Compulsive buying gradually increased during the first six months of the Covid-19 outbreak

doi:10.21203/rs.3.rs-640061/v1

Download PDF

Research Article

Compulsive buying gradually increased during the first six months of the Covid-19 outbreak

https://doi.org/10.21203/rs.3.rs-640061/v1

This work is licensed under a CC BY 4.0 License

Journal Publication

published 08 Mar, 2022

Read the published version in Journal of Behavioral Addictions →

Version 1

posted

You are reading this latest preprint version

Background and Aims: The current Covid-19 situation offers a natural experiment to explore the effect of a chronic stressor on compulsive buying tendencies over an extended period of time.

Design: Survey method of sampling every three days a new cohort during the first six months of the Covid-19 pandemic (March-October 2020) in the United States.

Participants: Total (clean) sample of N = 1430 (39.3% female, mean age = 36.4 years).

Measurements: Online and offline compulsive buying separately, distress, SES, income and age were assessed.

Findings: Both online and offline compulsive buying increased during the data collection period (𝜏 = 0.24, 𝜏 = 0.22, respectively, both p < 0.001). High-SES individuals reported the highest tendency for compulsive buying throughout the entire time frame, although the increase in compulsive buying tendencies over time was the highest among the socioeconomically less privileged. Online compulsive buying increased as a result of the CARES Act (first stimulus package) by an effect size of d = 0.33. When entered into a regression model, SES had the strongest effect on compulsive buying after accounting for the effect of distress, income and age (online: ß_PSS = 1.3***, ß_SES = 5.13***, ß_income = 2.6***, ß_age = -0.20*, F(4, 709) = 53.01, R² = 0.23, RSE = 29.5; offline: ß_PSS = 1.45***, ß_SES = 4.86***, ß_income = 2.16***, ß_age = -0.08 p > 0.4; F(4, 695) = 49.54, R² = 0.22, RSE = 30.41). The high-income group reported the strongest correlation between distress and compulsive buying (r = 0.67, p < 0.001, 95% CI: 0.57–0.76).

Conclusions: Compulsive buying tendency gradually increased during the first six months of the Covid-19 pandemic especially as a result of the CARES Act.

Psychology

chronic distress

SARS-CoV-2 pandemic

addiction

shopping addiction

compulsive buying-shopping disorder

socioeconomic status

The current Covid-19 situation provides a unique opportunity for behavioural addiction researchers to gain insight into the effects of chronic stress on shopping in a natural experiment setting. Although publicly available sales reports usually communicate that retail skyrocketed due to the pandemic, it is not certain whether the sharp increase in purchasing in this period is due to internal factors (e.g., a coping mechanism in response to heightened distress) or external factors (e.g., increased availability of money for shopping due to restrictions on travel and socialising).

In one of the first comprehensive phenomenological descriptions of compulsive buying (CB) O’Guinn and Faber (1989) consider buyers compulsive when they are unable to control their impulse to buy, which pervades their lives and sometimes results in severe consequences. Compulsive buyers typically have unmanageable amounts of debt, which create economic and emotional problems for themselves and for their families. Thus compulsive buying is more than excessive spending, it is viewed as a repetitive, uncontrollable behaviour, typically triggered by negative emotional states, where short-term positive rewards reinforce the behaviour resulting in delayed negative consequences (O’Guinn & Faber, 1989). Such rewarding behaviour can be completing a purchase, which was likely to be the reason for the observed elevated compulsive buying behaviour among survivors of a natural disaster, i.e. the Hurricane Katrina (Sneath, Lacey, & Kennett-Hensel, 2009).

Another natural disaster, the Covid-19 pandemic and the subsequent lockdowns had an unprecedented effect on people’s quality of life. The prolonged fear of the disease, the uncertainty, and the unavailability of other, healthier coping strategies e.g., socialising, outdoor activities or group exercise, as well as restrictions on recreational traveling, and serious reduction in social interactions for adolescents all impacted well-being. This has resulted in increased and chronic distress, especially for those with a pre-existing mental health condition (Rogers, Shepherd, Garey, & Zvolensky, 2020). Per definition, excessive behaviours are likely to be a mental and physical way of escaping stressful situations as a way of coping with distress (Király, Tóth, Urbán, Demetrovics, & Maraz, 2017). Supporting this assumption, the frequency of engaging in addictive activities intended for seeking relief from distress was reported to have increased during lockdowns. For example, researchers reported evidence for increase in porn use (Mestre-Bach, Blycker, & Potenza, 2020), online gaming (King, Delfabbro, Billieux, & Potenza, 2020), smartphone and social media use among schoolchildren (Chen et al., 2021), smoking (Patanavanich & Glantz, 2020) overeating (Di Renzo et al., 2020; Robinson et al., 2021) gambling (Håkansson, Fernández-Aranda, Menchón, Potenza, & Jiménez-Murcia, 2020) alcohol (Chodkiewicz, Talarowska, Miniszewska, Nawrocka, & Bilinski, 2020) and substance use (Czeisler et al., 2020; Rogers et al., 2020) during the pandemic, especially among those who already experienced loss of control over their behaviour before the pandemic. There is evidence, that compulsive buying also increased during the Covid-19 pandemic (Çelik & Köse, 2021; Islam et al., 2021; Jaspal, Lopes, & Lopes, 2020), however, authors typically report retrospective data collection or two-point comparison of CB behaviour in the sample. The only study which collected data regarding CB covered an extended period of seven days at the end of February 2020 in China used experience sampling in 150 individuals and concluded that perceived uncertainty influenced (and information overload and anxiety about Covid-19 mediated) daily impulsive buying behaviour (Xiao, Zhang, & Zhang, 2020).

However, not only internal (such as coping style), but also external factors (such as the availability of financial resources) influence people’s decision to buy. Contrary to expectations, however, representative studies typically report that compulsive buyers have a significantly lower income at their disposal than non-compulsive buyers as evidenced by data from Germany (Mueller et al., 2010) and from the United States (Koran, Faber, Aboujaoude, Large, & Serpe, 2006). Other studies, however, found no difference in perceived social class in Spain (Otero-López & Villardefrancos, 2014), Denmark (Reisch, Gwozdz, & Raab, 2011) or another study from Germany (Mueller et al., 2011). Consequently, some instruments assess income and overspending as a predictor of compulsive buying, while others explicitly leave out such items (Maraz, Griffiths, & Demetrovics, 2016). Thus, the relationship between income (social status) and compulsive buying is a yet unanswered and exciting question.

