot be determined directly from the

WDF tonnages. For most LAs the amount of food waste in this mixed stream was


calculated either from the number of households served or from the total waste

collected from this waste stream. More details can be found in Appendix 3.

● Other compostable waste: this category is highly uncertain as it could consist of

garden waste, food waste, cardboard or mixed food and garden wastes. For

authorities reporting tonnages in this category, reference has been made to data

supplied by WRAP on kerbside organics recycling scheme types for LAs; again, more

details can be found in Appendix 3.

WRAP scheme data was used to cross check the WasteDataFlow tonnages. The scheme

data used was financial year scheme data for separate food collections and garden

waste collections (which also includes mixed garden and food waste collections).

On occasion it was necessary to contact some LAs directly to fill data gaps. However, it

was not always possible to make contact with a LA officer who was able to help. In these

cases it was necessary to make the best estimate based on all the information available

for the authority in question. This was the case for a very small number of LAs.

2.4.3 Standard method: kerbside dry recycling

A similar procedure to kerbside residual was carried out for kerbside dry recycling:

where compositional data identified food waste contamination in kerbside dry recycling,

this was used to arrive at an average percentage of the dry recycling waste stream that

was food. This average percentage was multiplied by the total amount of dry recycling

collected in the UK to arrive at an estimate of food waste in kerbside dry recycling. This

waste stream has a minor contribution to the total amount of food waste from

households. The LAs were not stratified for this waste stream due to the small number

of studies available.

2.4.4 Standard method: HWRC residual

A similar procedure as for kerbside residual and kerbside dry recycling was applied to

arrive at an estimate of UK arisings of food waste in HWRC residual. Again, this waste

stream only had a small contribution to the total amount of food waste from

households. The LAs were not stratified for this waste stream due to the small number

of studies available.

2.5 Household and population data

Household and population counts were used to express the amount of food waste per

household and per person. The household-count data was also used within the

alternative method (see Appendix 4.3).

For the 2015 pooled estimates, the following sources were used to determine the

number of households in each LA in the UK:

● England: 2012-based household projections, ONS, December 2015

● Wales: 2011-based estimates of future numbers of households for LAs, StatsWales,

February 2014

● Scotland: Household projections for Scotland, 2012-based, National Records of

Scotland, 2014

● Northern Ireland: Household projections (2012-based), NISRA, March 2015.


The methodology for determining the number of households for the 2014 and 2015

pooled estimates was the same as used in the 2012 study and is described in the

Methods Annex for that study19. The number of households does not affect the standard,

pooled estimates presented in this report, although it does affect some of the other

estimates produced, such as those derived using the alternative method.

Population counts for all nations and periods were taken from the ONS Population

estimates time series dataset (which provides mid-year estimates from 1971 to 2015),

published June 2016.

In the waste industry, results are often presented as waste per household (e.g. kg per

household per week) as the household is the unit from which waste is measured. This is

how results have been presented in previous synthesis reports. However, the number of

people per household can vary over time and between different areas (e.g. between UK

nations). As the amount of food waste has previously been shown to vary with the

number of people in the household (i.e. more food waste in larger households)20, the

results are also presented per person in this report. Additional analysis of existing

datasets suggests that expressing the amount of food waste generated by a household

per person removes some of the variation, implying that comparisons made on a per

person basis are more robust than using per household data. Use of per person figures

also brings reporting more closely in line with other food waste measurement

initiatives21.

2.6 Confidence intervals

Confidence intervals were calculated for the main estimates in the report. These

measure the uncertainties arising from sampling errors: i.e. emanating from a sample of

LAs completing waste compositional analysis, and for each waste compositional

analysis, only a sample of houses being included (typically 135-250). For each waste

stream where sampling forms part of the estimation process (kerbside residual,

kerbside dry recycling and HWRC residual), the variation in the percentage of that waste

stream that is food was used to construct the confidence interval. The estimates for the

amount of food waste in collections that target food waste was assumed to be without

sampling error, as all the information was derived from WasteDataFlow (i.e. a census of

all LAs).

For the kerbside residual stream, the confidence intervals were calculated for each

stratum (LAs with collections targeting food waste and LAs without such collections).

These two confidence intervals were then combined to give an overall confidence

interval for kerbside residual. The confidence interval for all the waste streams was then

19 WRAP CFP102, Methods used for Household Food and Drink Waste in the UK 2012, WRAP 2013, Chapter 7

20 Household Food and Drink Waste in the UK 2012, WRAP 2013: http://www.wrap.org.uk/content/household-food-and-drink-

waste-uk-2012

21 E.g. Estimates of European food waste levels, FUSIONS, 2016: http://www.eu-

fusions.org/phocadownload/Publications/Estimates%20of%20European%20food%20waste%20levels.pdf



http://www.eu-fusions.org/phocadownload/Publications/Estimates%20of%20European%20food%20waste%20levels.pdf

http://www.eu-fusions.org/phocadownload/Publications/Estimates%20of%20European%20food%20waste%20levels.pdf


combined to give an overall estimate of confidence, quoted at the 95% confidence

level22.

These confidence intervals omit contributions from systematic errors, which are usually

difficult to quantify and / or combine with random sampling errors. Systematic errors

may stem from a number of sources including from: whether the LAs with waste

compositional analyses were representative of all UK LAs; the factors used in calculating

the amount of food waste in mixed garden and food waste; the time of year which waste

compositional analyses were undertaken. These systematic errors are discussed further

in Appendix 6.

Where two estimates were compared (e.g. between nations within the UK), t-tests were

used to inform whether the differences were significant.

2.7 Interpretation of trends in food waste arisings

Over recent years, there have been a number of studies to estimate the amount of LA-

collected food waste from UK households. Within these studies, there have been

improvements to the estimation methodology over time, so that previously published

estimates for 2010 and 2012 have been updated in this report, using the most recent

methodology23. Comparisons are also made to the 2007 estimate even though some

small methodological differences exist. Given the scale of change between 2007 and

2010, these small differences are unlikely to change the conclusions relating to these

years.

The details of how the methodology has been improved over time are detailed in

Appendices A4.5 and A4.6. In addition, where methodological differences are still

present, these are discussed.

Table 1 summarises key data used for current and historic estimates. This includes

which periods of WasteDataFlow were used, the fieldwork dates for waste compositional

analyses, and how many LAs were covered by these waste compositional analyses.

22 Combining confidence intervals in this way assumes that the estimates and their uncertainties are independent. Where waste

compositional analyses have measured waste in multiple streams within the same LA, there may be some correlation between

the uncertainties calculated, but the effect on the overall confidence intervals (which are dominated by the kerbside residual

waste stream) is likely to be small.

23 On revision, the UK estimate for 2012 was revised upwards by 3,000 tonnes (less than 0.1%). For 2010, the UK estimate was

revised upwards by 142,000 (c. 3%).

WRAP – 23

Table 1: Comparison of key features of household food waste estimates for different studies

Target period for estimates

2007 2010 2012 2014 2015

Nations from which compositional studies were included

England England, Wales, Scotland, Northern Ireland

England, Wales, Scotland

England, Wales, Scotland

England, Wales, Scotland, Northern Ireland

Time period of WDF data

UK: Jan2007 to Dec 2007

UK: Oct 2009 to Sept 2010

England: Oct 2011 to Sept 2012; Wales and Northern Ireland: April 2011 to Mar 2012; Scotland: Jan to Dec 2011

England and Wales: Oct 2013 to Sept 2014; Northern Ireland: Jan 2014 to Dec 2014; Scotland: Jan 2013 to Dec2013

England and Wales: Oct 2014 to Sept 2015; Northern Ireland: Apr 2014 to Mar 2015; Scot: Jan 2014 to Dec 2014

Time period of compositional data

February 2005 to September 2008

January 2009 to April 2011

February 2011 to March 2013, though excluding any

studies carried out in 2011 that had been included in the 2010 estimates

April 2013 to March 2015

April 2014 to March 2016

No. of LAs from which kerbside residual compositional data included

120 87 63 87 116

Organisation commissioning the ‘synthesis’ study

Main study = Defra (additional calculations performed for WRAP)

WRAP WRAP WRAP WRAP

Link http://randd.defra.gov.uk/Default.aspx?Module=More&Location=None&ProjectID=15133

www.wrap.org.uk/sites/files/wrap/Synthesis%20of%20Food%20Waste%20Compositional%20Data%202010%20FINAL.pdf


This report











WRAP – 24

3.0 Results

§3.1 presents estimates of local authority (LA) collected household food waste arisings

for the UK for 2014 and 2015, including a comparison with historical estimates.

Estimates for England, Wales, Scotland and London are included in §3.2. Capture rates

for household food waste arisings are included in §3.3. The estimates presented in this

section are standard, pooled estimates, as described in §1.1. The alternative method is

presented in Appendix 4 and single year estimates are presented in Appendix 5.

3.1 UK results

In this section, results are presented for 2014 estimates (§3.1.1) and 2015 estimates

(§3.1.2). Trends over time are presented in §3.1.3.

3.1.1 UK arisings in 2014

Table 2 shows estimates for household food waste collected by LAs in the UK in 2014.

Food waste arisings expressed in terms of kg per household and kg per person are

shown in Table 3.

Table 2: Estimates of household food waste collected by local authorities in the UK 2014,

thousand tonnes

Waste Stream Food waste

arisings

95%

Confidence

Interval24

Kerbside residual 4,198 ±148

Kerbside collections targeting food

waste 602 025


(contamination) 78 ±39

HWRC residual 57 ±25

TOTAL 4,934 ±15526

There was an estimated 4.93 million tonnes of food waste collected from UK households

by LAs in 2014. This is equivalent to 76.4 kg per person. The 95% confidence interval is

±155,000 tonnes, ±2.4 kg per person or ±3.1%.

24 Confidence intervals include sampling errors, but do not include other uncertainties.

25 Given that this information is derived from WasteDataFlow, there is no sampling error, but there will be other

uncertainties associated with it (discussed in Appendix 6).

26 The combined confidence interval is smaller than the sum of the confidence intervals for each waste stream as

the errors are assumed independent and therefore balance one another out to an extent.



kg per household and kg per person

Waste Stream

kg per household in 2014 kg per person in 2014

Food waste

arisings

95%

Confidence

Interval

Food waste

arisings

95%

Confidence

Interval

Kerbside residual 154.4 ±5.4 65.0 ±2.3

Kerbside collections

targeting food waste 22.1 ±0.0

9.3 ±0.0

Kerbside dry

recycling

(contamination)

2.9 ±1.4

1.2 ±0.6

HWRC residual 2.1 ±0.9 0.9 ±0.4

TOTAL 181.5 ±5.7 76.4 ±2.4

The vast majority of this food waste (4.20 million tonnes; 85% of the total) was found in

the kerbside residual waste stream, with most of the rest (0.60 million tonnes; 12% of

the total) in kerbside collections targeting food waste (either separate food waste

collections or mixed food and garden collections).

3.1.2 UK arisings in 2015

Table 4 shows estimates for household food waste collected by LAs in the UK in 2015.

Food waste arisings expressed in terms of kg per household per year and kg per person

per year are shown in Table 5.

The estimate for food waste collected from UK households by LAs in 2015 was 4.90

million tonnes. This is equivalent to 75.2 kg per person. The 95% confidence interval is

±134,000 tonnes, ±2.1 kg per person or ±2.7%.

As with 2014, the vast majority of this food waste (4.12 million tonnes; 84% of the total)

was found in the kerbside residual waste stream, with most of the rest (0.64 million

tonnes; 13% of the total) in kerbside collections targeting food waste (Figure 1).



thousand tonnes

Waste Stream Food waste

arisings

95%

Confidence

Interval

Kerbside residual 4,117 ±125

Kerbside collections targeting food

waste 639 0


(contamination) 80 ±40

HWRC residual 60 ±27

TOTAL 4,896 ±134


kg per household and kg per person

Waste Stream

kg per household in 2015 kg per person in 2015

Food waste

arisings

95%

Confidence

Interval

Food waste

arisings

95%

Confidence

Interval

Kerbside residual 150.0 ±4.6 63.2 ±1.9

Kerbside collections

targeting food

waste

23.3 ±0.0 9.8 ±0.0

Kerbside dry

recycling

(contamination)

2.9 ±1.4 1.2 ±0.6

HWRC residual 2.2 ±1.0 0.9 ±0.4

TOTAL 178.4 ±4.9 75.2 ±2.1


Figure 1: Household food waste collected by local authorities in the UK 2015, by waste

stream

3.1.3 UK trends in LA collected household food waste

For the purposes of comparing household food waste arisings for different periods, only

food waste in kerbside residual and collections targeting food waste for treatment has

been reported, as this represents the vast majority of food waste arisings (c. 97%), and

estimates for the other (minor) streams are relatively uncertain due to a lack of data.

The 2014 and 2015 estimates should not be compared directly with one another, as the

two estimates have significant overlap in the compositional studies that are included in

each estimate, and so are not independent. However the earlier estimates (2007, 2010

and 2012) are all entirely independent of the 2014 and 2015 estimates, and independent

of each other, and so can be compared in order to assess trends in LA collected

household food waste arisings in the UK.

Figure 2 shows total household food waste arisings, in tonnes, from the current and

previous studies, and Figure 3 shows the same estimates, expressed as kg per person in

a given year. These include estimates using the standard pooled method, along with

95% confidence intervals for each estimate. The same data is illustrated in Table 6 and

Table 7.

The results show that there was a substantial (and statistically significant) reduction in

the amount of food waste collected from UK households by LAs from 2007 to 2010. This

reduction amounted to 962 thousand tonnes less household food waste being

generated in 2010 compared to 2007. When comparing the amount of food waste per

person, the reduction between the two years was 17 kg per person. It should be noted

that there are some small differences between the method used to calculate the 2007


estimate and that used for subsequent estimates (2010, 2012, 2014 and 2015). These

are discussed in §2.7 and Appendix A4.5. The effect of these differences on the results is

thought to be small, such that the above conclusion is still valid.

Figure 2: Estimated arising of local authority collected household food waste in the UK

2007 to 2015, million tonnes (±95% confidence interval)27



Table 6: Estimated local authority collected household food waste in the UK 2007 to

2015, thousand tonnes

Waste stream 2007 2010 2012 2014 2015

Food waste in kerbside residual 5,488 4,322 4,040 4,198 4,117

Food waste in kerbside collections

targeting food waste 88 293 537 602 639

Total food waste (kerbside

residual plus collections

targeting food waste)

5,577 4,615 4,577 4,799 4,756

95% confidence interval ±225 ±147 ±171 ±148 ±125

27 In the 2012 synthesis study, a pooled estimate for 2009 was reported. Due to the practical limitations within the current

project, it was not possible to apply the most recent improvements in the methodology to this year; it has therefore not been

reported here.

4.0

4.2

4.4

4.6

4.8

5.0

5.2

5.4

5.6

5.8

6.0

2007 2008 2009 2010 2011 2012 2013 2014 2015

Mill

ion

s to

nn

es


Since 2012 – the last synthesis report – there has not been a significant change in the

food waste in these two streams. The apparent increase from 4.58 million tonnes to 4.76

million tonnes (an increase of 179,000 tonnes, equivalent to 3.9% increase) is not

statistically significant. There was also an apparent increase in the ‘per person' levels of

food waste between the two years, from 71.9 to 73.1 kg per person. Again, this apparent

increase was not statistically significant.

