key: cord-0817410-kj1rhh22
authors: Chenarides, Lauren; Grebitus, Carola; Lusk, Jayson L.; Printezis, Iryna
title: Food consumption behavior during the COVID‐19 pandemic
date: 2020-12-15
journal: Agribusiness (N Y N Y)
DOI: 10.1002/agr.21679
sha: 33666341de1bc58473b64423a0158ca76686f789
doc_id: 817410
cord_uid: kj1rhh22

We conducted an online consumer survey in May 2020 in two major metropolitan areas in the United States to investigate food shopping behaviors and consumption during the pandemic lockdown caused by COVID‐19. The results of this study parallel many of the headlines in the popular press at the time. We found that about three‐quarters of respondents were simply buying the food they could get due to out of stock situations and about half the participants bought more food than usual. As a result of foodservice closures, consumers indicated purchasing more groceries than normal. Consumers attempted to avoid shopping in stores, relying heavily on grocery delivery and pick‐up services during the beginning of the pandemic when no clear rules were in place. Results show a 255% increase in the number of households that use grocery pickup as a shopping method and a 158% increase in households that utilize grocery delivery services. The spike in pickup and delivery program participation can be explained by consumers fearing COVID‐19 and feeling unsafe. Food consumption patterns for major food groups seemed to stay the same for the majority of participants, but a large share indicated that they had been snacking more since the beginning of the pandemic which was offset by a sharp decline in fast food consumption.

To answer these questions, we began by collecting data in Detroit, MI, and Phoenix, AZ, using an online survey. We focused on these two metropolitan areas for the following reasons. First, both areas are similar in that they are two of the most populated metropolitan statistical areas, according to the U.S. Census, yet they are located in very different parts of the country. Second, we chose to interview urban shoppers, as their shopping patterns as well as internet access is crucial for delivery and pick-up order placement yet very different from those in rural areas (Devadas & Lys, 2011 , Hassan, 2015 , Lennon et al., 2009 , Mahmood et al., 2004 , Patel et al., 2015 , Sehrawet & Kundu, 2007 . With regard to  (AZDHS, 2020 , Michigan, 2020 . The survey was programmed by the researchers in the platform Qualtrics. Data were collected by the consumer panel company Dynata between May 13, 2020, and May 30, 2020. The study was approved by the IRB of Arizona State University. Data were analyzed using Stata version 14.

In the survey, we asked a series of questions that sought to investigate how individuals' food shopping behaviors and consumption patterns changed during COVID-19. Shopping behaviors included whether participants bought what they could due to empty shelves, whether they stockpiled food, and how often they went to the food store.

We also investigated participation in grocery delivery and pick-up services before and since COVID-19. Our main focus, however, was on changes in dietary patterns and food consumption during COVID-19. To measure dietary changes, we asked "How much has your diet changed since COVID-19 started?" From a list of responses participants could select any of the following answer categories (multiple options allowed): eat more, eat less, eat about the same, eat less healthy, eat more healthy. To measure changes in food consumption, we asked "How much more or less have you consumed these foods since COVID-19 started?" for 10 major food groups: fresh produce, dairy, meat, grains, snacks, fast food, frozen food, canned food, prepped food, and bottled water. The answer categories were based on a five-point Likert scale: A lot more (5), A bit more (4), About the same (3), A little less (2), A lot less (1) and Do not consume. The five responses were recoded, such that "a lot more" and "a bit more" received a value of 3, "about the same" received a value of 2, and "a little less" and "a lot less" received a value of 1. Those who answered "Do not consume" were treated as missing.

To answer the first three research questions, we rely on univariate statistical analysis, and the results are reported in Sections 3.1-3.4. To estimate the relationship between individual characteristics and the likelihood that a respondent made changes in their consumption during COVID-19 (RQ4), we apply an ordered probit model to each food category.

As described above, we ask how much respondents made changes to their diet across 10 food categories (fresh produce, dairy, meat, grains, snacks, fast food, frozen food, canned food, prepped food, and bottled water). We use the ordered probit model because it not only takes into account that the responses to our survey instrument are categorical and implicitly rank-ordered, but that the alternatives are correlated, that is, an alternative ("eat less") is more similar to one ("eat the same") than the other ("eat more"). As is the case for other probit models, the ordered probit model assumes a linear functional form for each participant's indirect utility function. The unobserved preference obtained by consumer i to maintain the respective level of consumption during COVID-19 is:

where x i is the vector of independent variables including, among others, socio-demographics, such as age, gender, and education. β is a vector of coefficients associated with x i , and an error term, ε i , which is assumed to follow a standard normal distribution. y i is the observed ordinal variable, denoted as the consumption frequency following this equation:

where j = 0, …, M is the number of possible y outcomes where the "highest category is M. u j 's are unknown cut-off values. In this study, M is equal to three. By assuming the error term ε i to follow a standard normal distribution, the probabilities for y i are ϕ ε ε β

where ϕ and Φ are the standard normal probability density and cumulative distribution functions, respectively.

The survey produced an eligible sample of 861 participants, with 47.9% of respondents (n = 412) residing in Phoenix and 52.1% (n = 449) in Detroit. 

We begin our analysis of how food shopping behaviors and patterns changed during COVID-19, the time-period from March until May 2020. As shown in Figure 1 , 75% of the sample bought "what they can get due to empty shelves," which is in line with reported out-of-stock situations at many retailers across the country. We also addressed whether people stockpiled food, or bought more than usual. Forty-seven percent bought more food than usual and 33% stockpiled food, which is consistent with previous research conducted in March (Redman, 2020) .