The possibility that SES and/or income may play a role in the experience of distress and compulsive buying has not been explored to the best of our knowledge. However, there is evidence in other disorders that distress primarily affects those with lower income. For example, that income is the strongest (negative) predictor of distress among men, and education among women (Kessler, 1982). In line with this, unskilled manual worker men are more prone to depression than their skilled counterparts, whereas high-status women appeared to be protected from depression (Kosidou et al., 2011). Based on these findings, it is reasonable to suppose that low-SES individuals will experience greater distress and thus are more prone to distress and short-term rewards in the form of compulsive buying as opposed to more wealthy middle- or high-SES individuals.

Publicly available sales reports usually rocketed due to the pandemic despite the fact that personal income remained unchanged, or even decreased as a result of increasing unemployment. For example, net sales at the largest online retail company, Amazon increased by 29% in 2020 Q2 and by 27% in Q3 (from $63.4 in 2019 Q2 to $88.9 in 2020 Q2 and from $70.0 billion in 2019 to $96.1 in 2020). About 8.2% in Q2 was probably due to the pandemic alone (expected $81.6, but realised $88.9 billion¹). During the same period, unemployment increased from around 5% to around 15% between April and June 2020 (then slowly decreased to 6–7%). As an attempt to reduce the economic damage (partially as a result of elevated unemployment) the administrative power issued the first Stimulus Package on the 27th March 2020 arriving on bank accounts about two weeks later, which amounted to 10% of the total U.S. gross domestic product. The Coronavirus Aid, Relief, and Economic Security (CARES) Act allocated money to individuals ($1200 per adult and $500 per child) for the low- and middle income social class among other provisions (H.R.748). Although a certain portion of the fund received by households was likely to have been spent on essential products, such as food or everyday household goods, unplanned products might have also been bought from the extra fund. Thus, the availability of an unexpected financial resource provided a unique opportunity to assess the possibility that this may have boosted compulsive buying patterns in different social classes and income groups.

Neither the impact of elevated distress nor the effect of a large financial aid on compulsive buying has been explored before. The only study, which focused on the effects of a natural disaster (i.e., Hurricane Katrina) found an increase in compulsive buying tendencies as a result of depressive states (Sneath et al., 2009). In line with this finding, a recent study reported, that the loss of safety (fear of Covid-19 and low political trust) significantly increased compulsive buying (Lopes & Jaspal, 2020). Consequently, although the relationship between status (income), indebtedness, distress and compulsive buying has been previously explored, the Covid-19 pandemic situation and the unexpected financial aid from the government provided a unique context for assessing the relationships among the variables.

More specifically we addressed the following research questions:

(1) We aimed to explore whether compulsive buying increased in the first six months of Covid-19 and if this increase is related to SES / income group status.

(2) We wanted to assess the possibility that compulsive buying tendencies further increased once disposable money became available for Americans due to the first Stimulus Package via the CARES Act. We further aimed to explore whether this increase is particularly pronounced across SES/income groups.

(3) We aimed to examine whether the degree of compulsive buying was associated with the experience of distress during Covid after controlling for SES, income and age.

(4) We wanted to investigate whether distress was more strongly associated with compulsive buying during Covid-19 in low SES / income participants than among those with average or high SES / income.

Procedure

A previous article has already been published from the same data collection on addiction-related behaviours (Maraz, Katzinger and Yi, under review). Data were collected every three days from Amazon’s MTurk between 26/03/2020 and 02/10/2020 (inclusive) covering a period of 191 days (from Day 14 till Day 204 of the pandemic). Each time a new cohort of 25 participants were sampled excluding the participants who had previously taken part. Participants were able to take part if they were above the age of 18 (as verified by Amazon) and they were logged in from a US-based IP address. Participants were paid $2.1 per questionnaire, which was an adequate payment based on their feedback.

Measures

The start of the pandemic was defined as the 13th of March 2020 when President Trump issued the Proclamation on Declaring a National Emergency Concerning the Novel Coronavirus Disease (COVID-19) Outbreak, declaring a national state of emergency. The first stimulus package was issued on the 27th March 2020 (Day 15) but arrived at bank accounts on the 11th April 2020 onwards (Day 30). Thus “pre” stimulus package period was defined as the first 15 days of data collection (Day 14 - Day 29), and “post” stimulus package as 15 days after the money arrived to bank accounts (Day 30 - Day 45).

Offline compulsive buying was measured with the Bergen Shopping Addiction Scale (BSAS, Andreassen et al., 2015). The BSAS contains 28 items, four for each of seven addiction criteria (salience, mood modification, conflict, tolerance, withdrawal, relapse, and problems) based on Griffiths’s (2005) “components” model of addiction. After administering the scale to over 23 000 participants, the factor structure, convergent and divergent validity of the scale was found to be acceptable. In the current survey the question to participants was to tell why they enjoyed (or didn’t) brick-and-mortar (offline) shopping in the past 30 days (i.e. “I thought about shopping/buying things all the time.” or “I decided to shop/buy less, but have not been able to do so.”), and answer alternatives were from Completely disagree (= 1) to Completely agree (= 5). The total score ranged between 28 and 140 with higher scores indicate higher probability for the presence of shopping addiction / compulsive buying. The BSAS was only presented to those who made an offline purchase in the past seven days, and items were presented in random order. In the current sample McDonald’s omega (Dunn, Baguley, & Brunsden, 2014) was 0.99, indicating a very good fit to the data.

Online compulsive buying was assessed with the twenty-eight-item Compulsive Online Shopping Scale (COSS) (Manchiraju, Sadachar, & Ridgway, 2017). This scale was adapted from the BSAS (Andreassen et al., 2015) by adding the word “online” to items (i.e. “I thought about shopping/buying things online all the time.” or “I decided to shop/buy less online, but have not been able to do so.” The new instrument was then tested on a US panel sample of 313 adults (mostly females) with the same answer alternatives as for the BSAS (Completely disagree = 1, to completely agree = 5, total score ranging between 28 and 140) and the instrument with seven factors was found to fit adequately to the data. The COSS assessed past-30-day purchase attitudes, and was only presented to those who made an online purchase no more than seven days prior to assessment. Higher scores indicate higher problematic behaviour and the internal consistency of the measure was excellent (McDonald’s omega = 0.98).