Figure 3: Estimated arising of local authority collected household food waste in the UK

2007 to 2015, kg per person (±95% confidence interval)



Table 7: Estimated local authority collected household food waste in the UK 2007 to

2015, kg per person

Waste stream 2007 2010 2012 2014 2015

Food waste in kerbside residual 89.5 68.9 63.4 65.0 63.2


targeting food waste 1.4 4.7 8.4 9.3 9.8


residual plus collections

targeting food waste) 90.9 73.5 71.9 74.3 73.1

95% confidence interval ±3.7 ±2.3 ±2.7 ±2.3 ±1.9

As noted earlier in this section, most of the food waste has been found in the residual

waste stream. The amount of food waste in this waste stream is the product of the total

50

55

60

65

70

75

80

85

90

95

100

2007 2008 2009 2010 2011 2012 2013 2014 2015

Food waste (kg per person per year)


amount of waste (of all materials) in the residual stream and the percentage of this that

is food. The trends in these two components are presented below.

The average percentage of the residual waste stream that is food waste was 33% in

2015. This average percentage has varied slightly since 2007, with a high of 35% in 2007

and a low of 32% in 2012 (Figure 4). The equivalent percentage for LAs targeting food

waste was generally lower, but followed a similar pattern, ranging from 28% - 32%. For

those LAs without collections targeting food waste, the pattern was similar with higher

levels (34% - 38%).

Figure 4: Trend in average percentage of kerbside residual waste collections measured

to be food waste, UK data

Whilst the percentage of residual waste that is food has not varied greatly over time, the

amount of residual waste collected from UK households has declined by approximately

22% between 2007 and 2015(Figure 5), despite a 6.2% increase in population.

Figure 5: Trend in kerbside residual waste from UK households


3.2 Arisings in England, Wales, Scotland and London

There were a sufficient number of waste compositional studies to calculate separate

estimates of household food waste for England, Wales and London in 2015 and Scotland

in 2014. However, there was insufficient data to produce an estimate for Northern

Ireland. The estimates for England, Wales, Scotland and London include only food waste

in kerbside residual and food waste collected at the kerbside for treatment because

there was insufficient data for the minor waste streams (HWRC residual and

contamination of dry recycling) for each of these geographical areas.

It is important to note that methodological issues may have influenced these national

results to a small degree. In particular, the latter half of Appendix A1.1 and Appendix 6

contain discussion of some of the uncertainties associated with these estimates.

3.2.1 Arisings in England

Estimated arisings of LA collected household food waste in England in 2015 are shown in

Table 8, in terms of tonnes, kg per household and kg per year.

Table 8: Estimated arising of local authority collected household food waste in England,

2015

Waste Stream Thousand

tonnes

kg per

household kg per year

Food waste in kerbside residual 3,493 152.2 63.7


targeting food waste 474 20.7 8.6


residual plus collections targeting

food waste)

3,967 172.9 72.4


The 2015 estimates have been compared (in terms of kg / person / year) to estimates for

England for 2009, calculated for the current study using the same method. The amount

of food waste in these two streams was higher in 2015 (72.4 (±2.4) kg / person / year)

compared to 2009 (71.6 (±4.0) kg / person / year). This difference of 0.8 kg / person /

year was not statistically significant. The amount of food waste collected for treatment

more than doubled.


Table 9: Estimated arising of food waste in residual and collections targeting food waste

in England in 2009 and 2015 (kg / person / year)

Waste Stream 2009 2015 Difference

Food waste in kerbside residual 68.1 63.7 -4.4


targeting food waste 3.5 8.6 +5.2


residual plus collections targeting

food waste)

71.6 72.4 +0.8

95% Confidence interval ±4.0 ±2.4 ±4.7

3.2.2 Arisings in Wales

Estimated arisings of LA collected household food waste for Wales in 2015 are shown in

Table 10.

The average amount of food waste per person generated in Wales in 2015 (66.2 ± 1.6 kg

per person) was significantly lower than for England (72.4 ± 2.4 kg per person). The

amount of food in collections targeting food waste is much higher than the rest of the

UK, with 31.6 kg per person collected in 2015, 48% of the total food waste. This is a

reflection that all local authorities in Wales have collections targeting food waste that

cover almost all households.

Table 10: Estimated arising of local authority collected household food waste in Wales

2015


Tonnes

kg per

household

kg per

person

Food waste in kerbside residual 107 79.9 34.6

Food waste in kerbside collections targeting

food waste 98 73.1 31.6

Total food waste (kerbside residual plus

collections targeting food waste) 205 153.0 66.2


The previous year for which there were a sufficient number of waste compositional

studies available in Wales to make an estimate was 2009. A 2009 figure for Wales was

recalculated using a methodology consistent with the 2015 estimate (Table 11). These

show that the central estimate for 2015 (66.2 kg per person) is 12% lower than for 2009

(75.4 kg per person), a difference which is statistically significant.


Table 11: Estimated arising of food waste in residual and collections targeting food

waste in Wales in 2009 and 2015 (kg per person)


Food waste in kerbside residual 68.0 34.6 -33.5

Food waste in kerbside collections targeting food

waste 7.3 31.6 +24.2


collections targeting food waste) 75.4 66.2 -9.2


3.2.3 Arisings in Scotland

Estimated arisings of LA collected household food waste for Scotland in 2014 are shown

in Table 12 using the standard method.

Table 12: Estimated arising of local authority collected household food waste in Scotland

2014 (using the standard method)


Tonnes

kg per

household

kg per

person



food waste 40 16.4 7.4




As described in the methodology section (§2.4), the result for Scotland in 2009 using the

standard method was not robust due to a lack of data in one of the two strata. For this

reason, the results using the alternative method have also been presented, both for

2014 (Table 13) and a comparison of 2009 and 2014 (Table 14). The difference between

the standard and alternative method in 2014 was 6,000 tonnes or 1.6%.

These two different estimates for 2014 were not statistically different – on a per person

basis – to the rest of the UK in that year. This suggests that levels of household food

waste in Scotland are similar to those in the rest of the UK.


Table 13: Estimated arising of local authority collected household food waste in Scotland

2014 (using the alternative method)


Tonnes

kg per

household

kg per

person



food waste 40 16.4 7.4




Table 14 makes a comparison between 2009 and 2014 in Scotland using the alternative

method. The difference in central estimate (–4.3 kg per person, equivalent to –6%) is not

statistically significant at the 95% confidence level. The apparent reduction between

central estimates in Scotland would be greater if the standard method (used in the other

UK nations and regions) was applied, though it would still not be statistically

significant28.

Table 14: Estimated arising of food waste in residual and collections targeting food

waste in Scotland in 2009 and 2014 (kg / person / year) using the alternative method


Food waste in kerbside residual 74.3 63.4 –10.9

Food waste in kerbside collections targeting food

waste 0.8 7.4 6.6


collections targeting food waste) 75.1 70.8 –4.3


Table 14 also illustrates that the amount of food waste in collections targeting food

waste increased greatly over this time period, from 0.8 kg per person in 2009 to 7.4 kg

28 The headline figures for Scotland here are calculated using the alternative method, whereas figures for other nations and

regions are presented using the standard method and so it is not always appropriate to compare them directly. Comparison can

be made between the standard estimate for Scotland (2014) and other estimates also calculated using the standard method.

However, for the 2009 estimate for Scotland (using the alternative method) and the trend in Scotland between 2009 and 2014

(also using the alternative method), comparison should be made only to estimates for other nations that are also calculated

using the alternative method. Such data for the UK can be found in Appendix 4. More discussion of this is found in the report by

Zero Waste Scotland: Household food and drink waste in Scotland 2014, ZWS (2016).



per person in 2014. This increase mirrored the introduction of collections targeting food

waste over this period. There was a corresponding drop in the amount of food waste in

kerbside residual collections (from 74.3 to 63.4 kg per person).

3.2.4 Arisings in London

Estimated arisings of LA collected household food waste for London in 2015 are shown

in Table 15. This estimate is based on 17 phases of waste compositional data from 14

London boroughs. The amount of food waste collected by LAs in London was 660,000

(±61,000) tonnes or 76.2 (±7.1) kg per person. There are no previous estimates for

London to compare these against.

The estimate – expressed per household or per person – is not significantly different

from the rest of England, implying that arisings in London are broadly in line with

England (and the UK as a whole).

Table 15: Estimated arising of local authority collected household food waste in London

2015


Tonnes

kg per

household

kg per

person



targeting food waste 86 24.8 10.0



95% Confidence interval ±46 ±13.2 ±5.3

3.3 Capture rates for food waste

The capture rate – the percentage of total household food waste that is captured in

kerbside (KS) collections targeting food waste – has been calculated using the formula

below:

Capture rate = Kerbside collections targeting food waste

KS residual + KS collections targeting FW + KS dry recycling (FW contamination) + HWRC residual


The UK capture rate during 2015 was 13.1% ± 0.4%29, a total of 639,000 tonnes in

targeted collections out of a total 4,896,000 tonnes. This covers LAs with collections

targeting food waste as well as those LAs without such collections.

Figure 6 illustrates the increase in capture rates over recent years: in 2007, the capture

rate was 1.6% (±0.1%), and this increased to 13.1% in the eight years to 2015. A large

part of this change is related to the increased coverage of collections targeting food

waste during this period.

Figure 6: Trend in UK capture rates of household food waste (±95% confidence interval)

Capture rates were also calculated separately for LAs that target food waste at the

kerbside for treatment. These capture rates related only to kerbside residual and

kerbside collections targeting food waste for treatment, and excluded food waste in

kerbside dry recycling and HWRC residual waste. This is because the data for these

kerbside dry recycling and HWRC residual waste are not very accurate for small subsets

of LAs.

Capture rate = Kerbside collections targeting food waste

KS residual + KS collections targeting food waste

For the 2014 pooled estimates for the UK, a distinction was made between authorities

providing food waste collections where the food is presented separately by

householders and presented mixed with garden waste, as this factor is known to affect

the capture of food waste. The following average capture rates were calculated:

● All LAs collecting food waste (separately or mixed with garden waste): 25%30

29 95% confidence interval includes contribution from sampling error, and assumes no sampling error in FW in collections

targeting food waste as this information is derived from WasteDataFlow.

30 For comparison, the corresponding capture rate for all local authorities in 2014 using this modified capture rate is 12.5%.


● Food presented separately: 29% (n = 60 LAs in the sample)

● Food presented mixed with garden waste: 16% (n = 33)

The figures above include LAs with varying levels of coverage of food waste collections –

i.e. they include authorities with only partial coverage of collections, such as those at an

early stage of being rolled out more widely. To investigate these figures further,

corresponding calculations were carried out for authorities providing a high coverage of

food waste collections, defined as a service provided to more than 80% of residents:

● High coverage, food presented separately: 35% (n = 44)

● High coverage, food presented mixed with garden waste: 16% (n = 28).

These figures demonstrate that the capture rate was higher on average when food

waste was presented separately than when it was presented mixed with garden waste.

For authorities with mixed garden and food collections and high coverage, only around

one-sixth (16%) of the food waste was found in the collection; the other 84% was in the

residual kerbside collection. Even for LAs with food waste presented separately and high

levels of coverage, approximately two-thirds of the food waste was still in the residual

waste stream on average, with 35% captured in the collections. This indicates that the

proportion of people using collections where available was relatively low and/or that

people were not using them for all food waste.

Due to limitations of time, the full analysis in this was not repeated using the 2015 data,

but the capture rate for all authorities targeting food waste for treatment (whether

separately or mixed with garden waste) was calculated at 31% (n=121), higher than the

corresponding estimate from the 2014 estimates of 25%. This is likely to be because the

2015 estimates included data for all authorities in Wales, where capture rates for food

waste were generally higher.

All LAs in Wales target food waste for treatment, most through separate collections and

a few through mixed food and garden collections. This resulted in an overall capture

rate of 48% of kerbside food waste, which is higher than the UK average for local

authorities with high coverage of separate collections presented above (35%). This

suggests that people in Wales are using their collections – where present – to a greater

extent compared to the rest of the UK.


4.0 Conclusions

For the UK, per capita food waste arisings were much lower in 2015 compared to 2007

(the first year with estimates that are sufficiently similar to allow comparison). The

majority of the reduction occurred between 2007 and 2010. Since 2012, there has been

no significant evidence of further reductions.

The 2015 estimate for the average amount of household food waste per person for

England was not significantly different to that for the UK as a whole. Similarly, the

amount in London was not significantly different to the rest of England.

In Wales, there was a reduction of 12% in the total amount of food waste per person

generated from 2009 to 2015. In 2015, the amount wasted per person was significantly

lower than for England by around 9%. However, methodological issues may have

influenced these results to a small degree (see Appendix A). Potential reasons for this

difference between Wales and England are discussed in Household Food Waste in the UK,

201531.

The most recent estimate for household food waste in Scotland (2014) was similar to

that for the rest of the UK for the same year. The average amount of HHFW per person

was 6% lower in 2014 compared to 2009; however, this difference was not statistically

significant.

There is some evidence that the presence of collections targeting food waste is

associated with slightly lower levels of food waste generated (the total of that in

kerbside residual and in collections targeting food waste). The average effect was 6.5

(±6.7) kg / hh / yr, which is marginally non-significant (p = 0.058). This could be due to

greater awareness of the amounts of food waste disposed by households that use

collections targeting food waste, leading to a change in actions (e.g. shopping, food

preparation). This is discussed in more detail in a WRAP report from 201332. However, as

this factor is marginally non-significant, the result should be interpreted with caution

and further analysis of the dataset would be beneficial.

The capture rate – the proportion of LA-collected household food waste found in

collections targeting food waste – increased between 2007 and 2015 from around 2% to

13%. However, this means that the proportion of food waste collected in targeted

collections in 2015 was still relatively low, with the vast majority – 87% – being found in

the residual waste streams (or contamination of dry recycling). In LAs that had a wide

coverage of separate food waste collections, the capture rate was around one-third

(35%) with two-thirds still going into the residual waste – implying that there is potential

to increase the use of these collections where they are present. Wales has a higher

capture rate (48%) than the rest of the UK, due to a higher proportion of households

with a collection targeting food waste and greater use of these collections.


32 Effect of food waste collections on arisings: recent evidence, WRAP (2013)

http://www.wrap.org.uk/sites/files/wrap/Effect%20of%20food%20waste%20collection%20on%20arisings%20WRAP%20UK_0.pdf




Appendix 1: Regression analysis of influences on food

waste arisings

Multiple regression analysis was carried out to help explain differences in the amount of

household food waste produced between Local Authorities (LAs). In lay terms, multiple

regression is a type of analysis that attempts to explain the variation in a key variable

(the dependent variable) by identifying factors from a list of potential explanatory

variables that exhibit a relationship with it (referred to as explanatory or independent

variables).

The objective of this appendix is to determine which explanatory variables are important

to the analysis in the main body of the report. In particular, it seeks to identify

explanatory variables for which it is important that the sample of waste compositional

analyses is representative of the wider population. In cases where the sample is not

representative, weighting or stratification can be introduced to adjust for this

unrepresentativeness and minimise biases in the estimates.

In this study, two dependent variables were explored:

● The percentage of the residual waste stream which is food waste (as measured by

waste compositional datasets): results presented in A1.1.

● The combined amount of food waste collected by LAs in the residual and in

collections targeting food waste (i.e. collections of either separate food waste or

mixed organics collections), quantified in kg / household / year (A1.2).

For each dependent variable, several models were produced in order to explore which

model provided the best statistical evidence and explanatory framework for

understanding how the independent variables tested impact on the dependent variable.

For each dependent variable, two models are presented here, an ‘inclusive’ model which

includes all of the independent variables tested, and a ‘parsimonious’ model which used

a step-wise approach (eliminating independent variables that were least significant in

each iteration of the model) to arrive at a model that only includes independent

variables found to be significant in affecting the dependent variable.

