Bags of money or bags of Impulsiveness? Psychological and Demographic Predictors of Plastic Bag Consumption in Big Data

Abstract

Despite the success of plastic bag charges in the UK, there are still around a billion single-use plastic bags bought each year in England alone (DEFRA, 2018), and politicians are currently debating whether to double the 5 pence charge to 10 pence (Rawlinson, 2018). Previous research has identified motivations for bringing personal bags to the supermarket (Jakovcevic et al., 2014; Poortinga et al., 2016) but little is known about whom is continuing to regularly purchase plastic bags after the levy. In this study, we harness the power of big data to investigate who the regular plastic bag buyers are that future campaigns ought to target.

We utilise a dataset of 12,968 questionnaire responses measuring demographics and individual differences matched to respondents' loyalty card transaction data from a UK health and beauty retailer. We used the transaction data to identify regular single-use plastic bag buyers and non-bag buyers in the real world. We optimised different prediction models and evaluated them using out-of-sample testing. A logistic regression model performed the best, predicting regular bag buyers with 76% accuracy.

The results showed that England had significantly less bag sales than Wales, Northern Ireland, and Scotland per 1000 items in the 9 months after the English levy. Across the UK, more bags per 1000 items were bought during the Christmas festive period. Shopping frequency, amount spent, age (negatively), frugality (negatively), income, impulsivity, a present time focus, and Openness significantly predicted regular bag buying. Notably, climate change concerns did not. In summary, this research has identified a young, high-income demographic of regular bag buyers to target. The additional insights that this sub-population are impulsive and unconcerned with climate change also helps to inform the content type and communication style for future plastic bag reduction campaigns.

Full text

Bags%of%money%or%bags%of%
Impulsiveness?%
Psychological%and%Demographic%
Predictors%of%Plastic%Bag%
Consumption%in%Big%Data
Rosa%Lavelle-Hill

Introduction
➢Psychology%and%N/LAB
➢Big%data%not%just%for%profit,%but%for%social%good%(e.g.%Health,%Environment)
➢10%years%of%Boots%loyalty%card%data%
➢234,938,722%baskets,%%and%718,105,269%items%bought
➢12,968%fully%completed%personality%questionnaire%responses

Plastic%Bags

Who%are%the%“culprits”%of%regular%plastic%
bag%consumption?
➢7.6%billion%plastic%bags%per%year%in%English%supermarkets%
(GOV,%2015;%WRAP,%2015).%
➢On%average%140%bags%per%person%(GOV,%2015)
➢5p%levy%introduced%in%2015
➢Still%a%billion%single-use%plastic%bags%bought%each%year%
(Defra,%2016).%
➢Who%is%still%regularly%buying%plastic%bags?
➢Will%a%10p%levy%help?%(Guardian,%2018).%
➢Who%does%the%government%need%to%target?

Previous%research%on%plastic%bag%charges
➢Comparisons%of%behaviour,%motivations,%and%attitudes%before%and%after%(e.g.%
Jakovcevic et%al.,%2014;%Poortinga et%al.’s,%2016)
Limitations:
➢Don’t%know%who%is%still%buying%plastic%bags%
➢Diary%studies%-limited%samples%
➢Questionnaires%used%self-reported%measures%of%behaviour

Method
➢Records%of%actual&shopping%behaviour
➢Bottom%up,%data%driven,%inductive%approach
➢Binary%prediction%(4%or%more%plastic%bags%vs%none)%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
9%months%after%the%levy
➢Balanced%sample%(N=%2,880)
➢Demographics,%attitudes,%and%individual%differences%
➢Research%has%only%previously%focused%on%supermarkets

Figure%1. A%graph%to%show%the%number%of%sales%of%single%use%plastic%bags%per%week%in%the%survey%
sample%(N%=%12,968).%

t(40)=−3.55,%p=.001,%R²=0.24

Figure%2.%A%map%of%the%UK%showing%in%each%region%the%percentage%of%bag’s%sold%by%the%
retailer%per%1000%items%bought%by%the%survey%sample%of%12,968.

!"#$%&'()*+,
-"&,$.'/*01
2"%3/'&4,*.%/"53,"4
(/',0*+46'%4+$.,3#"&
7(".8*.%94
,)$((*.%4
:&";3".+1
<00*038",
=$04'4,*.%/"53,"4
(/',0*+46'%4+$.,3#"&
Method
>
?
Split& training& (80%)&and&
test&data&(20%)
5-fold&cross& validation&
to&optimise&parameters
Test&the&optimised&
models& using& accuracy
Use&the&best&model&
to&extract&variable&
importance
?@47(/*048'0'A B@4C(0*#*,"4#$8"/,A D@4E",04#$8"/,A F@4G'&*'6/"4H#($&0'.+"A

Prediction%results
•Unseen%test%data%accuracy%(N=576)
•Variables%removed%from%logistic%regression%due%to%high%VIF
•Occam’s%razor%=%if%performance%is%the%same,%simplest%best
Logistic&regression Random&forest

Frugal
BAS-Fun%Seeking
‘I%would%rather%have%£55%
now%than%£75%in%3%months’
Openness
BAS-RR%(p=0.054)%
‘Environmental%considerations%
affect%the%products%that%I%purchase’.%
‘I%am%
concerned%
about%climate%
change’%
Shopping%more%frequently%
and%spending%more%
Older%people%
Income:%
£25,000%to%
£49,999
‘Homemakers’
Variable%Importance%Results

➢BAS-FS:%pos%r%and%neg%β%(interactions)
➢BAS-RR:%sig.%Pearson%r
➢The%interactions%and%correlations%don’t%affect%prediction%(RF,%VIFS)%but%do%they%matter%for%interpretation?
Impulsivity

Figure%3.%Number%of%bag%buyers%and%non%bags%buyers%in%each%income%category%for
the%balanced%sample%used%in%the%logistic%regression%analysis%(N%=%2880).

Figure%4. Standardised%beta%coefficients%showing%a%direct%effect%of%age%on%bag%buying,%three%
different%indirect%mediation%pathways%(through%impulsivity,%income%and%frugality),%and%the%
covariances%between%the%mediators.%
SEM

Summary
➢General%climate%change%concerns%didn’t%predict%bag%buying
➢The%levy%has%worked%best%for%low%income%groups
➢Significant%psychological%variables%reflect%a%lack%of%
planning
➢Impulsivity,%income,%and%frugality%mediate%the%effect%of%age
Implications:
➢Jan%2020%10p%levy%might%not%be%the%most%effective%
intervention%for&everyone
➢Need%to%target%high%income%groups,%young%people,%who%
don’t%plan%ahead%– “busy%lifestyle”
➢Convenient%plastic%alternatives
Martin%Creed,%‘Work%No%2814’%
Bishopsgate

Thank-you.
Rosa.Lavelle-Hill@nottingham.ac.uk
https://neodem.wp.horizon.ac.uk
https://www.rosaellenlavelle-hill.site