Distress was addressed via the 14-item Perceived Stress Scale (Cohen, Kamarck, & Mermelstein, 1983). This time frame for reporting distress was modified from past month to past seven days in the current survey in order to capture the dynamic experience of distress in the Covid-era. Items represent the subjective experience of distress, for example by asking “In the last month, how often have you found that you could not cope with all the things that you had to do?” Pre-determined answers were provided from Never (= 1) to Very often (= 5) with the total possible score ranging between 14 and 70. The instrument contains several reversed items, and higher scores indicate higher levels of distress. McDonald’s omega of factor saturation was 0.84 in the current sample.

Covid-19 related stress was assessed with a single item: “How stressful do you feel about the current situation caused by the corona virus outbreak?” Participants responded on a ten-point scale : 1 (Not at all stressful) to 10 (Very stressful).

In order to increase data quality, the Marlowe-Crowne Social Desirability Scale Short form (M-C Form A, 10 items) was presented to participants (Reynolds, 1982). This instrument has no validated cut-off to exclude socially desirable respondents. We chose the maximum score (10/10) to indicate biased responding.

Socioeconomic status (SES) was assessed with two measures. First, we assessed participants’ subjective sense of SES with the following question: “How wealthy do you think you are compared to others?”. Seven alternatives were offered between “Among the poorest” (= 1) and “Among the wealthiest” (= 7) (Maraz, van den Brink, & Demetrovics, 2015). Income, a proxy of SES, was assessed with the folowing item: What was your income in the past 30 days? Please include your net salary and all other "extra" earnings too. Answering was optional. Seven alternative categories were provided, with options for yearly and the corresponding monthly income, i.e. “$25,000-$34,999 a year ($2084-$2916 a month)”. According to the United States Census Bureau's Annual ASEC survey, the median personal income for 2020 was $43,206.00 (with an average of $62,518.13). Thus income under $50,000 was defined as low, between $50,000 and $75,000 as medium, and above $75,000 as high.

Debts were assessed with two questions. One question was about participants’ credit card balance (“Do you have an unpaid amount on your credit balance that is overdue?”), and the other asked about overdue payment for previous shopping (debt for company, “Do you currently owe money to a company because of shopping? (i.e. you bought something and the bill is overdue)”). Both questions were optional. Seven response alternatives were categorised for both questions: less than $100, $100-$249, $250-$599, $600-$1199, $1200-$1999, $2000-$3999, $4000-$9,999 and over $10,000.

Finally, the past seven-day online spending was assessed with categories such as grocery, clothes, shoes, electronics, hobby, jewellery, etc. with the following question: “Now we would like to know what you bought in the past seven days. Have you bought any of the following items?” when answered positively, then a follow-up question was displayed: “How much did you pay for [name of the category] in the past seven days (in total, if you bought several items)?”. Categories were defined as: less than $50, $50-$99, $100-$249, $250-$499, $500-$1000, and over $1000.

Data cleaning and statistical analysis

Several measures were taken to increase the quality of data. Three attention check items were hidden among regular items (e.g. “Please check “true” here.”). Participants’ age was asked twice, and they scored an error if the answers were different. Those scoring more than one (out of four) attentional errors were excluded from the sample. Finally, those with maximal score (10 out of 10) on the Marlowe-Crowne Social Desirability Scale were also excluded. Out of the initially collected sample of 1885, 1605 were complete. After excluding those with more than one error (145) and those with 10/10 lie score (30), 1430 participants’ data were left for the analyses. The median time of filling out the questionnaire was 11.1 minutes (mean: 26.2).

Given that group sample size across different SES groups were very different, we merged the poor groups (poorest, poorer and poor) and the rich (richest, richer and rich) groups to increase statistical power in the analyses. To test associations we used Pearson correlation between continuous variables, and Kendall’s tau (𝝉) for ordinal (and continuous + ordinal) measures, which can handle ties in the data where members of the pair have the same ordinal value (Khamis, 2008). Regular t-test were used when comparing a continuous variable between groups, or it’s non-parametric equivalent, Welch test when data were non-normally distributed. F-test was used to compare more than two groups. Cohen’s d estimates were used to assess the effect size between two groups, and 95% confidence interval (CI) around the correlation coefficients to assess true differences between point-estimates. Regression model was used to test predictors of online / offline compulsive buying and the significance of their (standardised) effect.

Data were collected using formr (Arslan, Walther, & Tata, 2020), analysed and visualised in R (R Core Team, 2013) and partially in ggplot2 (Wickham, 2016). The data collection procedure was pre-registered prior to data collection under the following link: https://osf.io/m5kw9. All data, material and scripts of analyses are available open-access under https://osf.io/qdhp4/ and https://github.com/anikomaraz/shopping_covid19.

Ethics

The study procedures were carried out in accordance with the Declaration of Helsinki. The Institutional Review Board of the [name of the Institute removed for anonymity] approved the study. All subjects were informed about the study in compliance with the GDPR, were at least 18 years old, and all provided informed consent before entering the study.

Sample description

Overall data were available from 1430 participants following data cleaning (i.e. attention check items). Average age was 36.4 years (SD: 11). Sixty percent of our participants were male, 39.3% female, and 0.5% did not indicate their gender. Most participants had an undergraduate degree (43.8%), whereas 31.3% had lower, and 24.91% had a higher educational level. The majority of participants had a full time job (80%), and were in a relationship (76%). Overall 1111 (78%) participants made an offline purchase, and 1136 (80%) an online purchase in the past seven days prior to responding to the survey items. Only a minority (5.5%, n = 79) did not make any purchase, and about 63% of them (n = 896) made both online and offline purchases in the past seven days.

As it appears in Suppl. 1, most participants spent on health and beauty products, followed by groceries, and to a decreasing extent books, movies, music and games. As the pandemic got longer, participants tended to spend more in each category. Participants reported having overdue credit card balances and debts to a company, although most were debt-free (see Suppl. 2).