The findings from analysing the first dependent variable (the percentage of the residual

waste stream which is food waste) are useful for quantifying household food waste

using the ‘standard’ method (§2.4) as this uses this particular percentage as the basis of

the calculations. The calculations within the ‘alternative’ method (Appendix 4) use the

combined amount of food waste collected by LAs in the residual and in collections

targeting food (measured in kg / household / year). Hence, the second set of regression

models (A1.2) informs that approach.

As with the main analysis, multi-phase waste compositional analyses were split into their

constituent phases in order to maximise the sample size available for analysis (see

§2.3.4). In addition, waste compositional data from the current study (2014 and 2015

estimates) and previous studies (2010 and 2012) was included in the dataset. The total

number of phases of waste compositional analysis explored in the regression analysis

was 449.

For both models, the explanatory variables tested in the analysis were:


● Level of deprivation (measured in terms of proportion of population in the LA

belonging to Social Grade D or E) – see A2.1 for a description of the factor used

● Population density (inhabitants per hectare)

● Whether or not kerbside fortnightly residual collections are provided (as opposed to

weekly)

● Whether or not food is collected at the kerbside from households for treatment

● Whether or not the waste compositional study used to produce the percentage

arising of food waste in kerbside residual for the LA included a ‘packaged food’

subcategory

● Nation in which the waste compositional study was carried out

● The year in which studies were carried out (these were grouped into two-year blocks,

i.e. 2009 & 2010, 2011 & 2012, etc. as a control variable in order to smooth out the

effects of variations between individual years)33

● The quarter of the year during which the waste compositional study was carried out

(to examine seasonality as a factor)

For the last two variables (year and quarter), the variables were treated as two groups.

In practice, this meant that their inclusion or exclusion from the parsimonious model

was determined on the significance of the inclusion of all the years (or quarters) in the

model, rather than each variable separately. This is why, for example, all the years and

seasons were included in the parsimonious model in A1.1 – they were significant as a

group, even if some years and quarters were not significant individually. Variables

relating to neither year nor quarter were found to be significant in the second set of

models (A1.2).

A further independent variable that was explored in the first set of models was the

contractor undertaking the waste compositional analysis. This was to explore whether

there were any systematic variations in measurements between contractors. This was a

problematic factor to account for and so was included in a separate regression model. A

discussion relating to this is included in Appendix A1.1.

There are likely to be other variables that could influence food waste: for instance, the

degree to which food waste prevention campaigns have been run locally (e.g. by the LA

or a local charity). However, data for each LA relating to such variables was not possible

to obtain for this study and consequently it was not possible to include these in the

regression analysis.

A1.1 Regression results: percentage of kerbside residual which is food waste

This section presents the results and discussion relating to the first set of regression

models, focusing on the percentage of kerbside residual waste that is food waste. Table

16 shows the ‘inclusive’ model (see comments above) and Table 17 shows the

‘parsimonious’ model.

33 In contrast to the other explanatory variables, the year was used as a control variable in the modelling to ensure that any

trend in the amount of household food waste generated over time did not adversely affect the regression modelling. As the main

analysis in this report focused on specific years that are then compared with previous estimates, the ‘year’ variable was not

considered as a factor for weighting or stratification.


The R2 for the ‘inclusive’ model (Table 16) is 0.230, indicating that the model explains

23.0% of the variation in the dependent variable. The R2 is 0.224 for the ‘parsimonious’

model (Table 17) indicating that the model explains 22.4% of the variation in dependent

variable – a small reduction due to the removal of many non-significant factors from the

model.

Table 16: ‘Inclusive’ regression model for percentage of kerbside residual that is food

waste

Explanatory variable

Unstandard-

ised

coefficient

Signifi-

cance

95% CI

lower

bound

95% CI

upper

bound

Constant 34.4% < 0.0005 31.3% 37.5%

Food waste targeted for

treatment -5.6% < 0.0005 -7.0% -4.2%

Kerbside residual fortnightly -0.9% 0.239 -2.3% 0.6%

Deprivation level 2.5% 0.597 -6.9% 12.0%

Population density

inhabitants per hectare 0.0% 0.820 0.0% 0.0%

Wales -2.6% 0.023 -4.8% -0.4%

Scotland 0.0% 0.971 -1.9% 1.8%

Packaged food waste included

in food waste subcategories 0.9% 0.224 -0.6% 2.5%

Quarter 1 (Apr-Jun) -3.0% 0.001 -4.8% -1.2%

Quarter 2 (Jul-Sep) -0.6% 0.434 -2.2% 1.0%

Quarter 4 (Jan-Mar) 0.2% 0.828 -1.5% 1.9%

Years 2011 and 2012 0.6% 0.538 -1.3% 2.6%

Years 2013 and 2014 3.0% 0.001 1.2% 4.8%

Years 2015 and 2016 0.6% 0.574 -1.4% 2.5%

Both models only explain a relatively small proportion of all the variation between LAs.

However, due to the fact that data for individual LAs comes from a relatively small

number of households (typical sample sizes were 135-250), we would expect a relatively

large amount of variation in the data from this fact alone34. Where variation stems from

34 This variation arising from small sample sizes comprises a) differences in the long-term average food waste arisings between

households selected and b) week-to-week variation in the amount of food waste produced by a single household. For more


relatively small sample sizes, we would not expect any variable to be able to explain the

variation within the regression modelling. Therefore, there is a limit on the proportion of

the observed variation that can be explained in regression modelling.

Table 17: ‘Parsimonious’ regression model for percentage of kerbside residual that is

food waste

Explanatory variable Unstandardised

coefficient

Signifi-

cance

95% CI

lower

bound

95% CI

upper

bound

Constant 35.0% < 0.0005 33.4% 36.7%


treatment -5.6% < 0.0005 -7.0% -4.3%

Wales -2.0% 0.032 -3.8% -0.2%

Quarter 1 (Apr-Jun) -3.0% 0.001 -4.9% -1.2%

Quarter 2 (Jul-Sep) -0.7% 0.395 -2.2% 0.9%

Quarter 4 (Jan-Mar) 0.2% 0.777 -1.4% 1.9%

Years 2011 and 2012 0.3% 0.737 -1.5% 2.1%

Years 2013 and 2014 2.5% 0.004 0.8% 4.1%

Years 2015 and 2016 0.3% 0.711 -1.3% 2.0%

Figure 7 illustrates the distribution of the percentage of residual stream that is food

waste for phases of waste compositional analysis used to produce the UK 2015

estimate. It should be noted that the wide variation in this percentage is partly a

reflection of the relatively small number of households discussed above. The variation

between LAs (i.e. if it were possible to measure the percentage food waste in all of the

residual waste from households) would be much smaller.

discussion on this topic, see Appendix 1 of The Milk Model: Simulating Food Waste in the Home, WRAP (2013)

http://www.wrap.org.uk/sites/files/wrap/Milk%20Model%20report.pdf

http://www.wrap.org.uk/sites/files/wrap/Milk%20Model%20report.pdf


Figure 7: Distribution of the percentage of residual stream that is food waste measured

by waste compositional analysis. Data includes all phases used for UK 2015 estimate.

In the parsimonious model, the amount of food waste in the kerbside residual waste

stream was 5.6 (±1.3) percentage points lower in LAs that had collections targeting food

waste at the kerbside, a highly significant difference (p < 0.0005). This is clear evidence

that food waste collected for treatment is reducing the proportion of food waste in the

residual waste stream. This is an important finding for scaling up results from a sample

of LAs with waste compositional data – it is important to stratify the data if there is any

under- or over-representation of food waste collections in the sample relative to all LAs

in the UK. This is an important feature of the standard method used for this and

previous reports.

Figure 8 illustrates this point: that waste compositional analyses in LAs with collections

targeting food waste measured, on average, lower percentages of food waste compared

to those LAs without such collections.


Figure 8: As for Figure 7, but split by whether the waste compositional analysis is carried

out in a LA with collections targeting food waste

Whether waste compositional analysis was performed in Wales was a significant factor

in the model, with lower arisings in Wales compared to other parts of the UK (2.0 (±1.8))

percentage points lower, p = 0.032). It is interesting to compare this to the fact that

Wales is not a significant factor in the models of food waste per household (A1.2). Taking

these two results together, this suggests that the presence of Wales as a factor relates to

diversion of food waste away from the residual towards targeted collections (rather than

a lower level of food waste generated across all years). All Welsh LAs have some form of

targeted collections and, in recent years, the level of diversion is higher compared to

other UK LAs with targeted collections: in 2015, the average yield in Wales was 73 kg / hh

/ yr, whereas across the rest of the UK it was 37 kg / hh / yr (for those LAs with

collections targeting food waste). This increased yield is taken into account in the

adjustment factor (see Appendix 4.1.2) and therefore there is no need to stratify or

weight to account for under- or over-representation of Welsh authorities, as this would

lead to adjusting for the effect twice.


Seasonality appeared as a significant factor in the model, with a lower percentage in the

April to June quarter (3.0 percentage points lower than the reference period (Oct-Dec),

p = 0.001). The authors were not aware of any research that explains these results,

although there are many potential explanations. Significant differences were not seen

between the other three quarters of the year. Therefore further investigation would be

useful to understand this effect. Consideration was given as to whether to stratify the

sample to account for seasonal variation in food waste arisings, particularly as the

coverage of audits from April to June is underrepresented in the sample used for the

2015 pooled estimates (standard method). However, such stratification only had a

minimal impact on the results (approximately 12,000 tonnes lower estimate for

household food waste in UK kerbside residual waste, c. of the total 0.3%) and therefore

was not used in the final estimates as it would greatly increase the complexity of the

calculations with negligible impact on the results; see comments in Appendix A2.4.

The year in which the waste compositional analysis was undertaken exhibited a

significant relationship with the percentage of residual which is food waste. However,

there wasn’t a clear trend over time. Food waste was found to be 2.5 percentage points

higher (p = 0.004) in 2013 & 2014 compared to the reference period (2009 & 2010), all

other factors having been accounted for. However, there was no significant difference

between the other two periods (2011 & 2012, 2015 & 2016) and the reference period.

This higher percentage for 2013 & 2014, but no other period, is difficult to explain – it

could be a real (but temporary) effect, or related to the LAs that decided to perform

compositional analysis during this time. As mentioned previously, this group of variables

was used to control for trends over time, and so it was not necessary to stratify or

weight the main analysis by year, since the estimates produced by the main analysis are

time bounded.

The model is also interesting in terms of the factors that were not found to be significant

(Table 16):

● Level of deprivation (measured in terms of proportion of population in the LA

belonging to Social Grade D or E)

● Population density (inhabitants per hectare)

● Whether kerbside fortnightly residual collections are provided

● Whether packaging was included in subcategories for food waste in waste

compositional studies.

Overall, the model provides justification for the stratification approach adopted for the

‘standard’ method (§2.4), particularly in relation to the decision to only stratify the

sample by whether or not food waste was targeted at the kerbside for treatment. This is

further discussed in the coverage assessment of Appendix 2.

Another independent variable was explored in the regression modelling, relating to

whether there were differences in the percentage of food waste arisings found by some

contractors undertaking waste compositional analysis compared to others. Differences

were found between two groups of contractors35 in the regression modelling that were

35 The different groups of contractors are not named – data for these studies was provided on the basis that it would not be

linked back to contractors or individual local authorities.


significant. The average difference between these two groups, as measured by this

additional regression analysis, was 2.9% (the 95% confidence interval ranges from 1.7% -

4.1%).

It has been suggested that the difference could stem from the use of a subcategory

within food waste of ‘packaged food’ – i.e. that a small element of packaging was being

included in the percentage of food waste and increasing the estimate. However, it was

found that this was unlikely to be the explanation – controlling for the inclusion of a

‘packaged food’ subcategory in the regression model did not eliminate the effect relating

to the contractors.

It is possible that the contractor effect relates to differences in methodological

approach. However after several discussions with the main contractors who contributed

compositional studies to this research, it was not possible to establish whether this was

the case, and neither was it possible to find any evidence for differences in

methodological approach that could explain systematic differences in percentages of

food waste in kerbside residual found by different contractors.

Nevertheless, a sensitivity analysis was performed to understand the impact on the

main estimates in this report of this effect. This was performed by repeating the main

(standard, pooled) calculations with two sets of modifications:

● Increased the percentage of kerbside residual that is food waste for the group of

contractors with lower levels, by the amount found in the regression analysis (2.9%)

● Decrease the percentage of kerbside residual that is food waste for the group of

contractors with higher levels, by the amount found in the regression analysis (again,

2.9%)

Table 18 shows the resultant effect on the estimates for household food waste in

kerbside residual waste in the UK. It is assumed that tonnage estimates for household

food waste in all other streams would not be impacted by this effect.

Table 18: Impact of ‘contractor effect’ on estimated tonnages of household food waste

in kerbside residual in the UK (all figures in thousands of tonnes)

2010 2012 2015

Main estimate 4,322 4,040 4,117

Difference if adjusted downwards -254 -200 -159

Difference if adjusted upwards 140 172 200

Average of the absolute change

(irrespective of the sign) 197 186 179

Note: previously published figures for household food waste in kerbside residual UK for 2010 and 2012 show some differences to

those in the above table. For these years, the results have been recalculated using the methods described, which include some

improvements to the method over time. This allows the results to be compared.

Since there is no firm evidence on which set of contractors was measuring more

accurately, it is not possible to say where the correct value lies. However, they do give


some indication of the likely level of uncertain stemming from which contractor

performs the analysis – somewhere in the region of 179 thousand tonnes in the UK in

2015, or c.4% of the total for the residual waste stream.

It is important to note that this factor has a smaller impact on trends over time for UK

household food waste, assuming that the contractor effect is present to a similar extent

over time. However, if it turns out that one set of contractors are over- or

underestimating food waste in the residual stream, this would substantially alter the

actual amount of food waste, and could be up to the 4% discussed above.

The above discussion underlines the importance of waste compositional analyses being

performed using consistent methodologies. Zero Waste Scotland has published some

useful guidance aimed at achieving a higher level of consistency in waste composition

methodologies36. It is worth the relevant actors across the rest of the UK considering

whether a standardised approach would be practical, as such a change would benefit

many studies (such as this one) that use that data.

A1.2 Regression results: food waste at the kerbside per household per year

This section presents the results and discussion relating to the second set of regression

models, focusing on the amount of food waste in kerbside residual and collections

targeting food waste (in kilogrammes per household per year). Table 19 shows the

‘inclusive’ model, with the same independent variables as used for the ‘inclusive’ model

in the previous section.

The R2 for the ‘inclusive’ model (Table 19) is 0.126, indicating that the model explains

12.6% of the variation in dependent variable. The R2 for the ‘parsimonious’ model (Table

20) is 0.096, indicating that the model explains 9.6% of the variation in dependent

variable. Both models only explain a small proportion of all the variation between LAs,

and this is probably for similar reasons to those for the first set of models (i.e. stemming

from relatively small sample sizes for individual waste compositional analyses). Despite

this, two variables exhibited a significant relationship with levels of food waste, and a

third has a value just above the significance threshold used in this report. These are all

discussed below.