Compulsive buying in the first six months of the pandemic (Aim 1)

As seen in Fig. 1 both online and offline compulsive buying tended to increase as the pandemic grew, especially after the introduction of the first stimulus package. The association between compulsive buying and time are significant both in the online (𝜏 = 0.24, p < 0.001) and in the offline context (𝜏 = 0.22, p < 0.001). We assessed the possibility that this pattern may have been different across SES and income groups. Within the online context, the poor group reported significant increase within their compulsive buying tendency (𝜏 = 0.24, p < 0.001), the average and high income groups’ compulsive buying tendencies remained non-significant (average: 𝜏 = 0.07, p = 0.14; rich: 𝜏 = -0.06, p = 0.18). The pattern was similar within the offline context, although the association was weaker but significant in the average group as well (poor: 𝜏 = 0.22, p < 0.001; average: 𝜏 = 0.11, p = 0.02; rich: 𝜏 = -0.01, p = 0.74). As for income-derived groups, individuals with both low (𝜏 = 0.24, p < 0.001), and middle (𝜏 = 0.30 p < 0.001) income increased their online compulsive buying, whereas the increase was marginally significant in the high-income group (𝜏 = 0.12, p = 0.05). Similarly, offline compulsive buying tendency significantly increased over time for both low (𝜏 = 0.23, p < 0.001) and middle (𝜏 = 0.28, p < 0.001) income groups, while the pattern was not significant for the high income group (𝜏 = 0.07, p = 0.24). This pattern was essentially the same as the findings derived from subjective SES-derived groups as visualized in Suppl. 3.

Note

smoothed lines are presented to reduce noise in visualising. Case numbers are taken from https://www.ecdc.europa.eu/en/publications-data/download-todays-data-geographic-distribution-covid-19-cases-worldwide

The effect of the stimulus package (Aim 2)

To calculate the effect of the stimulus package, the first 15 days of data collection (before the financial aid arrived at bank accounts) was compared to the next 15 days when the money was available for spending. As seen in Table 1, compulsive buying significantly increased in the online (but not in the offline) context following the stimulus package with a medium effect size (Cohen’s d = 0.33). However, significance was undetectable within each SES group (see Table 1). The offline shopping context remained unaffected.

Table 1

Compulsive buying before and after the stimulus package
	N		Compulsive Buying mean (SD)		Pre-post Difference
Social status	Stimulus Package		Stimulus Package		Welch test t (df)	Cohen’s d
Social status	Pre	Post	Pre	Post	Welch test t (df)	Cohen’s d
Online compulsive buying (COSS)
Poor	32	27	44.09 (14.6)	50.67 (23.3)	1.27 (42.2) n.s.
Average	33	19	51.88 (18.0)	46.79 (16.0)	-1.05 (41.4) n.s.
Rich	3	6	108.33 (42.9)	108.33 (24.3)	0 (2.67) n.s.
All	68	52	50.09 (22.6)	58.33 (27.6)	2.33 (193.0)*	0.33 (95% CI: 0.05, 0.60)
Offline compulsive buying (Bergen Shopping Scale)
Poor	28	20	40.6 (14.7)	44.3 (21.8)	0.65 (30.94) n.s.
Average	37	16	48.6 (16.2)	40.4 (14.4)	-1.83 (32.02) n.s.
Rich	6	6	96.5 (31.9)	96.2 (22.6)	-0.02 ( 9.01) n.s.
All	71	42	48.5 (23.4)	54.4 (27.5)	1.56 (156.2) n.s.

Note: *<0.05. “Pre” stimulus package was defined as the first 15 days of data collection (Day 14 - Day 29), and “post” stimulus package as 15 days after the money arrived to bank accounts (Day 30 - Day 45). Categories within “poor” (poor, poorer and poorest) and “rich” (rich, richer, richest) were merged to increase statistical power.

Compulsive buying in response to distress during Covid-19 (Aim 3)

In Fig. 2, we plotted online as well as offline compulsive buying tendencies per group derived from SES and income measures. Groups were visualised separately, but merged to increase statistical power (poorest, poorer and poor groups were combined to “poor”, and richest, richer and rich groups to “rich”). Overall, both online and offline compulsive buying tendencies were higher for high-SES participants. In terms of mean differences in online compulsive buying (COSS), the rich groups reported the highest values (mean: 106.2, SD: 18.3), followed by the poor groups (67.5, SD: 31.4) and the average-SES participants (55.8, SD: 25.9) (F(2, 711) = 222.3, p < 0.001). Results were similar in the offline context (mean_rich=105.1, SD = 18.6; mean_poor=67.11, SD = 32.8; mean_average=55.9, SD = 28.5), and all group differences were significant at the level of p < 0.05 (F(2,697) = 188).

The zero-order correlation between distress and online compulsive buying was r = 0.45 (p < 0.001, 95% CI: 0.41–0.50), and the one between distress and offline compulsive buying was r = 0.47 (p < 0.001, 95% CI: 0.43–0.52). Furthermore, rich-SES participants experienced the highest distress (41, SD = 4.1), followed by the poor-SES group (38.2, SD = 9.9), and the average group (36.4, SD = 10.2), and all group differences were significant (Bonferroni-corrected group differences: all p < 0.03; F-statistics: F(2, 907) = 20.8, p < 0.001, R² = 0.04). Correlation between SES and income was 𝜏 = -0.01, p = 0.64; whereas between age and SES was 𝜏 = -0.05, p = 0.03 and between age and income was 𝜏 = 0.05, p = 0.01 suggesting that SES and income measure different constructs, and their association with age is only weak (although significant).

Note

violin plots reflect the distribution of the data, whereas lower and upper hinges of the boxplots correspond to the first and third quartiles (the 25th and 75th percentiles) of the data. Data beyond the end of the whiskers are outliers. Participants had the option to refuse answering any of the items visualised on the plot. None of the participants indicated belonging to the “Richest” category, thus this was removed from the figure.

In order to assess the relationship between distress and compulsive buying above and beyond the association of income and SES, we separately regressed BSAS and COSS on distress, income, SES and age. We found that although online compulsive buying was significantly predicted by income, SES and age, distress was also a significant positive predictor (ß_PSS = 1.3***, ß_SES = 5.13***, ß_income = 2.6***, ß_age = -0.20*, F(4, 709) = 53.01, R² = 0.23, RSE = 29.5). Similarly, offline compulsive buying was positively predicted when the effects of income, SES and age were taken into account (ß_PSS = 1.45***, ß_SES = 4.86***, ß_income = 2.16***, ß_age = -0.08 p > 0.4; F(4, 695) = 49.54, R² = 0.22, RSE = 30.41). Thus the distress has a weak, though independent effect from SES, income and age on compulsive buying. Furthermore, the effect of SES appears to be stronger than the effect of income.

The relation between distress and compulsive buying: differences among SES / income groups (Aim 4)

Figure 3 displays the correlation between distress and compulsive buying in the total sample and by each SES group. As a visual trend, participants with medium distress (about 35 to 50 on the PSS) appear to report the highest compulsive buying tendencies in each SES group.

The correlation between distress and online CB was significantly positive in the three income-derived groups (N_low = 727, r = 0.38, p < 0.001, 95% CI: 0.31–0.44; N_medium = 200, r = 0.62, p < 0.001, 95% CI: 0.52–0.70; N_high = 133, r = 0.67, p < 0.001, 95% CI: 0.57–0.76). Similarly, the Pearson correlation between distress and online compulsive buying was significantly positive in all the SES-derived groups (N_poor = 297, r = 0.31, p < 0.001, 95% CI: 0.21–0.41; N_average= 184, r = 0.4, p < 0.001, 95% CI: 0.27–0.51; N_rich= 233, r = 0.31, p < 0.001, 95% CI: 0.19–0.42). Given their non-overlaping CIs, the low-income group reported significantly lower correlation between distress and compulsive buying compared to the high-income group.

In terms of offline shopping all within-group correlations were significant in the income groups (N_low = 723, r = 0.42, p < 0.001, 95% CI: 0.36–0.48; N_medium = 181, r = 0.62, p < 0.001, 95% CI: 0.52–0.70; N_high = 130, r = 0.61, p < 0.001, 95% CI: 0.49–0.71). Similarly, coefficients were significant within the SES-groups (N_poor = 285, r = 0.36, p < 0.001, 95% CI: 0.25–0.46; N_average= 191, r = 0.38, p < 0.001, 95% CI: 0.25–0.50; N_rich= 233, r = 0.30, p < 0.001, 95% CI: 0.17–0.41). Given the 95% confidence intervals of the point-estimates, the difference in low vs. medium, and low vs. high income were meaningful, but the between-group differences according to SES-derived groups were nonexistent.

The dashed line is the linear regression line between distress and shopping. Online shopping was measured with the Compulsive Online Shopping Scale. Offline shopping was assessed via the Bergen Shopping Scale. Social-economical status was assessed as “How wealthy do you think you are compared to others?”

The current study aimed to explore compulsive buying tendencies during the first six months of the Covid-19 pandemic in a sample of 1430 US participants. Overall, compulsive buying patterns (online and offline) increased during this period, especially among less wealthy individuals (Aim 1). As a result of the CARES Act (first stimulus package), compulsive buying activity increased in our sample, but only in the online context (Aim 2). Although socioeconomic status was the strongest predictor of compulsive buying in our model both in the online as well as in the offline context, the intensity of distress experienced positively influenced compulsive buying above and beyond the effect of SES, income and age (Aim 3). The correlation between distress and compulsive buying was higher among individuals with high income compared to the below-average income group (Aim 4). Thus, the tendency to engage in compulsive buying as an attempt to cope with distress during the period of chronic stress was more pronounced among people with high social status than among those with low or middle socioeconomic status, which cannot be attributed to distress alone.

In terms of our first aim, we found a gradually increasing trend of compulsive buying patterns during an extended period of the Covid-19 pandemic from our sample. Although there was an indication that compulsive buying increased after the outbreak of pandemic compared to before (Lopes & Jaspal, 2020; Xiao et al., 2020), the gradual increase of compulsive buying over an extended period during the course of pandemic has never been reported before. On the other hand, we also found that the increase in compulsive buying pattern was significant only in the low-SES and low-income groups both in the online and in the offline context. The trend was not significant for the high-SES and high-income groups (and was significant but weak for the average-SES). The overall increase in compulsive buying tendency in our sample is not surprising. Given the strong association, elevated compulsive buying tendency is likely to be the result of increased and chronic stress, and the restricted availability of other, healthier coping strategies (social interactions, most forms of physical exercise, recreational traveling, etc.), which possibly resulted in escaping into compulsive buying (Király et al., 2017). There was previous evidence, that people seek substitutes when a ban is introduced on recreational but potentially harmful activities such as alcohol and cigarette consumption in South Africa (Sinclair et al., 2021) or abuse alcohol when less adaptive coping strategies were favoured (Chodkiewicz et al., 2020) especially in those with high worry about Covid-19 situations (Czeisler et al., 2020; Rogers et al., 2020). This is probably due to the fact that addiction, in this case compulsive buying, is an attempt to adapt to interpersonal trauma, as a result of “compromised abilities to form healthy attachments and decreased capacity for self-regulation” (p. 352. Padykula & Conklin, 2010). However, regarding the findings according to SES / income group the increase in compulsive buying was only significant among the poorest and those with average or below-average income, which may have come as a surprise. It is possible, that high-status participants could afford to shop throughout the study period as a way of lifting their spirits, whereas their less wealthy counterparts restricted their compulsive buying habits, and then increased their buying as a result of (ego-control) fatigue as the pandemic grew. This is an especially likely explanation given according to a recent meta-analysis, socioeconomic inequalities were clearly present in terms of unemployment, infection risks as well as the severity of the course of the disease, with the less privileged individuals being hit harder (Wachtler et al., 2020).

Our second aim was to assess the effect of the CARES Act (the first stimulus package) on compulsive buying. This Act provided financial help for the middle and lower social classes with the aim of boosting the economy. Economists generally found that people spent about the third of the funds within 10 days of receiving it, and those with less initial money in their bank accounts tended to spend more of the financial aid right away. This effect partially appeared in our sample. Total compulsive buying tendency to buy online (but not offline, because brick-and-mortar shops were closed in most states during this period) significantly increased from 14 days before to 14 days after the financial aid was available on bank accounts. However, the increase in compulsive buying within each SES / income group was not significant. More data are needed to draw firm conclusions on the effect of the stimulus package according to SES and income.

Our third aim concerned the relationship between distress and compulsive buying. We found that distress was a significant and strong predictor of compulsive buying, although this effect was rather small when SES, income and age were added as predictors. Thus it is rather the standard of living, i.e. the availability of funds that triggers compulsive buying in the presence of distress, rather than distress per se. Therefore it might be possible that in the presence of distress, people choose addictive behaviours depending on the availability of other instruments, such as money. It is likely that individuals who reported problematic addictive behaviours (i.e. heightened tendency for overbuying as a result of higher standard of living) before the pandemic would be especially likely to lose control over their behaviour during the pandemic as a result of elevated distress (and develop compulsive buying). There is already evidence for this mechanism (Czeisler et al., 2020; Rogers et al., 2020).