36 Guidance on the Methodology for Waste Composition Analysis, Zero Waste Scotland, 2015


Table 19: ‘Inclusive’ regression model for food waste at the kerbside per household per

year

Explanatory variable

Unstandardised

coefficient (kg /

hh / yr)

Signifi-

cance

95% CI

lower

bound

95% CI

upper

bound

Constant 164.4 < 0.0005 148.1 180.8


treatment -4.9 0.192 -12.2 2.5

Kerbside residual

fortnightly -19.9 < 0.0005 -27.7 -12.2

Deprivation level 120.1 < 0.0005 70.3 170.0

Population density

inhabitants per hectare -0.2 0.053 -0.3 0.0

Wales -7.2 0.232 -19.0 4.6

Scotland -7.7 0.123 -17.6 2.1

Packaged food waste

included in food waste

subcategories

2.4 0.549 -5.6 10.5

Quarter 1 (Apr-Jun) -8.4 0.088 -18.0 1.3

Quarter 2 (Jul-Sep) -6.0 0.161 -14.4 2.4

Quarter 4 (Jan-Mar) -1.1 0.813 -10.0 7.8

Years 2011 and 2012 -5.2 0.327 -15.5 5.2

Years 2013 and 2014 5.3 0.279 -4.3 14.8

Years 2015 and 2016 -1.9 0.717 -12.1 8.4

The presence of kerbside fortnightly residual collections was found to be a significant

factor associated with lower arisings (15 (±6.4) kg / hh / yr lower, p < 0.0005). A potential

explanation for this is that, where the amount of residual waste collected is constrained

(in this case due to lower frequency of collection, but in other cases by the size of

residual waste container), this could divert residual waste from kerbside collections to

HWRC (i.e. more people take residual waste to HWRCs). This would be consistent with

recent WRAP research for 2012/1337, which concluded:

37 Analysis of recycling performance and waste arisings in the UK 2012/13, WRAP (2015):

http://www.wrap.org.uk/content/factors-influencing-recycling-performance

http://www.wrap.org.uk/content/factors-influencing-recycling-performance


“There is no significant relationship between effective weekly residual

containment capacity and total waste arisings. The fact that this predictor

is not significant when looking at total waste arisings indicates that even

though some authorities may experience reduced kerbside arisings

when residual containment is reduced, there is not necessarily a

reduction in total waste arisings so the material is likely to be diverted to

other streams (i.e. HWRCs [Household Waste Recycling Centres]).”

Table 20: ‘Parsimonious’ regression model for food waste at the kerbside per household

per year

Explanatory variable Unstandardised

coefficient

Signifi-

cance

95% CI

lower

bound

95% CI

upper

bound

Constant 161.4 < 0.0005 148.6 174.2

Kerbside residual fortnightly -15.0 < 0.0005 -21.4 -8.5

Deprivation level 94.1 < 0.0005 50.8 137.4


treatment -6.5 0.058 -13.1 0.2

If residual waste is being diverted from the kerbside to HWRCs when containment

capacity of residual is reduced (e.g. by reducing the frequency of collections), we would

expect the percentage of kerbside residual waste that is food to remain approximately

constant with residual collection frequency; this is consistent with the regression models

presented in A1.1. In contrast, we would expect the amount of food waste collected at

kerbside (in kg / hh / yr) to decrease as residual material is diverted to HWRCs –

consistent with the regression model in this section. Therefore, we cannot conclude that

reduced frequency of collections is leading to food waste prevention in the home – it is

more likely that a small amount of food waste is being diverted to HWRCs.

Deprivation is also a significant factor, with an additional 1% of the residents of a LA in

social grades D & E being associated with an extra 0.94 (±0.43) kg / hh / yr (p < 0.0005) of

food waste. The confidence intervals for this factor are large, so the magnitude of the

effect is relatively uncertain. The result could potentially stem from a range of effects:

number of people in each household, use of kerbside residual streams (as opposed to

HWRC streams), as well as actual amounts of food waste generated in the home. A

potential relationship between household food waste and social grade was also

explored in Household Food and Drink Waste: A People Focus38. In contrast to the current

report, this previous report found that those in social grades D & E had similar levels of

food waste to the rest of the population (controlling for household size and prevalence

38 WRAP (2014) http://www.wrap.org.uk/sites/files/wrap/People-focused%20report%20v6_5%20full.pdf

http://www.wrap.org.uk/sites/files/wrap/People-focused%20report%20v6_5%20full.pdf


of home composting). Therefore, the above result should be treated with caution – more

research is required to understand whether there is an effect.

The targeting of food waste for treatment appeared as being a marginally non-

significant factor , indicating that the presence of such collections could be connected

with lower amounts of food waste being generated (6.5 (±6.7) kg / hh / yr, p = 0.058).

This could be due to greater awareness of the amounts of food waste disposed by

households that use collections targeting food waste, leading to a change in actions (e.g.

shopping, food preparation). This is discussed in more detail a WRAP report from 201339.

However, as this factor is marginally non-significant, the result should be interpreted

with caution and further analysis of the dataset would be beneficial. In particular, it

would be useful to analyse whether total arisings are affected by different types of

collection (separate versus mixed garden and food), their frequency or the degree of

diversion.

The coverage of the samples for the 2014 and 2015 calculations in terms of kerbside

residual collection frequency and deprivation (see Appendix 2) were deemed to be

sufficiently good as to not merit stratifying the samples when calculating national

household food waste arisings estimates using the alternative method.

Some factors were not found to have a significant relationship with household food

waste levels. These included:

● Population density inhabitants per hectare – this indicates that levels of household

food waste do not vary along the urban to rural spectrum

● Seasonality – the analysis suggests that the total amount of food waste produced

does not exhibit pronounced seasonality.

● Year – similar to the main analysis in the report, there is no significant trend found in

this analysis for the amount of food waste generated between 2010 and 2015.

● Factors relating to Wales and Scotland – these do not appear as significant. It should

be noted that the regression analysis uses data from 2009 to date, so results

suggesting lower levels of household food waste in Wales more recently (2015) are

obscured by the inclusion of a wide range of data and not looking at interaction

terms.

In Appendix 2, the results from this regression analysis were considered in light of how

representative the sample is of the UK population for the factors included in this

analysis.

39 Effect of food waste collections on arisings: recent evidence, WRAP (2013)




Appendix 2: Coverage Assessment

The coverage assessments presented here relate to the ‘pooled’ 2014 and 2015

estimates for the UK. Coverage assessments for each period are carried out in respect of

levels of deprivation (A2.1), region and nation (A2.2), population density (A2.3), collection

system (A2.4), to assess the degree to which the LAs represented in the study data is

representative of all LAs in the UK. The assessment focuses on the coverage of studies

with data on food waste arising in kerbside residual, which is the most important waste

stream for food waste. Assessments of the coverage of waste compositional studies

over time (year and season) are also included in A2.5, which also indicates the degree to

which there is overlap in the waste compositional datasets used to produce the 2014

and 2015 estimates. A2.6 contains information on the number of phases of waste

composition analysis and LAs included in each set of estimates.

The results from the coverage assessment should be considered alongside those from

the regression analysis presented in Appendix 1. Where a factor exhibits a relationship

with the amount of food waste collected and there is a mismatch in coverage (e.g. there

is an under- or over-representation with regard to the factor in the sample compared to

the population), then there is a case for adjusting the results to account for this

mismatch (e.g. by weighting or stratification).

A2.1 Coverage by levels of deprivation

Coverage in terms of levels of deprivation is illustrated for UK LAs in Figure 9 for the

2014 estimates and in Figure 10 for the 2015 estimates. Indices of Multiple Deprivation

are also available for all the UK nations, but unfortunately these indices are not mutually

comparable between nations. One way around this problem is to measure levels of

deprivation in terms of the proportion of the population found to be of Social Grade D

or E in the 2011 Census. This is a relatively recent measure of deprivation which is

consistently measured in all UK nations, and so is the most appropriate for our

purposes.

The coverage in terms of level of deprivation (using the Social Grade D or E measure) is

illustrated for all authorities in the UK in Figure 9 (2014 estimates) and Figure 10 (2015

estimates). The black line indicates proportions of LA populations of Social Grade D or E,

and the vertical blue bars represent those LAs contributing data on food waste in

kerbside residual to the study.


Figure 9: Coverage of LAs performing compositional studies (levels of deprivation): UK –

2014 estimates

Figure 10: Coverage of LAs performing compositional studies (levels of deprivation): UK

– 2015 estimates

Information on the number of waste compositional phases from each of the quartiles,

as defined by deprivation, is provided in Table 21 for the 2014 estimates and Table 22

for the 2015 estimates.

0%

5%

10%

15%

20%

25%

30%

35%

40%

45%

50%

% p

op

ula

tio

n S

oci

al G

rad

e D

or

E (2

01

1

Cen

sus)

High deprivation local authorities ordered by level of deprivation Low deprivation

0%

5%

10%

15%

20%

25%

30%

35%

40%

45%

50%

% p

op

ula

tio

n S

oci

al G

rad

e D

or

E (2

01

1

Cen

sus)

High deprivation local authorities ordered by level of deprivation Low deprivation


Table 21: Coverage by LA level deprivation in the United Kingdom by quartile – 2014

estimates

Quartile Range for UK LAs, (%

population of Social Grade D/E)

No. sample

LAs

Percentage of

sample LAs

1 6.0% - 20.1% 34 39.1%

2 20.1% - 25.1% 14 16.1%

3 25.1% - 30.6% 20 23.0%

4 30.6% - 43.1% 19 21.8%

Table 22: Coverage by LA level deprivation in the United Kingdom by quartile – 2015

estimates

Quartile Range for UK LAs, (%

population of Social Grade D/E)

No. sample

LAs

Percentage of

sample LAs

1 6.0% - 20.1% 46 26.6%

2 20.1% - 25.1% 32 18.5%

3 25.1% - 30.6% 48 27.7%

4 30.6% - 43.1% 47 27.2%

For the 2014 estimates, Figure 10 and Table 22 show that coverage across the UK for

this measure of deprivation is generally good. However there is a bias towards some of

the least deprived LAs in the country, many of which are from a project that studied all

Surrey districts. This is not thought to have any major bearing on the results as the

average level of deprivation of the audited authorities (22.9%) is within three percentage

points of the national average (25.2%). Regression modelling (Appendix 1) did not find

that deprivation was significant in affecting the percentage of the kerbside residual

waste stream which is food, which is the basis of the ‘standard’ method (§2.4) used for

the main results presented in this study. Hence the decision was taking not to stratify by

this factor.

For the 2015 estimates, Figure 10 and Table 22 show that coverage across the UK for

this measure of deprivation is generally good. As regression modelling (Appendix 1) did

not find that deprivation was significant in affecting the percentage of the kerbside

residual waste stream which is food the decision was taking not to stratify by this factor.


A2.2 Coverage by nation

The coverage of the sampled LAs by UK nation is shown in Table 23 for the 2014

estimates and Table 24 for the 2015 estimates.

For the 2014 estimates, coverage in England and Wales is broadly proportional to the

number of LAs in each nation. Scotland is proportionally over-represented as Scottish

authorities make up 18.4% of the sample but only 8.0% of the UK. There are no datasets

for Northern Ireland.

For the 2015 estimates, Wales is overrepresented due to a major waste composition

study in 2015 that carried out two phases of auditing in each Welsh authority. Coverage

in Scotland is broadly proportional to the number of LAs in Scotland compared to the

UK. England is slightly underrepresented and Northern Ireland is severely

underrepresented due to only one study being available. The regression modelling

(Appendix 1) found that Wales has lower percentage arisings of food waste in kerbside

residual compared to other nations, and this is mostly likely due to the high capture rate

of food waste in Wales (§3.3). However this factor was taken into account through

applied an adjustment in the standard method calculations to account for differences in

food waste yields between the sample and the population (see Appendix 4.1.2, also

Appendix 1), and therefore it was not necessary to stratify the sample by nation.

Table 23: Breakdown of LAs by UK Nation – population and sample – 2014 estimates

Nation Population Sample

No. LAs % of LAs No. LAs % of sample

England 321 80.0% 67 77.0%

Wales 22 5.5% 4 4.6%

Scotland 32 8.0% 16 18.4%

Northern Ireland 26 6.5% 0 0.0%

Total 401 100.0% 87 100.0%


Table 24: Breakdown of LAs by UK Nation – population and sample – 2015 estimates

Nation Population Sample

No. LAs % of LAs No. LAs % of sample

England 321 80.0% 109 63.0%

Wales 22 5.5% 44 25.4%

Scotland 32 8.0% 19 11.0%

Northern Ireland 26 6.5% 1 0.6%

Total 401 100.0% 173 100.0%

A2.3 Coverage by population density

Assessments of coverage by population density in UK authorities are presented in Table

25 for 2014 estimates and Table 26 for 2015 estimates. For both the 2014 and 2015

estimates the sample authorities are spread well across all four quartiles of all UK LA

population densities. Regression analysis suggests that there was not a significant

relationship between the level of food waste and population density (see Appendix 1),

and hence this factor is not used to weight the data.

Table 25: Coverage by LA level population density in the United Kingdom – 2014

estimates

Quartile Range for UK LAs, (inhabitants /

hectare)

No. sample

LAs

Percentage of

sample LAs

1 0.09 – 1.69 21 24.8%

2 1.69 – 5.02 18 23.3%

3 5.02 – 17.89 27 31.0%

4 17.89 – 138.85 21 20.9%


Table 26: Coverage by LA level population density in the United Kingdom – 2015

estimates

Quartile Range for UK LAs, (inhabitants /

hectare)

No. sample

LAs

Percentage of

sample LAs

1 0.09 – 1.69 43 24.9%

2 1.69 – 5.02 40 23.1%

3 5.02 – 17.89 56 32.4%

4 17.89 – 138.85 34 19.7%

A2.4 Coverage by collection system

Table 27 (2014 estimates) and Table 28 (2015 estimates) compare the coverage of

authorities with and without collections targeting food waste for a) the sample of

authorities with waste compositional analyses and b) all authorities in the UK.

For both the 2014 and 2015 estimates, there is significant over-representation of

authorities with collections targeting food waste. This means that the sample is skewed

towards LAs which tend to have a lower percentage of residual waste that is food as

some of the food waste is diverted to collections targeting food waste. To avoid this bias,

the samples have been stratified by the presence of collections targeting food waste,

thus taking this effect into account and producing more robust estimates (§2.4).

Table 27: Coverage of LAs by food waste collection system – 2014 estimates

Food waste collection system

UK Sample LAs

No. LAs % of LAs No. LAs % of

sample

Collections targeting food waste 240 59.9% 60 69.0%

No collections targeting food waste 161 40.1% 27 31.0%

All LAs 401 100.0% 87 100.0%


Table 28: Coverage of LAs by food waste collection system – 2015 estimates

Food waste collection system

UK Sample LAs


sample

Collections targeting food waste 245 61.1% 124 71.7%

No collections targeting food waste 156 38.9% 49 28.3%

All LAs 401 100.0% 173 100.0%

Coverage by kerbside residual collection frequency is shown in Table 29 (2014

estimates) and Table 30 (2015 estimates). For both the 2014 and 2015 estimates, the

split between fortnightly and weekly residual collections in the sample authorities

matches reasonably closely with the split in all UK authorities.

Table 29: Coverage of LAs by kerbside residual collection frequency – 2014 estimates

Kerbside residual collection

frequency

UK Sample LAs

No. Las % of LAs No. LAs % of

sample

Weekly 130 32.4% 26 29.9%

Fortnightly 271 67.6% 61 70.1%

All LAs 401 100.0% 87 100.0%

The regression analysis in Appendix 1 found that kerbside residual collection frequency

was not a significant factor in affecting the percentage of food waste in kerbside

residual, and therefore no weighting or stratification was carried out when using the

standard method. Kerbside collection frequency was found to be significant in affecting

total arisings of food waste at the kerbside (in kerbside residual and food waste

collected for treatment); Appendix 1. This would potentially impact results for

calculations using the alternative method. However the coverage in the samples was

deemed to be sufficiently good as to not merit weighting or stratification for kerbside

residual collection frequency for the alternative method.