To further clarify the role of SES and income in the relationship between distress and compulsive buying, we calculated the association within each SES / income group as our Aim 4. Although the correlations were equally strong between SES groups, high- and average-income participants reported higher association between distress and compulsive buying than low-income participants. This indicates that participants with average or high income are more likely to loose control in the presence of distress than those with lower income, which fits the pattern of the regression model (Aim 3), in which we found that much of the effect of distress is conveyed by the standard of living. However, this finding is contrary to most representative studies reporting that low-SES individuals have higher compulsive buying tendencies than high-SES people, which is usually attributed to the fact, that low-SES individuals use lower-quality coping and higher resting Cortisol levels than high-SES individuals (Schmeelk-Cone, Zimmerman, & Abelson, 2003). In any case, financial status cannot be ignored when measuring compulsive buying, i.e. by omitting items of a questionnaire that assess the financial implications of buying activity (Ridgway, Kukar-Kinney, & Monroe, 2008), since SES / income were more important predictors of compulsive buying than distress during our time frame.

Our findings appear controversial regarding the role of socioeconomic status in compulsive buying. One one hand, high SES participants reported the highest compulsive buying values throughout the time frame, but the increase in compulsive buying pattern was only significant among the poor-SES participants. On the other hand, the association between distress and compulsive buying was the strongest among the high-status participants. One explanation for these findings could be that a substantial proportion of high-status individuals recruited in our sample may already have had high levels of compulsive buying at the time of the pandemic outbreak, achieved a plateau in their compulsive buying activity, and as a result of high distress and high resources continued to buy. In addition, less wealthy participants increased their spending as a result of withholding spending early in the pandemic, and then gradually losing control over shopping behaviour as a result of the chronic stress. This hypothesis gains further support by examining the spending patterns (Suppl. 1). As the pandemic grew, the buying of clothes, shoes, electronics and gifts showed a steep increase, but spending on most other goods was stagnant until Day 80, when distress and case numbers also started to increase. Economists also observed a shift in patterns. After the normalisation of panic buying, spending in home apparel, beauty, and electronics increased, as extra money normally spent on restaurants and travel was suddenly available.

In many regions in the United States, part of the lockdown was the periodic restriction on brick-and-mortar shopping opportunities. This could have resulted in binge-shopping before and after the measure came to power. However, our data suggest that online and offline shopping were very similar in terms of trends and financial and/or psychological predictors. The only difference was that the stimulus package boosted shopping only in the online but not in the offline context, which was probably due to the fact that in many states brick-and-mortar shops were closed during this period.

This study has several limitations. First of all, we sampled a different cohort of people at each time point instead of using a follow-up design. We decided on this approach in order to avoid fatigue, dropout and the impact of major individual effects (i.e. becoming unemployed) in the sample. However, one disadvantage of our approach was that we were not able to conduct a longitudinal assessment of prolonged effects during the Covid-19 situation. Second, since the sample consisted of MTurkers, some caution may be necessary before generalizing our findings to other people. However, it has been found that after careful attention check and data cleaning measures, MTurk samples are more informative about the general population than most ad-hoc gathered (especially panel) samples (Kennedy et al., 2020; Smith, Roster, Golden, & Albaum, 2016). Thirdly, the instruments we used to assess online and offline compulsive buying (COSS and BSAS) are almost identical (M. D. Griffiths et al., 2016), which limits their validity. The similarity in wording could also be the reason why online and offline buying models appeared very similar to each other in our data. Fourth limitation is the differing time-frames of assessment: distress was measured in the past seven days, whereas compulsive buying was assessed for the past 30 days prior to data collection. We opted for this approach to reflect the effect of the quickly-changing pandemic measures (distress), whereas participants’ spending habits, attached emotions and attitudes are best captured by a longer, thirty-day period to reflect the effect of income arriving on a monthly basis. However, since we tested the effects of distress on subsequent compulsive buying, a longer time frame for assessing the latter than the former may have had an impact on our calculations. Finally, despite the significant correlations, it is possible, that third factors might mediate the effect of distress, such as health and economic fears (Eger, Komárková, Egerová, & Mičík, 2021), the de-facto cohabitation status (i.e. living alone or with family) the quality of social network (Lopes & Jaspal, 2020), pre-pandemic levels of use or abuse (Rogers et al., 2020), or daily fluctuations in affect (Buecker et al., 2020).

In conclusion, we found that compulsive buying increased during the first six months of the Covid-19 pandemic in the USA especially as a result of the first stimulus package. Although compulsive buying is affected by fluctuations in distress, socioeconomic status (SES) has a markedly large influence on the heightened compulsive buying tendency beyond the effect of distress. High-income people are more prone to compulsive buying than low-income individuals, thus financial status plays an important role in the maintenance of compulsive buying in addition to psychological factors such as distress. Further caution is necessary to track the changes in consumer behaviour, and how it affects well-being of those, mostly at risk of suffering the consequences of losing control over buying habits during these challenging times.

Funding sources: Institute of Psychology at the Humboldt University of Berlin

Authors’ contribution: AM designed the concept and delivery of the study, wrote the statistical analysis and interpreted the data. YS provided consultation on the design and interpretation of findings. Both authors had full access to all data in the study and they all take responsibility for the integrity of the data and the accuracy of the data analysis.

Conflict of interest: The authors declare no conflict of interest.

Acknowledgement: I would like to thank my family for their support during the study in these challenging times.

Andreassen, C. S., Griffiths, M. D., Pallesen, S., Bilder, R. M., Torsheim, T., & Aboujaoude, E. (2015). The Bergen Shopping Addiction Scale: Reliability and validity of a brief screening test. Frontiers in Psychology, 6. https://doi.org/10.3389/fpsyg.2015.01374

Arslan, R. C., Walther, M. P., & Tata, C. S. (2020). formr: A study framework allowing for automated feedback generation and complex longitudinal experience-sampling studies using R. Behavior Research Methods, 52(1), 376–387. https://doi.org/10.3758/s13428-019-01236-y

Buecker, S., Horstmann, K. T., Krasko, J., Kritzler, S., Terwiel, S., Kaiser, T., & Luhmann, M. (2020). Changes in daily loneliness for German residents during the first four weeks of the COVID-19 pandemic. Social Science & Medicine, 265, 113541. https://doi.org/10.1016/j.socscimed.2020.113541

Çelik, S., & Köse, G. G. (2021). Mediating effect of intolerance of uncertainty in the relationship between coping styles with stress during pandemic (COVID-19) process and compulsive buying behavior. Progress in Neuro-Psychopharmacology and Biological Psychiatry, 110, 110321. https://doi.org/10.1016/j.pnpbp.2021.110321

Chen, I.-H., Chen, C.-Y., Pakpour, A. H., Griffiths, M. D., Lin, C.-Y., Li, X.-D., & Tsang, H. W. H. (2021). Problematic internet-related behaviors mediate the associations between levels of internet engagement and distress among schoolchildren during COVID-19 lockdown: A longitudinal structural equation modeling study. Journal of Behavioral Addictions. https://doi.org/10.1556/2006.2021.00006

Chodkiewicz, J., Talarowska, M., Miniszewska, J., Nawrocka, N., & Bilinski, P. (2020). Alcohol Consumption Reported during the COVID-19 Pandemic: The Initial Stage. International Journal of Environmental Research and Public Health, 17(13), 4677. https://doi.org/10.3390/ijerph17134677

Cohen, S., Kamarck, T., & Mermelstein, R. (1983). A Global Measure of Perceived Stress. Journal of Health and Social Behavior, 24(4), 385. https://doi.org/10.2307/2136404

Czeisler, M. É., Lane, R. I., Petrosky, E., Wiley, J. F., Christensen, A., Njai, R., … Rajaratnam, S. M. W. (2020). Mental Health, Substance Use, and Suicidal Ideation During the COVID-19 Pandemic—United States, June 24–30, 2020. MMWR. Morbidity and Mortality Weekly Report, 69(32), 1049–1057. https://doi.org/10.15585/mmwr.mm6932a1

Di Renzo, L., Gualtieri, P., Pivari, F., Soldati, L., Attinà, A., Cinelli, G., … De Lorenzo, A. (2020). Eating habits and lifestyle changes during COVID-19 lockdown: An Italian survey. Journal of Translational Medicine, 18(1), 229. https://doi.org/10.1186/s12967-020-02399-5

Dunn, T. J., Baguley, T., & Brunsden, V. (2014). From alpha to omega: A practical solution to the pervasive problem of internal consistency estimation. British Journal of Psychology, 105(3), 399–412. https://doi.org/10.1111/bjop.12046

Eger, L., Komárková, L., Egerová, D., & Mičík, M. (2021). The effect of COVID-19 on consumer shopping behaviour: Generational cohort perspective. Journal of Retailing and Consumer Services, 61, 102542. https://doi.org/10.1016/j.jretconser.2021.102542

Griffiths, M. (2005). A ‘components’ model of addiction within a biopsychosocial framework. Journal of Substance Use, 10(4), 191–197. https://doi.org/10.1080/14659890500114359

Griffiths, M. D., Andreassen, C. S., Pallesen, S., Bilder, R. M., Torsheim, T., & Aboujaoude, E. (2016). When Is a New Scale not a New Scale? The Case of the Bergen Shopping Addiction Scale and the Compulsive Online Shopping Scale. International Journal of Mental Health and Addiction, 14(6), 1107–1110. https://doi.org/10.1007/s11469-016-9711-1

Håkansson, A., Fernández-Aranda, F., Menchón, J. M., Potenza, M. N., & Jiménez-Murcia, S. (2020). Gambling During the COVID-19 Crisis – A Cause for Concern. Journal of Addiction Medicine, 14(4), e10–e12. https://doi.org/10.1097/ADM.0000000000000690

Islam, T., Pitafi, A. H., Arya, V., Wang, Y., Akhtar, N., Mubarik, S., & Xiaobei, L. (2021). Panic buying in the COVID-19 pandemic: A multi-country examination. Journal of Retailing and Consumer Services, 59, 102357. https://doi.org/10.1016/j.jretconser.2020.102357

Jaspal, R., Lopes, B., & Lopes, P. (2020). Predicting social distancing and compulsive buying behaviours in response to COVID-19 in a United Kingdom sample. Cogent Psychology, 7(1), 1800924. https://doi.org/10.1080/23311908.2020.1800924

Kennedy, R., Clifford, S., Burleigh, T., Waggoner, P. D., Jewell, R., & Winter, N. J. G. (2020). The shape of and solutions to the MTurk quality crisis. Political Science Research and Methods, 8(4), 614–629. https://doi.org/10.1017/psrm.2020.6

Kessler, R. C. (1982). A Disaggregation of the Relationship between Socioeconomic Status and Psychological Distress. American Sociological Review, 47(6), 752. https://doi.org/10.2307/2095211

Khamis, H. (2008). Measures of Association: How to Choose? Journal of Diagnostic Medical Sonography, 24(3), 155–162. https://doi.org/10.1177/8756479308317006

King, D. L., Delfabbro, P. H., Billieux, J., & Potenza, M. N. (2020). Problematic online gaming and the COVID-19 pandemic. Journal of Behavioral Addictions, 9(2), 184–186. https://doi.org/10.1556/2006.2020.00016

Király, O., Tóth, D., Urbán, R., Demetrovics, Z., & Maraz, A. (2017). Intense video gaming is not essentially problematic. Psychology of Addictive Behaviors, 31(7), 807–817. https://doi.org/10.1037/adb0000316

Koran, L. M., Faber, R. J., Aboujaoude, E., Large, M. D., & Serpe, R. T. (2006). Estimated prevalence of compulsive buying behavior in the United States. American Journal of Psychiatry, 163(10), 1806–1812.