Table 30: Coverage of LAs by kerbside residual collection frequency – 2015 estimates

Kerbside residual collection

frequency

UK Sample LAs


sample

Weekly 119 29.7% 43 24.8%

Fortnightly 282 70.3% 130 75.1%

All LAs 401 100.0% 173 100.0%

A2.5 Coverage by period and season

For the 2014 estimates, the studies included in the sample took place between April

2013 and March 2015 (see §2.2.2). Table 31 shows coverage for all compositional data

for kerbside residual waste included in the pooled estimates for 2014. As shown in Table

31, Scotland and England are best represented in the sample, and there are no studies

from Northern Ireland. Only 16 studies from 2015 were included in the 2014 estimates –

this is due to the relatively small window for inclusion (January – March 2015).

Table 31: Coverage by year during which waste compositional audits were carried out -

2014 estimates

Nation 2013 2014 2015 Total

England 34 56 10 100

Wales 4 1 1 6

Scotland 2 16 5 23

Northern Ireland 0 0 0 0

Total 40 73 16 129

For the 2015 estimates, the studies included in the sample took place between April

2014 and March 2016. Table 32 shows coverage for all compositional data for kerbside

residual waste included in the pooled estimates for 2015. As shown in Table 32, Wales

and England are best represented in the sample, and there are no studies from

Northern Ireland. For the 2015 estimates, only 20 studies from 2016 were included – his

is due to the relatively small window for inclusion (January – March 2016).

The overlap of compositional studies used in both the 2014 and 2015 estimates is

discussed in A2.6.


Table 32: Coverage by year during which waste compositional audits were carried out -

2015 estimates

Nation 2014 2015 2016 Total

England 54 35 20 109

Wales 0 44 0 44

Scotland 14 6 0 19

Northern Ireland 1 0 0 1

Total 69 85 20 173

Table 33 (2014 estimates) and Table 34 (2015 estimates) show the number of phases of

auditing by season for the sample, relating to compositional data for kerbside residual

included for each set of estimates. The seasons have been defined using the four

reporting quarters for WDF that are in line with the financial year, and are as follows:

● Quarter 1: April, May, June;

● Quarter 2: July, August, September;

● Quarter 3: October, November, December; and

● Quarter 4: January, February, March.

For the 2014 estimates the samples are spread relatively evenly across the four

quarters, with only a marginal bias towards October to December (Table 33).

Table 33: Seasonal coverage by number of phases of auditing for kerbside residual

compositional data – 2014 estimates

Season No.

studies

%

studies

Quarter 1: April to June 32 24.8%

Quarter 2: July to September 31 24.0%

Quarter 3: October to December 39 30.2%

Quarter 4: January to March 27 20.9%

For the 2015 estimates, the coverage for Quarter 1 was lower compared to other

quarters (Table 34). Regression analysis showed that lower arisings in Quarter 1 were a

significant factor in the models produced, being associated with slightly lower arisings of

food waste. However it was not possible to arrive at a plausible hypothesis as to why

household food waste arisings would be lower at this time of year (April to June).

Consideration was given to stratifying the sample by season for the 2015 estimate. The

sample was stratified to test for this, but had minimal impact on the results40. Given this

40 Approximately 12,000 tonnes lower estimate for household food waste in UK kerbside residual waste in 2015; see Appendix 1.


and the increased complexity of the analysis if stratifying for seasonality, it was decided

not stratified by season.

Table 34: Seasonal coverage by number of phases of auditing for kerbside residual

compositional data – 2015 estimates

Season No.

studies

%

studies

Quarter 1: April to June 26 15.0%

Quarter 2: July to September 54 31.2%

Quarter 3: October to December 58 33.5%

Quarter 4: January to March 35 20.2%

A2.6 Number of samples included in each set of estimates

Table 35 compares the sample sizes for the 2014 and 2015 presented in this report, and

those for the previously published 2010 and 2012 estimates.

There is considerable overlap between the 2014 and 2015 estimates, since these are

both ‘pooled’ estimates that use compositional datasets either side of the calendar year

to which the estimate relates, in order to maximise the sample of compositional

datasets contributing to each estimate (see §2.2.2). 76 compositional datasets from 63

LAs were used in both the 2014 and 2015 estimates. The 2010 and 2012 estimates are

both independent of each other (i.e. no overlap in compositional datasets used to

calculated them) and of the 2014 and 2015 estimates.

Table 35: Number of waste compositional phases used for pooled estimates by year

Number of waste

compositional phases Number of LAs

2010 estimate 153 87

2012 estimate 82 63

2014 estimate 129 87

2015 estimate 173 116

The information in appendices 1 and 2 suggests that the practice of stratifying the

sample using the presence of collections targeting food waste is still justified for the

standard method. The analysis presented did not suggest the results of the standard

method would be substantially improved by further stratification or weighting.


Appendix 3: Detailed method for estimating food

waste in collections targeting food waste

There are four ways that a LA may have reported food waste tonnages. These are

described below and then illustrated in flow charts.

A. No tonnages reported under ‘waste food only’, ‘mixed garden and food waste’ or ‘other

compostable waste’

This suggests that there is no food waste being collected. To confirm this, the scheme

data is checked. If the scheme data is consistent with this reporting (i.e. it shows that

there is neither a separate food waste scheme in operation, nor is food waste accepted

as part of the garden waste collection), then the food waste tonnage can be confirmed

as zero.

For a very small number of LAs, there is an inconsistency: the scheme data shows that

food waste is targeted, but no organic tonnage is reported. In this case, the LA’s website

is checked to see what scheme is in operation. If the website says that food waste is not

targeted then the tonnage is confirmed as zero (and the scheme data assumed to be

wrong). If both the scheme data and the LA’s website say that food is targeted, then the

LA is contacted to obtain tonnages where food is targeted and the number of

households served. From here, the authority is treated as if they had reported those

tonnages in WasteDataFlow, using the method in part B.

B. A tonnage is reported for ‘waste food only’, and not for ‘mixed garden and food waste’

or ‘other compostable waste’

This suggests that the authority operates a separate food waste collection as its only

means of collecting food waste. If the scheme data confirms this – which is usually the

case – this tonnage should be used as the authority’s food waste estimate.

However, in a small number of cases there is an inconsistency, with a separate food

waste tonnage reported, but no separate food waste scheme described in the scheme

data. In this case, the LA is contacted. The LA may confirm that a separate food

collection is the only means of collecting food waste (i.e. the scheme data is incorrect),

and in these cases the reported tonnage is used as the food waste estimate.

In other cases, the LA may confirm that they operate a mixed garden and food waste

collection (sometime just this scheme, sometimes in combination with a separate food

waste collection). In either case, the omission of mixed food and garden waste from the

reported tonnages needs clarifying: the LA is asked to supply the tonnages and

household numbers for all schemes that target food waste. Any mixed garden and food

waste tonnages they supply are treated as per part C, and any separately collected food

waste tonnages are added to the estimate.

There may be rare cases where the reported food waste tonnage seems highly

unrealistic for the number of households that are served by the scheme. It is recognised

that food waste yields can vary quite substantially, and so a tonnage will only be

considered unrealistic if it is outside the wide range that would normally be expected

(anything less than 20 kg/hh/yr or more than 100 kg/hh/yr). Additionally if a scheme was


only in operation for part of the year then a lower yield would be expected (quarterly

tonnages can be used to confirm this).

Once a food waste tonnage has been identified as highly unrealistic for the number of

households, the LA is contacted to confirm that the number of households from the

scheme data is correct. If the tonnage still seems unrealistic – which is extremely rare –

then it is assumed to be inaccurate, and a tonnage is estimated using the modified food

waste ready reckoner (see below).

In rare cases the food waste tonnage for a quarter may be missing, when it is known

from the scheme data and LA website that the scheme is in operation throughout the

year. In this instance, the missing quarter’s tonnage is estimated using an average of the

other three quarters.

Modified food waste ready reckoner

The modified food waste ready reckoner applies an estimated yield per household, and

takes into account the socio-economics of the LA, and whether the scheme is only

serving flats.

The ready reckoner was initially derived from a predictive model for separate food waste

yields that was developed during the evaluation of the WRAP Separate Food Waste

Collection Trials41. The ready reckoner was updated for the WRAP LEN002-003 Modelling

Support Data and Assumptions project to use the percentage of households in social

grade D and E for the deprivation measure, rather than the Index of Multiple

Deprivation (IMD), as the social grade measure can be used consistently for LAs

throughout the UK.

The predictions of the updated ready reckoner were then checked against actual

performance of separate food waste collection schemes in the UK from 2013/14

WasteDataFlow (analysis carried out internally by WRAP and reviewed by Resource

Futures). This showed that on average, authorities were achieving 66% of the yield

predicted by the ready reckoner. WRAP have found that for individual authorities the

ready reckoner provides a good predictor of anticipated performance, so long as the

quality of the service and supporting communications are sufficient, and importantly

that free liners are provided to the householders. However in many instances this is not

the case, which probably explains the lower yields found in the WDF analysis. For the

purposes of the current project, the food waste ready reckoner has been modified by

the factor of 66% found in the WDF analysis.

C. A tonnage is reported for ‘mixed garden and food waste’ and/or ‘other compostable

waste’

The majority of these tonnages relate to mixed garden and food waste collections. The

scheme data is checked to confirm that this is the case. If so, then the tonnage of food

waste is estimated using either the kg per household or percentage arising method;

see below.

41 WRAP EVA037, Evaluation of the WRAP Separate Food Waste Collection Trials, Resource Futures, 2009


In some cases the scheme data may indicate that food waste is presented separately by

the householder and then mixed with garden waste. In these cases, it is considered

likely that in terms of amounts of food waste presented by householders, the scheme

will perform as if it were a separate collection of food waste (since householders are

interacting with and presenting food waste as a separate stream). This is likely to lead to

– on average – higher levels of food waste collected compared to a mixed food and

garden waste scheme. Therefore in these cases the modified food waste ready reckoner

is applied (see part B above).

Another check that is performed is to ensure that the tonnages are consistently

reported under one category (either ‘mixed garden and food waste’ or ‘other compostable

waste’) for all quarters of the year. It is not expected that tonnages would be similar for

each quarter, because the garden waste component would be larger in spring and

summer. However, if one or more quarters are missing, or data is reported

inconsistently between different categories then the data is treated separately. This

process is described in part D.

Kg per household method

For the mixed garden and food waste tonnages that have been confirmed to relate to a

mixed collection, and are reported consistently for all quarters, the food waste

component is estimated. This is calculated by applying a yield 44.1 kg/hh/yr for weekly

collections, and 24.0 kg/hh/yr for fortnightly collections. These figures were produced by

Resource Futures through combining data analysed as part of WRAP’s LEN002-003

project42 with additional kerbside organics that have been collated for the current

project. For future projects, they should be reviewed to ensure they reflect up-to-date

information.

Percentage arising method

In some instances the tonnage calculated by applying the yield above may be greater

than the reported tonnage for mixed garden and food waste. This is clearly implausible,

and so an alternative method must be used to determine the food waste component. A

percentage split is applied to the mixed garden and food waste tonnage; 27.0% is

assumed to be food for weekly collections, and 14.0% for fortnightly collections. These

assumptions have been derived from the data that formed the basis of the analysis for

WRAP’s LEN-003 project (see above), and may be updated on the basis of additional

kerbside organics that are in the process of being collated for the current project. Again,

they should be reviewed for future projects of this nature to ensure they reflect up-to-

date information.

D. Tonnages that may contain food are reported inconsistently between categories or

quarters

These cases require particular attention to establish what the tonnages relate to, by

cross-checking with scheme data and contacting the LA if necessary.

It may be that tonnages reported under ‘mixed garden and food waste’ and ‘other

compostable waste’ are reported in such a way that they are likely to refer to the same

42 WRAP LEN002-003, Material Splits in Co-mingled Recycling, Resource Futures, 2014


scheme. For this to be the case, tonnages would need to be of comparable magnitude,

whilst acknowledging the seasonal variations in garden waste arisings. In these

instances the food waste component would be calculated using the method described in

part C.

Another likely reason for inconsistent reporting between quarters is if a mixed garden

and food waste scheme was launched during the year, or if it is not in operation in

winter. The food waste component for these LAs can be estimated using the method

described in part C, but the yields will be adjusted to account for the length of time the

scheme is active. For example, the yield will be halved if the scheme is only active in two

quarters.

There may be occasions where there is no apparent reason for the inconsistent

reporting of tonnages. In such cases, the LA will be contacted to establish tonnage and

scheme details.


Obtain for all local authorities (LAs), and for

the required period:

WasteDataFlow (WDF) kerbside organics tonnages, all quarters

Organics scheme data

Are tonnages

for any of

these

categories

>0?

The WDF organics categories that potentially

contain food waste (FW) are:

Waste food only (WFO)

Mixed garden and food waste (MGFW)

Other compostable waste (OCW)

Does scheme

data indicate

that FW is not

targeted?

NO

FLOWCHART A

YES

Proceed to

FLOWCHART B

Assign zero FW

tonnage to the LA

Check the LA’s website

and/or contact the LA to

confirm if FW targeted at the

kerbside

Does the LA

target FW

at the

kerbside?

NO

YES

NO

Discuss with LA why WDF

tonnages do not show any

FW collected. Obtain data

from the LA on tonnages of

organics collected where FW

is targeted, and number of

households included on the

scheme(s).

Proceed to

FLOWCHART B,

although use LA

provided tonnages in

lieu of WDF tonnages

YES


FLOWCHART B

The LA has non-zero tonnages for one or

more of the following:

Waste food only (WFO)



Is a tonnage

reported only

for WFO, and

zero for MGFW

& OCW?

Does scheme

data confirm

that FW is

only targeted

via separate

collections?

YES

*

FW tonnage realistic for

number of households

served? Take into

account if scheme was

launched partway during

the year.

YES

Proceed to

FLOWCHART C

NO

Assign all of “Waste

food only” tonnage to

the LA

Contact the LA to obtain

details on the organics

collection scheme(s) which

target food waste.

Does the LA

target FW

only through

separate

collections? YES

Obtain tonnages from the

LA for organics schemes that

target FW.

NO

Contact LA to confirm

number of households

served with separate FW

collection.

NO

NO

Apply modified food

waste ready reckoner

to estimate FW

tonnage

YES


FLOWCHART C

The LA has non-zero tonnages for one or

more of the following:



Do the schemes

targeting food

waste include

separate

collections?

Estimate FW tonnages for the

separate FW scheme: go to

FLOWCHART B and follow the

instructions from the red

asterisk onwards *

YES

Food waste is also collected in mixed

organics and an estimate of food waste

collected in this scheme needs to be

calculated. FW collected separately &

misreported as MGFW/ OCW must be

deducted.

Is a tonnage

reported

consistently under

either MFGW or

OCW for all

quarters?

** If mixed

organics are

collected, is the

FW presented

separately by

residents?

NO

Proceed to

FLOWCHART D

NO

YES

NO

Apply kg per household

method to estimate FW in

mixed organics.

YES

Is the FW

estimate

greater than the

reported

tonnage of

mixed organics?

Apply percentage

arising method to

estimate FW in

mixed organics.

YES

Assign the FW

identified to arise in

mixed organics to the

LA’s FW tonnage

NO

Apply modified food

waste ready

reckoner to

estimate FW

tonnage


FLOWCHART D

The LA has reported mixed organics tonnages

that could include FW, but these have not

been reported consistently in one category

(MGFW) and/or across all quarters.

Are MFGW and OCW

tonnages spread

across quarters in

such a way that they

are likely to refer to

the same scheme?

Proceed to FLOWCHART C

and follow the instructions

from the double red asterisk

onwards **

If the scheme was launched

part way through the year,

account for this (by applying

the relevant percentage to the

tonnage estimated)

YES

Can the pattern of

MFGW / OCW tonnage

reporting be explained

by the launched of a

mixed organics scheme

partway through the

year?

Contact LA to confirm

scheme details.

YES

NO


Appendix 4: Scaling results from sample to national

level – detailed method

This appendix contains methodological details relating to scaling information from a

sample of LAs to the whole of the UK. Section A4.1 discusses the methods used to

account for differences in the coverage of LAs between the sample and the population;

in particular it deals with differences in yield for collections targeting food waste. Section

A4.2 details the exact method for obtaining the estimate of food waste in the residual

stream. Section A4.3 describes the method used for the ‘alternative’ method, the results

of which are presented in section A4.4.

A4.1 Accounting for differences between sample and population in the standard

method

A4.1.1 Stratification

The degree to which the collated waste compositional studies are representative of LAs

in the UK was assessed via regression analysis (Appendix 1) and coverage information

(Appendix 2). For the standard method, the sample and population were stratified by

presence of collections targeting food waste to ensure that any mismatch in coverage of

these collections between the sample and the population was accounted for. This

means that, for each stratum, the proportion of residual waste that was food was

calculated; this stratification approach therefore took into account the lower proportion

of food in residual waste for those LAs collecting food waste for treatment. This is in

keeping with previous studies.

This stratification was used when calculating food waste in the kerbside residual and

kerbside collections targeting food waste. It was not applied to calculations for kerbside

dry recycling and HWRC residual for two reasons. Firstly, there are too few data points to

allow effective stratification; secondly, the interaction in food waste between these two

waste streams and kerbside collections targeting food waste is not known (and likely to

be small). Given the minor contribution of HWRC residual and kerbside dry recycling to

household food waste estimates, this decision is unlikely to have a substantial impact on

the results.

From the testing of different stratification approaches in the 2010 and 2012 studies, it

was found that the most robust approach is to place LAs into two strata according to

whether or not they had collections targeting food waste. Each stratum contained

reasonable numbers of authorities and, furthermore, there was a large difference in the

amount (per household) and proportion of food waste in the residual waste between

these two strata. For example, for the LAs in the sample (for the 2015 pooled estimate)

with collections targeting food waste, there was on average 28.5%43 food waste in the

residual stream, or 116 kg / household / year; for those authorities in the sample

without collections targeting food waste, the corresponding figures were 36.5% and 183

kg / household / year.

43 Once the adjustment described later in this section was applied, the effective percentage was 31%. This adjusted figure is

presented in the results section.


Therefore, stratification helps the standard methodology account for the effects of

collections targeting food waste. Furthermore, stratification improves the precision of

the final estimates by grouping similar LAs together, reducing the variability between

those in the same stratum.

A4.1.2 Adjustment for differences in yield

After this stratification, one remaining effect is that the level of diversion of food waste

from the residual stream to collections targeting food waste varies greatly when

assessed per household (averaged over the whole LA). This could be for a number of

reasons including different coverage of collections targeting food waste within each LA,

and different participation and capture rates for those areas covered. However, the

stratification method above places all LAs that collect some food waste in targeted

collections in a single stratum, irrespective of the level of diversion.

A couple of options were considered to account for this effect. The first is to increase the

number of strata in the stratification, so that those authorities diverting some food

waste to targeted collections are further subdivided depending on the amount of food

waste diverted to targeted collections (assessed in kg / household / year). This was

explored in the 2012 study but not used because the number of studies in each stratum

was relatively small; the upshot was that small changes to the position of the boundaries

between strata could alter the results substantially.

A second method was developed to adjust for differences in diversion between the

sample (LAs with waste compositional analysis data) and the population (all LAs). This

method calculates the average amount of food waste in collections targeting food waste

for a) authorities with a waste compositional analysis, b) all authorities in the population.

This is possible because data on food in collections that target food waste comes from

WasteDataFlow rather than compositional analyses. By way of example, in 2015, the

amount of food in targeted collections was 49.1 kg / household / year for authorities

with a waste compositional analysis and 40.6 kg / household / year for all LAs. The

difference in these average yields was then used to calculate the additional amount of

food waste that was diverted to collections by those authorities with a waste

compositional analysis (compared to all UK authorities). For 2015, a difference of 8.5 kg /

household / year equates to 138,000 tonnes additional food waste collected by LAs in

the sample. This was then added to the amount in the kerbside residual waste to correct

for the underestimate (in the case of 2015).

Given the above, it was decided to adjust the data using the second of the two methods

(i.e. using the difference in yield to adjust the results), and this has been applied to the

2010, 2012 and 2014 studies.

The ‘alternative method’ (see A4.3) does not suffer from this effect as it sums the food

waste in the residual and targeted collections before extrapolating to LAs outside the

sample.

In addition to this diversion effect, there is the potential for collections targeting food

waste to affect the total quantity of food waste generated (e.g. a prevention effect).


However, the direct evidence for this effect is limited (see Appendix 1 for further

information44); there is considerable uncertainty about its magnitude and what factors

influence it (e.g. frequency of residual collections). Therefore, no changes to the

methodology were undertaken to account for this potential effect.

Given the methodology adopted, the analysis removes most distortions in the estimates

emanating from over-representation in the sample of LAs targeting food waste for

treatment. However, as stated above, tests were also performed for the current study to

determine if other factors are influencing the results: see Appendices 1 and 2.

A4.2 Methodological details for the ‘standard’ method

In the past, a number of slightly different methods have been used in the standard

method to scale information from the LAs with waste compositional data to the whole of

the UK for the kerbside residual stream. The different methods are detailed below,

alongside a rationale for why the method described in §2.4 has been used.

● Method 1a: Determine the average percentage of the residual stream which is food

(for each stratum) from the sample and multiple by the amount of residual waste for

the population (in that stratum). This gives equal weight to each LA.

● Method 1b: As above, but equal weight given to each phase (so additional weight

given to multi-phase studies).

● Method 1c: As above, but weighted by residual waste in LA.

● Method 2a: Determine the amount of food waste in the residual stream for each LA

in the sample first, and then extrapolating to the non-sample from the average

percentage (as in method 1a).

● Method 2b: As above, but equal weight given to each phase (so additional weight

given to multi-phase studies).

Although all of these have been calculated in the past, only methods 2a and 1c have

been reported as main estimates.

The main difference between the two groups (methods 1a-c and methods 2a-b) is that

for methods 2x, within the sample, information from the LA in question is used to

calculate the food waste in the kerbside residual and only use averages (derived from

the sample) for the non-sample LAs. In contrast, the 1x methods use averages from the

sample for sample and non-sample LAs alike.

The most appropriate method (1x or 2x) depends on the degree to which information

from a compositional analysis represents the whole LA in question. In general, sample

sizes for a single phase of field work are in the region of 135-250 households sampled

for one collection cycle (either one or two weeks). There is a large degree of variability in

the amount of food waste generated between households when sampled over such

short periods – some of this represents differences in the (long-term) average level of

food waste generated by different households; some is related to temporal variations in

the amount of waste generated by each household. For instance, previous analysis of

44Two other pieces of WRAP research focus on this topic:

Effect of food waste collections on arisings: recent evidence (WRAP, 2013):


Literature Review - Relationship between Household Food Waste Collection and Food Waste Prevention (WRAP, 2011):

http://www.wrap.org.uk/sites/files/wrap/Impact_of_collection_on_prevention_FINAL_v2_17_8_11.33a4f2d0.11159.pdf


http://www.wrap.org.uk/sites/files/wrap/Impact_of_collection_on_prevention_FINAL_v2_17_8_11.33a4f2d0.11159.pdf


data from LAs where waste compositional analyses had been performed multiple times

(e.g. in 2007 and again in 2010) showed low levels of correlation between the amount of

food waste found in different audits for the same authority.

The implication of these relatively small sample sizes and high variability is that, if the

amount of food waste in the kerbside residual waste stream for one LA is determined by

waste compositional analysis data from only that LA (i.e. methods 2x), the estimate may

be quite imprecise. Although this imprecision should be averaged out by combining

information from multiple LAs, it is, in effect, weighting the studies by the amount of

residual waste collected by each LA (which is highly correlated with the population of the

LA and the degree of diversion to recycling). There are likely to be much better factors to

weight the data (e.g. the sample size of the waste compositional analysis).

On the other hand, applying data from a waste compositional analysis to the LA in

question means that factors specific to that LA are taken into account – for instance, the

level of diversion of material away from the residual stream to dry recycling (which is not

accounted for elsewhere) will implicitly be taken into account using this method.

On balance, it was decided that method 1b was most appropriate for this analysis. Using

an equal weight for each phase of waste compositional analysis gives added weight to

larger (multiphase) studies. In addition, the scaling up method is not giving undue

weight to LAs with large amounts of residual waste. This method of scaling up is also

simpler than the other methods, and construction of confidence intervals is therefore

more straight-forward.

A4.3 ‘Alternative’ method for calculating UK estimates

The key difference between the alternative method and the standard method is that the

former determines the total food waste in the main kerbside waste streams (residual

and collections targeting food waste) for each LA before scaling this information to the

population. The alternative method essentially consists of:

● Calculating the amount of food waste per household in the kerbside residual waste

and collections targeting food waste fractions for each of the LAs included in the

study;

● Calculating the average arisings of food waste, in terms of kg / household / year,

across all the LAs included in the study; and

● Multiplying the mean arisings of food waste kg / household / yr by the number of UK

households.

Food waste in the kerbside dry recycling and HWRC streams is calculated separately and

added to the estimates for kerbside residual waste and collections targeting food waste.

This separation is necessary as there are too few studies with data for all four waste

streams to calculate a meaningful average.

For the 2010 and 2012 studies, there was no stratification of the sample for the

alternative method, as no factors were isolated during the course of that research that

influenced the amount of food waste collected at the kerbside (in the residual and

targeted collections combined). For this reason, no stratification of the alternative

method was applied to the 2014 and 2015 calculations either.


A4.4 Results and discussion of alternative and standard methods

The two methods (standard and alternative) were used to generate results to ensure

that artefacts from the calculations are not substantially influencing the results. The

results are presented in Table 36 & Table 37.

Table 36: Comparison of standard and alternative results for arisings of LA collected

household food waste in UK, 2015 (thousand tonnes)

Waste Stream Standard Alternative

Food waste in kerbside residual 4,117 4,021

Food waste in kerbside organics 639 639

Total food waste, kerbside

residual plus organics 4,756 4,660

95% confidence interval ±125 ±148

Table 37: Comparison of standard and alternative results for arisings of LA collected

household food waste in UK, 2015 (kg per person)

Waste Stream Standard Alternative

Food waste in kerbside residual 63.2 61.8

Food waste in kerbside organics 9.8 9.8

Total food waste, kerbside

residual plus organics 73.1 71.6

95% confidence interval ±1.9 ±2.3

The alternative method produced an estimate 2.1% lower for food waste in the kerbside

residual stream, which led to an estimate 2.0% lower for all LA collected food waste, in

comparison to the standard method. This is a larger discrepancy than for other years

(Figure 11) and was further investigated, including via regression modelling (Appendix 1)

and the coverage assessment (Appendix 2).


Figure 11: Comparison of standard and alternative pooled estimates of household food

waste arisings in the UK, tonnes; note y-axis does not start at zero (95% confidence

intervals shown for standard method)

The regression analysis suggests that there was a relationship between the amount of

amount of food waste produced by each LA, quantified in kg / household / year, and the

frequency of kerbside residual collections (lower levels if residual collection was

fortnightly) and deprivation level (less food waste for lower levels of deprivation). In

addition, there was some evidence, below the 95% confidence level, that collection of

food waste at the kerbside was associated with food waste levels.

The coverage assessment indicates that the sample is reasonably representative of the

UK for all of these factors with the exception of collection of food waste at the kerbside.

If stratification of the sample is applied to the alternative method to take into account

this over-representation of LAs with collections targeting food waste, then the difference

in results between the two methods (standard and alternative) reduces to 0.5%. This

suggests that, for 2015 at least, the standard method is the stronger and that the

alternative method may benefit from stratification. This should be considered for future

analyses.

However, the alternative method has the advantage that it directly takes into account

the diversion effect of collections targeting food waste. For the standard method to

function correctly stratification and adjustment is required to circumvent this issue.

Furthermore, the alternative method also takes into account the effect on total kerbside

residual waste of, for instance, different levels of kerbside dry recycling. (Currently, no

adjustments are made in the standard methodology to account for difference in total

amount of kerbside residual waste between the sample of LAs and the whole

population.)

On balance, the advantages and disadvantages of each method are of similar

magnitudes, and they produce similar results. Given the historic use of the standard

method in WRAP estimates of household food waste, and the use of the method by

research commissioned by Defra (e.g. WR0119), the results from the standard method

4.0

4.2

4.4

4.6

4.8

5.0

5.2

5.4

5.6

5.8

6.0

2007 2008 2009 2010 2011 2012 2013 2014 2015

Mill

ion

s to

nn

es

Standard (pooled)

Alternative (pooled)


are reported and used in other research by WRAP, although both have been included in

this report.

A4.5 Current and historic methodologies (comparison of estimates presented in this

report)

This section describes the difference between the methodology used in this study

(described above) and that used to obtain estimates for earlier years presented in this

report.

In short, the estimates of food waste in the kerbside residual waste stream – which

contains the majority of household food waste – presented in this report (§3.0) for the

following time periods were all calculated using the same methodology:

● Pooled estimates for 2010, 2012, 2014 and 2015

● All single-year estimates.

This is the result of improvements to the methodology being applied retrospectively to

historic data where possible.

For estimates of food waste within collections targeting food waste, the 2010 pooled

estimates and single-year estimates from 2006 to 2012 used an older method compared

to the 2012, 2014 and 2015 pooled estimates and 2013-2015 single-year estimates. The

more recent years used the method described in §2.4.2 of this report. The earlier years

used the method described in the 2010 and 2012 synthesis reports. The essential

difference was for food in mixed food and garden waste collections: the amount of food

waste was previously estimated by assuming that a certain percentage of the collected

material was food waste, based on analysis of compositional studies of mixed food and

garden waste collections. In the current study, the main approach to estimating food

waste in mixed food and garden waste collections is though applying an assumption for

the amount of food waste collected per household served in mixed food and garden

waste collections, with the assumption similarly being derived from analysis of

compositional studies of mixed food waste garden waste collections; see Appendix 3 for

a description of the methodology. There are only relatively minor differences in the

estimates produced by each method.

For the 2007 pooled estimate, the methodology was somewhat different. It was based

on an estimate produced for England for 2006/7, part of a Defra project (WR0119 project

– see glossary for more details). This estimate was used as there were very few

compositional studies for Wales, Scotland and Northern Ireland in the correct

timeframe.

With regard to whether LAs had targeted collections for food waste, the 2006/7 estimate

for England used neither stratification nor adjustment for yield. However, the proportion

of households with collections targeting food waste was very low and consequently the

amount of food collected in targeted collections was low – the capture rate was only

1.6% (see §3.3). Therefore, the impact of these two differences is likely to be very small.