Kosidou, K., Dalman, C., Lundberg, M., Hallqvist, J., Isacsson, G., & Magnusson, C. (2011). Socioeconomic status and risk of psychological distress and depression in the Stockholm Public Health Cohort: A population-based study. Journal of Affective Disorders, 134(1–3), 160–167. https://doi.org/10.1016/j.jad.2011.05.024

Lopes, B. C. da S., & Jaspal, R. (2020). Understanding the mental health burden of COVID-19 in the United Kingdom. Psychological Trauma: Theory, Research, Practice, and Policy, 12(5), 465–467. https://doi.org/10.1037/tra0000632

Manchiraju, S., Sadachar, A., & Ridgway, J. L. (2017). The Compulsive Online Shopping Scale (COSS): Development and Validation Using Panel Data. International Journal of Mental Health and Addiction, 15(1), 209–223. https://doi.org/10.1007/s11469-016-9662-6

Maraz, A., Griffiths, M. D., & Demetrovics, Z. (2016). The prevalence of compulsive buying: A meta-analysis: Prevalence of compulsive buying. Addiction, 111(3), 408–419. https://doi.org/10.1111/add.13223

Maraz, A., van den Brink, W., & Demetrovics, Z. (2015). Prevalence and construct validity of compulsive buying disorder in shopping mall visitors. Psychiatry Research, 228(3), 918–924. https://doi.org/10.1016/j.psychres.2015.04.012

Mestre-Bach, G., Blycker, G. R., & Potenza, M. N. (2020). Pornography use in the setting of the COVID-19 pandemic. Journal of Behavioral Addictions, 9(2), 181–183. https://doi.org/10.1556/2006.2020.00015

Mueller, A., Mitchell, J. E., Crosby, R. D., Gefeller, O., Faber, R. J., Martin, A., … de Zwaan, M. (2010). Estimated prevalence of compulsive buying in Germany and its association with sociodemographic characteristics and depressive symptoms. Psychiatry Research, 180(2–3), 137–142. https://doi.org/10.1016/j.psychres.2009.12.001

Mueller, A., Mitchell, J. E., Peterson, L. A., Faber, R. J., Steffen, K. J., Crosby, R. D., & Claes, L. (2011). Depression, materialism, and excessive Internet use in relation to compulsive buying. Comprehensive Psychiatry, 52(4), 420–424. https://doi.org/10.1016/j.comppsych.2010.09.001

O’Guinn, T., C., & Faber, R. J. (1989). Compulsive buying: A phenomenological exploration. Journal of Consumer Research, 16(2), 147–157.

Otero-López, J. M., & Villardefrancos, E. (2014). Prevalence, sociodemographic factors, psychological distress, and coping strategies related to compulsive buying: A cross sectional study in Galicia, Spain. BMC Psychiatry, 14(1), 101. https://doi.org/10.1186/1471-244X-14-101

Padykula, N. L., & Conklin, P. (2010). The Self Regulation Model of Attachment Trauma and Addiction. Clinical Social Work Journal, 38(4), 351–360. https://doi.org/10.1007/s10615-009-0204-6

Patanavanich, R., & Glantz, S. A. (2020). Smoking Is Associated With COVID-19 Progression: A Meta-analysis. Nicotine & Tobacco Research, 22(9), 1653–1656. https://doi.org/10.1093/ntr/ntaa082

R Core Team. (2013). R: A language and environment for statistical computing.

Reisch, L. A., Gwozdz, W., & Raab, G. (2011). Compulsive buying in Denmark: The first study on Danish consumers’ tendency to compulsive buying. Retrieved from https://openarchive.cbs.dk/handle/10398/8867

Reynolds, W. M. (1982). Development of reliable and valid short forms of the marlowe‐crowne social desirability scale. Journal of Clinical Psychology, 38, 119–125.

Ridgway, N. M., Kukar-Kinney, M., & Monroe, K. B. (2008). An Expanded Conceptualization and a New Measure of Compulsive Buying. Journal of Consumer Research, 35(4), 622–639. https://doi.org/10.1086/591108

Robinson, E., Boyland, E., Chisholm, A., Harrold, J., Maloney, N. G., Marty, L., … Hardman, C. A. (2021). Obesity, eating behavior and physical activity during COVID-19 lockdown: A study of UK adults. Appetite, 156, 104853. https://doi.org/10.1016/j.appet.2020.104853

Rogers, A. H., Shepherd, J. M., Garey, L., & Zvolensky, M. J. (2020). Psychological factors associated with substance use initiation during the COVID-19 pandemic. Psychiatry Research, 293, 113407. https://doi.org/10.1016/j.psychres.2020.113407

Schmeelk-Cone, K. H., Zimmerman, M. A., & Abelson, J. L. (2003). The Buffering Effects of Active Coping on the Relationship Between SES and Cortisol Among African American Young Adults. Behavioral Medicine, 29(2), 85–94. https://doi.org/10.1080/08964280309596061

Sinclair, D. L., Vanderplasschen, W., Savahl, S., Florence, M., Best, D., & Sussman, S. (2021). Substitute addictions in the context of the COVID-19 pandemic. Journal of Behavioral Addictions, 9(4), 1098–1102. https://doi.org/10.1556/2006.2020.00091

Smith, S. M., Roster, C. A., Golden, L. L., & Albaum, G. S. (2016). A multi-group analysis of online survey respondent data quality: Comparing a regular USA consumer panel to MTurk samples. Journal of Business Research, 69(8), 3139–3148. https://doi.org/10.1016/j.jbusres.2015.12.002

Sneath, J. Z., Lacey, R., & Kennett-Hensel, P. A. (2009). Coping with a natural disaster: Losses, emotions, and impulsive and compulsive buying. Marketing Letters, 20(1), 45–60. https://doi.org/10.1007/s11002-008-9049-y

Wachtler, B., Michalski, N., Nowossadeck, E., Diercke, M., Wahrendorf, M., Santos-Hövener, C., … Hoebel, J. (2020). Socioeconomic inequalities and COVID-19 – A review of the current international literature. https://doi.org/10.25646/7059

Wickham, H. (2016). ggplot2: Elegant graphics for data analysis. Springer. Retrieved from http://ggplot2.org

Xiao, H., Zhang, Z., & Zhang, L. (2020). A diary study of impulsive buying during the COVID-19 pandemic. Current Psychology. https://doi.org/10.1007/s12144-020-01220-2

Download PDF

Journal Publication

published 08 Mar, 2022

Read the published version in Journal of Behavioral Addictions →

Version 1

posted

You are reading this latest preprint version

Compulsive buying gradually increased during the first six months of the Covid-19 outbreak

Status:

Journal Publication

Version 1

Abstract

Figures

Introduction

Methods

Procedure

Measures

Data cleaning and statistical analysis

Ethics

Results

Sample description

Compulsive buying in the first six months of the pandemic (Aim 1)

The effect of the stimulus package (Aim 2)

Compulsive buying in response to distress during Covid-19 (Aim 3)

The relation between distress and compulsive buying: differences among SES / income groups (Aim 4)

Discussion

Declarations

References

Supplementary Files

Status:

Journal Publication

Version 1