The method to estimate food waste in collections accepting food and garden waste was

also based on different assumptions (as outlined in the report for WR0119) and the

amount of contamination (non-food material) in separate food waste collections was not


accounted for, but, again, as coverage of these collections was low, the impact on the

estimate is likely to be low.

As discussed previously, WR0119 only used multi-phase studies (§2.2.3) and did not split

out the individual phases of these studies (§2.2.4).

To derive an estimate of food waste for the UK for 2007, the percentage of household

waste collected at kerbside that was food waste for England in 2006/7 was applied to the

equivalent data for the UK in 2007. This assumes that food waste in England is similar to

the rest of the UK. Given the high proportion of the UK’s population in England, this

assumption is unlikely to have a large impact on the estimate of UK food waste.

As mentioned above, the similarity of pooled and single-year food-waste estimate for

the UK in 2007 suggests that, although the 2007 methodology is different, these

differences are unlikely to have substantially impacted on either the results or the

conclusions drawn from these results.

A4.6 Current and historic methodologies (comparison of previously published estimates)

Given that the methodology has been updated and improved over the years, it is useful

to document the differences in methodology between that used for this study and that

used in previously published studies. These differences explain the slight changes

between figures previously published for 2010 and 2012 and those estimates found in

this report for those years.

2010 study

The UK estimate for 2010 in this report (kerbside residual and collections targeting food

waste) is 4,615,000 tonnes, 142,000 tonnes higher than the figure previously published

(4,473,000 tonnes)45. This difference is due to the following differences in methodology:

● The previously published estimate calculated food waste for each nation and then

totalled these estimates (i.e. stratifying by nation). The current methodology does

not stratify by nation. (This is because, for some years, there is very little data for

some nations, so no robust results are possible.)

● The previous methodology did not split out the phases within multi-phase waste

compositional analyses. This impacted on the method to scale from the sample of

LAs with waste compositional analyses to the whole nation was different. Details of

the different methods are given in Appendix 4.2. The 2010 study (as published) used

method 1c; the current methodology (and updated 2010 estimate) uses 1b.

● No adjustment was made for yield in the previously published method (see Appendix

4.1.2); this has been applied in the current methodology.

● The assumptions used to estimate food waste in mixed food and garden collections

has been refined since the 2010 study.

● The amount of contamination (non-food material) in separate food waste collections

was not accounted for in the 2010 study as published, but now is accounted for.

45 http://www.wrap.org.uk/sites/files/wrap/Synthesis%20of%20Food%20Waste%20Compositional%20Data%202010%20FINAL.pdf



2012

The UK estimate for 2012 in this report (kerbside residual and collections targeting food

waste) is 4,577,000 tonnes, 3,000 tonnes higher than the figure previously published

(4,574,000 tonnes)46. This slight difference is due to the following differences:

● In the original analysis, equal weight was given to each LA (method 1a in Appendix

4.2); the current methodology uses 1b (equal weight to each phase of waste

compositional analysis).

● The assumptions used to estimate food waste in mixed food and garden collections

has been refined since the 2012 study.

● Quarterly WasteDataFlow data was used, matching residual waste data for the

quarter that the waste compositional analysis fieldwork was undertaken. This was

not possible for 2014 and 2015 as some nations stopped releasing quarterly data.

● The amount of contamination (non-food material) in separate food waste collections

was not accounted for in the 2012 study as published, but now is accounted for.

46 http://www.wrap.org.uk/sites/files/wrap/Synthesis%20of%20Food%20Waste%20Compositional%20Data%202010%20FINAL.pdf



Appendix 5: Single-year estimates

The main estimates presented in §3.0 are ‘pooled’ estimates, as described in §2.4. These

include waste compositional studies from outside the target time period. The rationale

for including studies from either side of the target year is that this substantially

increases the number of compositional studies included in the analysis, which reduces

the size of the confidence intervals around the estimates.

This appendix presents ‘single-year’ estimates from 2006 and 2015, as a different way of

obtaining trends in national household waste arisings over time. These single-year

estimates were derived from waste compositional analyses carried out during the year

in question. This reduces the number of waste compositional analyses available to make

an estimate for any given year, but ensures that information used is more temporally

relevant to the year in question.

The ‘single year’ estimates are presented in Figure 12 & Figure 13. These include

household food waste in kerbside residual and collected at the kerbside for treatment.

Arisings are expressed in kg per person, in order to control for the increase in

population over the period in question. Estimates are included using both the ‘standard’

and ‘alternative’ methods (see Section 2.4), with 95% confidence intervals included for

the single-year standard method estimates. In addition, the standard pooled estimate is

included for comparison.

Figure 12: ‘Single year’ estimates of household food waste arisings in the UK, 2006 to

2015, tonnes; note y-axis does not start at zero

Figure 12 illustrates that the results of the single-year estimates are relatively close to

those of the pooled estimates: the standard pooled estimates are within the confidence

intervals of the single-year standard estimates.

4.0

4.2

4.4

4.6

4.8

5.0

5.2

5.4

5.6

5.8

6.0

2006 2007 2008 2009 2010 2011 2012 2013 2014 2015

Ton

ne

s M

illio

ns Standard 'single-year'

Alternative 'single-year'

Standard pooled


Figure 13: ‘Single year’ estimates of household food waste arisings in the UK, 2006 to

2015, kg per person; note y-axis does not start at zero

It should be noted that, for some years – notably 2008 and 2013 – there are a small

number of data points within one of the strata for the single-year (standard) estimates,

as shown in Table 38. This helps to explain the wide confidence intervals for the

standard method estimates for 2008 and 2013. The standard method estimate for 2013

is notably higher than the alternative method estimate, and this is at least partly

explained by the small number of authorities in the strata for food waste not being

targeted for collections (only 8 authorities), with these authorities having relatively high

food waste arisings. Therefore the higher arisings for the standard method estimate in

2013 are most likely explained by a combination of small sample size for one of the

strata and scatter in the data. For years with a low number of studies available, the

advantages of pooling studies across multiple years are likely to outweigh the

disadvantages – hence the use of the pooled estimates in the main report, with the

single-year estimates being used as a shadow measure to check for any artefacts

relating to pooling the data.


Table 38: Sample sizes for ‘single year’ estimates using the ‘standard method’, by strata

Year No. of LAs in sample:

collections targeting

food waste

No. of LAs in sample:

collections not

targeting food waste

2006 21 57

2007 16 115

2008 3 21

2009 42 35

2010 33 19

2011 31 19

2012 21 12

2013 47 8

2014 44 29

2015 69 15


Appendix 6: Uncertainty

This appendix lists and discusses the uncertainties associated with the estimates of

household food waste presented in the main report. This list includes both random and

systematic errors. Random errors are largely associated with only sampling a small

proportion of the total population of households and lead to imprecision in the

estimate. Systematic errors are where an aspect of the methodology may lead to a bias

(i.e. net over- or under-estimate) in the estimate.

Where possible, quantification of these uncertainties has been made and the details of

this information, alongside how the confidence intervals were calculated, are presented

in this section.

Sampling of households within LA compositional analyses: Within individual waste

compositional analysis studies, it is not practically feasible or desirable to perform

compositional analysis on all households in a given LA, so samples were drawn. Random

errors in the results from each waste compositional study contribute to the variability

observed between LAs, which in turn influences the confidence intervals presented in

this report. If households are not selected randomly (which is often the case for practical

reasons), there could also be a systematic error if there is a bias in the selection of

households. This systematic uncertainty has not been accounted for in this report and –

should it exist – would be very difficult to quantify. These sampling issues are also

applicable to Household Waste Recycling Centres (HWRCs) and contamination of dry

recycling by food waste. In both cases, the random sampling errors have been estimated

and reported.

Having waste composition analyses from only a sample of LAs: Having data from a

sample of LAs (rather than the entire population) introduces random sampling error into

the calculations. The degree of sampling error has been quantified in this study using

confidence intervals, which take account of both the genuine variation in waste

composition between LAs and the additional variation caused by sampling households

within LAs (see previous uncertainty).

Representativeness of LAs covered by compositional analysis: There are many

factors that may influence whether a LA performs a waste compositional analysis.

Analysis of the data suggests that LAs that performed a compositional analysis included

in this synthesis were more likely to collect food waste for treatment. This factor has

been adjusted for in the calculations by stratifying the population and sample between

those authorities collecting food waste in targeted collection and those not. Adjustment

has also taken into account a difference in the yield of LAs targeting food waste with and

without a waste compositional study. Other potential differences between the LAs

covered by compositional analyses and those in the UK include the frequency of residual

collections, rurality and deprivation. The effect these could have on the results is

explored in appendices 1 and 2.

The ‘alternative method’ of scaling up from the sample to the population provides some

analysis of sensitivity to these issues. The results of the alternative and standard

methods for the UK in 2015 differed by around 2%.

Total quantity of waste in relevant streams (reported in WasteDataFlow):

Information reported to WasteDataFlow undergoes many checks before it is made


public. However, it is possible that measurement, classification and reporting errors

occur, which would feed through to the results in this report. It is not possible to

quantify the magnitude of these potential errors.

Estimate of food waste in mixed organics collections: The proportion of food waste

in mixed food and garden collections is highly variable and could depend on a range of

factors including how the collection is communicated to households and the length of

time since its introduction. This data is only available for a few authorities so it was

necessary to make assumptions for the other authorities with such collections, which

could have introduced a systematic error into the results (see Appendix 3 for method of

determining food waste in mixed collections). No correction has been made for this

uncertainty, but the overall impact is likely to be low.

Estimate of food waste in separate food waste collections: The analysis took into

account a small level (2%) of contamination in separate food waste collections. This is an

improvement on previous studies which assumed no contamination. This affects the

results by less than 10,000 tonnes (or 0.2%).

Fines: Food waste is likely to be present in the fraction of waste classified as ‘fines’ (i.e.

small particles of waste). In the original 2007 estimate, an estimate of fines was included

in the results (around 70, 000 tonnes). However, as it is difficult to find data on the

proportion of fines that is food waste, it has not been included in subsequent research

and has been stripped out of the 2007 estimate to allow comparison. The omission of

fines from the estimate means total food waste may have been slightly underestimated,

but that comparison over time is more robust. There is the possibility of a bias, but its

magnitude is likely to be small.

Seasonal variation: There was a near uniform distribution of studies by season and

hence the influence of seasonal variation on the results is likely to be small.

Methodology of waste compositional analyses: Differences have been found in the

regression modelling between the amounts of food waste measured by two groups of

contractors (see Appendix 1.1). It is likely that the contractor effect relates to differences

in methodological approach or thoroughness in separating food waste from other

materials when carrying out waste audits. However, it has not been possible to establish

the exact cause.

A sensitivity analysis was performed to understand the potential magnitude of this

effect (see A1.1). This gives some indication of the likely level of uncertainty stemming

from which contractor performs the analysis – somewhere in the region of 180,000

tonnes of waste in the UK in 2015, or c.4% of the total for the residual waste stream. It is

important to note that this factor has a smaller impact on trends over time for UK

household food waste, if the contractor effect is assumed to be present to a similar

extent for all studies.

No adjustment has been made to the estimates in relation to this uncertainty. To

mitigate this uncertainty in future, it is worth the relevant actors considering whether a

standardised approach to waste compositional analysis would be practical, as such a

change would benefit those using the resultant data.


Appendix 7: Peer review report

Technical peer review of CFP302-001: Synthesis of UK waste compositional data

2014-2015

An independent peer review was undertaken of Synthesis of Food Waste Compositional

Data 2014 & 2015, with the goal of ensuring that the methodology is statistically valid

and that the results stand up to technical scrutiny.

WRc first reviewed a draft methodology (v.6, 26th May 2015) and potential issues raised

by the review were discussed with WRAP to identify and agree appropriate changes. Two

draft versions of an interim report covering the 2014 results (v9, 16th July and v15, 19th

August 2015) were then reviewed, focusing on the presentation and interpretation of

the results, and a draft final version (v16, 1st September 2015) was circulated by WRAP so

that WRc could verify the edits that had been made. Following further analysis by

Resource Futures and WRAP to generate the 2015 results, WRc reviewed a draft report

(v2, 20th July 2016) and a draft final report (v10, 7 September 2016) prior to finalisation.

The main issues raised by the peer review process, and subsequently addressed by the

report authors, are detailed in the tables below.

The authors have made great efforts to select and apply appropriate statistical

techniques, and provide a clear, detailed and transparent justification for the methods

used. Notably, the representativeness of the waste composition studies has been

assessed, potential sources of bias have been identified and adjusted for where

possible, the results have been presented together with a measure of precision and

confidence, and sensitivity analysis has been used to test the influence on the results of

key assumptions. The report provides an accessible account of the study’s findings and

the conclusions are supported by the available evidence.

In summary, I am satisfied that the research presented in this report provides reliable

and up-to-date estimates of the quantity of food waste collected by (or on behalf of)

local authorities from UK homes.

Dr Andrew Davey, WRc plc

Peer Reviewer, 12/09/16

Name of document/s

reviewed:

CFP302 Household Food Waste Data Synthesis Methodology (v6a)

A. Issue to be

clarified

B. Thoughts or concerns of

peer

reviewer

C. Response from report

author

D. Response from

peer reviewer

1 Objectives The specific objectives of the

study should be stated at

beginning of methodology.

The paragraph with the aims

has been rewritten. It now

includes mention of the

national comparisons,

comparisons over time and

capture rates.

The full scope of the

project is now

conveyed in the

introduction.

2 Scope The report should clarify which This has been added in the This now dealt with


food waste streams (e.g. street

sweepings, waste to sewer and

home composting) are in and

out of scope.

text. comprehensively in

the Introduction.

3 Baseline

period

The baseline period should be

clearly defined to set the

scope of the study before

going to on to discuss the

extent to which the

component datasets fit with

the baseline.

The baseline period for wave

2 is January-December 2014

and this is clarified in the

text.

Required

information added.

4 Selection

criteria for

WCA studies

The strategy used to identify

candidate WCA studies should

be explained to give readers

some assessment of how

comprehensive the search has

been, and an understanding of

any factors that may have

limited access to WCA results.

Contact was made with all

the main contractors

undertaking WCA studies in

the UK. Some LAs decline

permission to use their

results. The exact number of

WCA studies (and phases)

identified and shortlisted will

be tabulated, with a

breakdown by UK nation and

type of waste stream

examined. This is described

in the text of the report.

The “vast majority”

of WCA studies are

believed to have

been sourced by

the project, but the

total number of

available studies is

not reported (or not

known?).

5 Selection

criteria for

WCA studies

The time span for WCA studies

used to derive the pooled

estimates should be clarified

and justified.

For the 2014 pooled

estimate, WCA studies

undertaken between April

2013 and March 2015 will be

used. The choice of two

whole financial years

provides a clear cut-off and

permits valid comparisons

with the 2010 estimate for

the purposes of evaluating

WRAP’s impact.

WCA studies undertaken

January-March 2015 will also

be used in wave 3 to

generate a 2015 estimate, so

direct comparisons between

the pooled 2014 and 2015

results will not be made.

This is now described in the

text.

Clarified in v15 of

report.

6 Selection

criteria for

WCA studies

It would be helpful to have

more information about the

type and level of socio-

demographic stratification that

WCA studies had to have

undertaken to be included in

the synthesis.

WCA studies simply had to

show some evidence of

stratification using a

recognised socio-

demographic classification

system. Further detail will be

added to the report.

Examples of

stratification system

added as a

footnote.


7 Comparability The draft report should

highlight any refinements

made to the methodology and

the extent to which these

affect comparisons with

previous (2010, 2012)

estimates of food waste

arisings.

This will be explored, either

by applying the revised

method retrospectively, or

by testing the sensitivity of

the 2014 results to the

choice of method.

The effect of

methodological

choices is explored

in detail in

Appendix 4.

8 Splitting multi-

phase studies

Clarification required that

splitting multi-phase studies

into their individual phases

shouldn’t introduce any bias

(i.e. systematic over- or under-

estimation), but that non-

independence of individual

phases could mean that

estimates of precision are

slightly optimistic.

This is clarified in the text. Clarification added.

9 Report

structure

The detailed discussion of

alternative methods is

excellent, but the final report

could be made more readable

by presenting the preferred

approach in the main text and

discussing other alternatives

as an appendix.

The final report includes an

appendix exploring the

sensitivity of the results to

the choice of method.

Detailed

description,

discussion and

justification of

methods have been

moved to

appendices.

10 Scaling from

sample to

population

Six possible methods are

described for scaling up to

obtain an estimate of food

waste in the kerbside residual

stream for the UK from the

sample of local authorities

with waste compositional

analysis data. Different

methods have been used to

produce previously reported

figures. It appears that some

of these methods (e.g. 1c and

2c) are directly equivalent. A

convincing argument has been

made for splitting multi-phase

WCA studies but the logic for

calculating a weighted average

estimate of %FW is less clear.

Weighting by residual waste

arisings does not reflect the

quality of the WCA study.

Weighting by number of

households surveyed may

provide little improvement in

the food waste estimate

because of the limited

variation among WCA studies

and the law of diminishing

returns. There are strong

Method 1b will be taken

forward at this stage

because it follows a simple,

statistically-based random

sampling logic, allows

confidence interval to be

calculated easily, and can be

readily extended to include

stratifying factors. This is all

covered in the text of the

appendix.

In addition, the duplication

of method 1c and 2c has

been removed from the

report.

Choice of method

now explained and

justified in

Appendix 4.


benefits in choosing a method

that is simple to understand

and easy to calculate.

11 Contamination

in food waste

collections

Previous studies have

assumed 100% food waste in

separate FW collections, but

available evidence suggests an

average contamination rate of

1-2%. Consideration should be

given to making some form of

adjustment to correct for this

low-level contamination.

We now adjust for

contamination – explained in

the text.

Now explained in

Section 2.4.2.

12 Mis-reporting

of food waste

tonnages

For a very small number of

local authorities, there is an

inconsistency between the

WRAP scheme data, which

shows that food waste is

targeted, and the WDF

tonnages, where no organic

tonnage is reported. Is this

due to non-reporting or mis-

reporting, and could this lead

to some food waste being

double-counted?

This situation occurs

infrequently and is almost

certainly due to food waste

being mis-reported under

another organics category

rather than as part of the

kerbside residual waste

stream. This will be clarified

in the report.

Explained in

Appendix 3.

13 Unrealistic

food waste

tonnages

Could partial coverage of

separate food waste

collections potentially explain

why reported food waste

tonnages occasionally seem

very low for the number of

households in the local

authority?

No; partial coverage is taken

into account using

information from WRAP’s

scheme data. This is now

clarified in the report.

Clarification added

14 Food waste in

mixed

collections

Are there any other WCA

studies that could provide

more up-to-date evidence of

the food waste composition of

mixed collections?

As detailed in appendix 3,

the most up-to-date factors

were used.

Detailed in

Appendix 3.

15 Confidence

intervals

Further detail should be

provided on the calculation of

confidence intervals.

Uncertainty is discussed in

appendix 6, alongside

section 2.6.

Covered in

Appendix 6.

Name of document/s

reviewed:

CFP302 Synthesis of Food Waste Compositional Data 2014 – Draft report

(v9)

A. Issue to be

clarified


peer

reviewer


author

D. Response from

peer reviewer

1 Coverage

assessment

Comparing the average %food

waste across the four quartiles

of each factor would provide

further information on the

adequacy of the sample

coverage. This would also be a

The regression results and the

coverage assessment are now

cross-referenced in the text.

The report has also been re-

organised to present the

regression analysis first. A

This strengthens

the justification for

the choice of

stratifying factors.


useful place to refer to the

results of the regression

analysis. A short conclusions

section would be helpful,

summarising and justifying the

choice of stratifying factors.

short conclusions section has

been added.

2 Confidence

intervals

Some readers will be puzzled

as to why the precision on the

food waste total isn’t a sum of

the precision figures for

individual waste streams. Need

to explain clearly in the

methodology, or with the use

of a footnote, that the errors

for each waste stream are

assumed to be independent,

and therefore balance each

other out at some extent.

Foot note added. Footnotes added

to relevant tables.

3 Presentation

of results

The results are much easier to

interpret when presented in

graphical form… are tables

needed as well?

Although I’d normally

recommend a parsimonious

report without undue

repetition, I think in this

instance presenting

graphically and in a table is

useful for comparison.

However, I think the graphical

representation has the most

impact, so I’ve moved that

first.

Agreed.

4 Capture rates Can the report clarify which

waste streams were used in

calculating capture rates?

Explanatory sentence added. Now clarified.

5 Regression

analysis

“Percentage of food waste in

kerbside residual” could be

mis-read as “Percentage of all

food waste that is found in the

kerbside residual waste

stream”

Report reworded to avoid this

ambiguous statement.

Now clarified.

6 Regression

analysis

It needs to be clear what your

model selection procedure

was. It’s fine to leave all the

independent variables in, if you

wish; alternatively, the analysis

should seek to find the most

parsimonious model, using a

defined model selection

criterion (e.g. adjusted r2 or

AIC).

The model reported contains

all the explanatory factors

used in the modelling.

Although some models were

constructed with fewer

explanatory factors, there was

not a systematic search for a

more parsimonious model.

Hence, the decision to report

the ‘full’ regression model.

The text of the report has

been amended accordingly.

The report now

focuses on

presenting and

interpreting the full

regression model.

7 Regression

analysis

A low r2 value does not

necessarily mean that the

results need to be interpreted

Agreed, and have amended

text.

The revised text

makes it clear why

the proportion of


with caution, as the p-values

tell us which factors are and

aren’t important. Rather, it

means that there are other

unmeasured influences on

%FW and/or uncertainties in

the WCA and WDF figures.

explained variation

is expected to be

low.

8 Regression

analysis

The reference category should

be made clear when

interpreting categorical

explanatory factors.

Text amended to draw clear

comparisons with the

reference group.

This has been

clarified.

9 Regression

analysis

Need to be careful not to over-

interpret non-significant

results.

Text amended. A more cautious

interpretation is

now provided.

Name of document/s

reviewed:

CFP302 Synthesis of Food Waste Compositional Data 2014 – Draft report

(v15)

A. Issue to be

clarified


peer

reviewer


author

D. Response from

peer reviewer

1 Standard

method

The detailed description of the

standard methodology is

complex and difficult to follow

in places.

Appendix 4 has been

reworked – it now contains an

introduction, additional

headings and much reworked

text.

Re-wording and

addition of sub-

headings has

improved the clarity

of the Appendix.

2 Presentation

of results

The 2012 study included a

pooled estimate for 2009 –

why hasn’t this been presented

also?

Due to resource constraints

in the project, it was not

possible to recalculate the

2009 estimate with the

improvements to the

methodology made for this

report. A footnote has been

added to explain this.

This is now

explained.

3 Confidence

intervals

The report notes that the CIs

measure the uncertainty

arising from sampling error,

but it would also be worth

listing briefly the sources of

uncertainty which are not

measured by the confidence

intervals.

Text has been added to the

main part of the report.

The sources of

uncertainty

measured and not

measured by the

confidence intervals

is now clear.

4 Confidence

intervals

Some errors were noted in the

calculation of confidence

intervals for 2009 and 2014

Scotland pooled estimates.

The calculation of other

confidence intervals presented

in the report should be

checked.

All the confidence-interval

calculations have been

checked and some (minor)

changes have been made to

the report. In addition, there

are some minor changes to

the estimates due to some

minor changes in the

underlying data. None of

these alters the conclusions

drawn from the report.

The confidence

intervals have been

updated.


5 Conclusions The report requires a

Discussion / Conclusion

section after the Results.

This will need to be written in

light of other information that

is published at the same time

as this report. This additional

information is likely to

present explanations for the

trends seen and it would be

useful for the discussion

section to cross- reference

any additional information.

This will be drafted when it is

known what else is being

published.

Acknowledged.

Name of document/s

reviewed:

CFP302 Synthesis of Food Waste Compositional Data 2015 - Draft Report

v2

A. Issue to

be clarified

B. Thoughts or concerns of peer

reviewer


author

D. Response

from peer

reviewer

1 Choice of

stratifying

factors

The regression analysis suggests

that the average %food waste in

kerbside residual is lower in

Wales than in other UK nations.

The over-representation of Welsh

local authorities in the sample

could therefore bias the results.

The most likely explanation

is that higher capture rates

of FW collections in Wales

are leading to lower % of FW

in the Welsh residual stream

(compared to LAs in the rest

of the UK with FW

collections). This effect is

already being accounted for

(via adjustment factor for

differences in yield

depending on

presence/absence of FW

collection), so no need to

stratify by country. Text

explaining and justifying this

decision has been added to

the regression modelling

appendix.

An explanation

for this result is

now included and

a justification

provided for not

stratifying by

nation.

2 Contractor

differences

The regression analysis suggests

that that the average %food

waste in kerbside residual

depends upon the identity of the

contractor who undertook the

compositional analysis (after

controlling for other factors).

Should an adjustment be made

for these potential differences

between contractors?

We can’t be sure which

contractors’ results are most

accurate – i.e. whether some

figures are too high or the

rest are too low. This effect

does not appear to be

explained by the presence of

packaged food as a category.

The decision was made to

exclude this variable from

the main regression analysis

presented, but to discuss the

issue in the regression

modelling appendix and

uncertainty log. Sensitivity

analysis has been used to

quantify the magnitude of

Agreed that this

is a sensible and

pragmatic

solution.


the uncertainty on the 2014

and 2015 estimates.

3 Regression

modelling

Factors such as Year and Season

should be treated as a single

explanatory variable to determine

whether there is evidence of a

significant overall effect on the

response variable. If there is a

significant effect, then a full set of

dummy variables can then be

used to explore where the

differences lay, relative to a

reference year or season.

This has been undertaken in

the regression modelling.

The report has been

updated accordingly

(appendix 1).

Categorical

factors are now

treated

appropriately in

the regression

models.

4 Confidence

intervals

Confidence intervals should be

quoted for all headline results,

notable the estimated kerbside

capture rate.

This has been resolved –

confidence intervals have

been added to the headline

results, both in the text and

in the figures.

Confidence

intervals now

presented clearly.

5 Scope Pooled estimates for 2014 were

calculated in Wave 2 of CFP302

project but not published. Rather

than refer to ‘previously

unpublished results’ it would be

simpler to say that this report

estimates arisings for 2014 and

2015.

The report has been

updated.

Scope of report

clarified as being

2014 and 2015.

6 Definition of

food waste

It would be helpful to explain how

the EU FUSIONS Project’s

distinction of edible vs inedible

food waste compares with

WRAP’s own

avoidable/unavoidable food

waste terminology.

Added a footnote to clarify in

terms of WRAP’s

terminology.

Terminology

clarified.

7 Statement of

methods

Presentationally, there might be

value in including, somewhere, a

table that compares the key

aspects of the 2010, 2012 and

present studies – i.e. provides at a

glance, the key similarities and

differences. At present, contrasts

are drawn at various points

throughout the report; it would

be good to have a consolidated

assessment of the comparability

of the results from the three

studies. Details of the 2010 and

2012 methods could then be

relegated to an appendix.

A new section (in Appendix

4) has been added. Most of

the comparison is in the text,

but a table has also been

added to show where data

has come from. This also

includes information for

point 11 on comparisons

with 2007. There is also a

comment in the appropriate

place in the results section.

New section in

Appendix 4

added to explain

how the

methodology has

evolved, and

which results are

comparable.

8 Terminology The term ‘baseline’ is used in two

different contexts – to refer to the

focal year of the pooled

estimates, and also to describe

the original 2006/7 estimate. Can

This has been updated –

baseline is now only referred

to with respect to Courtauld

targets (and in the peer

review statement).

Clarified.


this be clarified?

9 Calculation

method

For improved clarity, Section 2.4

could be restructured to have

four sub-sections, one for each of

the four waste streams analysed.

Clarification needed as to

whether or not the kerbside dry

recycling and HWRC residual

streams were stratified.

This change has been made. Text re-

structured for

clarity.

10 Trend testing A brief description of the

statistical methods used to

compare results between years

should be included at the end of

the Methodology.

Added to relevant section. Methods now

detailed.

11 Comparison

with 2007

Need to be very clear about

whether meaningful comparisons

can be drawn with 2007. If

comparisons are to be drawn

then, then the 2007 similarities

and differences between this

study and the 2007 study need to

be highlighted and discussed. If

not, then consider omitting all

references to 2007.

See comment for no. 7. Discussion added

to the report.

Name of document/s

reviewed:

CFP302 Synthesis of Food Waste Compositional Data 2015 - Draft Final

Report v10

A. Issue to be

clarified


peer

reviewer


author

D. Response from

peer reviewer

1 Nation

comparisons

When comparing nation X to

the UK average, need to clarify

whether or not nation X is

included in the UK figures

(Section 3.2.2, 3.2.3). If it is,

then it is not appropriate to

test the significance of the

difference.

Previously there was overlap

and these have been

amended. Now, the

comparison for Wales is now

with England, to avoid this

overlap. For London, I’ve

compared to England (minus

London), again to avoid

overlap. In both cases, the

broad conclusion is the same.

Fair comparisons

are now made.

2 Capture rates Details of how capture rates

were calculated could be

presented in the Methodology.

Given the unusually high

capture rates in Wales,

consideration should be given

to stratifying the capture rate

calculations by nation,

particularly if formal

I don’t think they should be.

The capture rates are

calculated from the estimates

of:

a) food waste in the

kerbside residual,

where effects relating

to coverage of

collections and higher

Agreed.


comparisons are to be made. capture rates are

taken into account via

the stratification and

adjustment factor

respectively.

b) Food waste in

collections targeting

food waste, which is

calculated for all local

authorities, and so

has complete

coverage.

Therefore, we don’t think that

the differences in capture

rates in any nation will be

affecting the results.

3 Capture rates Can reasons for the higher

capture rate in Wales be fully

explored? (Section 3.3)

Text has been added to the

Executive Summary and

section 3.3 to provide further

information on this effect.

The reasons for

the higher capture

rate in wales are

now discussed.

4 Time trends I suggest Section 2.7 could be

called “Trends in food waste

arisings”, and could simply say

that there have been changes

to the estimation methodology

over time, that previously

published pooled estimates for

2010 and 2012 have been

updated using the current

(2014, 2015) methodology, and

that comparisons are drawn

with 2007 even though there

are some small methodological

differences. This would avoid

the need for numerous

footnotes at various points in

the report.

Detailed comparisons of the

methodologies should be

reserved for the appendix.

Section 2.7 has been reworked

in line with these suggestions.

The detailed discussion is now

found within the appendix,

with a brief summary in

section 2.7, alongside the table

summarising similarities and

differences.

Report now

provides a clear

and succinct

assessment of the

comparability of

different studies.

www.wrap.org.uk

Synthesis of Food Waste Compositional Data 2014 & 2015 · 2019-05-09 · Synthesis of food waste...

Documents

Transcript of Synthesis of Food Waste Compositional Data 2014 & 2015 · 2019-05-09 · Synthesis of food waste...