Some 55% ate food they stockpiled but restocked it, while 28% ate from their stockpile without restocking it. Only 10% stockpiled food without the intention to touch it until the crisis is over. Aside from changes in general food shopping behavior, 66% stated to go less often to the store and 21% stated to go more often to the store. This finding is consistent with previous research which found that consumers rather not shop inside the grocery store when COVID-19 is actively spreading (Grashuis et al., 2020) .

Given that two-thirds of the sample went to the grocery store less often, the question arises whether shoppers replaced visits to the store with another shopping mode. To shed light on this, we asked respondents to indicate their level of participation in grocery delivery and pick-up services before and since COVID-19. Results show that, before the pandemic, 9% participated in grocery delivery and 15% in grocery pick-up. These percentages rose to 15% and 25%, respectively, since the pandemic indicating that a fair share of consumers makes use of these services (see Figure 2 ).

Because shopping at the store was viewed as risky behavior, grocery pick-up and delivery services saw a sudden spike in usage (Gray, 2020; Redman, 2020) . To take a closer look at changes in utilization rates across pickup and delivery, Table 2 provides insight into the different user-types, for instance, by differentiating whether respondents prefer pick-up or delivery exclusively or at the same time.

Continuing with those who use grocery delivery and/or pick-up programs we investigated which services are most preferred (multiple answer question), shown in Figure 3 . As a follow-up question, we inquired about the reasons why shoppers participate in grocery pick-up or delivery programs, and these responses are presented in Figure 4 . Overwhelmingly, responses were related to anxiety.

Some 75% stated they were scared of the pandemic and 66% said they were feeling unsafe. About one-third mentioned that they were too busy with work. About 21% said they had no childcare or were not healthy enough.

These two reasons are likely related to the virus, however, another 21% participate in the programs due to lack of transportation and store hours, which might be unrelated to COVID-19.

Before COVID-19, most of those who used take-out and delivery got meals or groceries once or twice a

week or less often (see Figure 5 ). While more than two-thirds of all respondents made use of take-out meals or meal-delivery, the opposite is true for grocery delivery, where more than two-third never made use of these services.

As displayed in Table 3 , over 50% of the sample uses grocery delivery more because of COVID-19, the same is true for about 40% of participants when it comes to meals. Some 30% state that their usage level is about the same as before. While about 15% get fewer groceries delivered than before, over 30% get fewer meals. This is in line with the customer-loss reported for restaurants.

While shifts in grocery shopping patterns during the pandemic had implications along the food supply chain, ultimately, we are interested in understanding the extent to which consumers' diets were shaped as a result.

Anecdotally, many pre-COVID-19 consumption habits, for example, purchasing premade salads or other fresh meals and dining out, were affected because of food safety concerns and changes in working conditions. Therefore, in seeking to better understand shifts in consumption patterns, we asked respondents to indicate how the volume and quality of food consumed changed during the lockdown. When respondents were asked "How much has your diet changed since COVID-19 started?" about 60% stated that they ate about the same amount of food as before, Grocery pick-up before COVID-19

Grocery delivery since COVID-19

Grocery delivery before COVID-19 F I G U R E 2 Participation in grocery delivery and pick-up before and since COVID-19 (%). Question: Have you participated in… Participants could enter multiple answers. Participants could enter N/A 13% said they ate less, and 21% said they ate more. Some 9% stated they eat healthier and 12% thought they eat less healthy (see Figure 6 ).

Next, we examine responses to the question "How much more or less have you consumed these foods since COVID-19 started?" and present these results in Table 4 , which shows the extent to which participants changed their volume of food consumption, ranging from eating a lot less to eating a lot more, across the 10 major food groups. Across most T A B L E 2 Participation in grocery delivery and pick-up, comparisons before and since COVID-19 and 70% of the sample, except for Snacks and Fast Food. Unsurprisingly, the majority of respondents (48.0%) indicated that meal take-out decreased. However, snack consumption increased (41.9%), most likely because people were working from home more. Aside from respondents who indicated "about the same," more people stated they ate less meat and prepped meals compared to those who stated they eat more; whereas more people stated they ate more fresh produce, dairy, grains, frozen food, canned food, and bottled water compared to those who stated they ate less.

Finally, we aim to shed light on what associations can be drawn between shopping behaviors and changes in consumption patterns during the COVID-19 lockdown period. We estimate the ordered probit model from Store hours don't work for me (n=282) F I G U R E 4 Reasons to participate in grocery pick-up or delivery programs. Question: What are reasons that you participate in grocery pick-up or delivery programs?. Participants could enter multiple answers. Participants could enter N/A. Note the category "other" (note specified) was chosen by 32.1% (n = 190) Section 2.3 using maximum-likelihood, with separate models for each food category, where the dependent variable for each model is the change in consumption since the beginning of COVID-19. In total, we estimate four model specifications, beginning with a model that only reflects demographics and income shocks due to COVID-19, and ending with a model that includes those variables as well as shopping frequency and shopping behaviors. Ultimately, model selection was determined according to the following criteria: (1) the minimum Get take-out meals or meal-delivery (e.g. from a restaurant)?

Get groceries delivered F I G U R E 5 Take-out and delivery usage before COVID-19 (% the model and comparing them with the sample probability distributions across food categories. According to these criteria, two models best fit the data. Model 1-our "preferred" model-most closely predicted observable probabilities (see Table A1 ), whereas Model 4 resulted in the lowest BIC. Thus, we present the results of the preferred model in Table 5 with marginal effects presented in Table 6 . The results of Model 4 are reported in appendix Tables A2 and A3. 1

The results in Table 5 describe the determinants of whether a consumer was more likely to move from one consumption amount to another since COVID-19 across each food group. Results from Table 5 can only be interpreted insofar as the significance and the sign; therefore, we focus our findings section on the marginal effects in Table 6 . Notably, findings show that associations between changes in food consumption and the presence of children in the household, race, and residency are statistically significant at the 90% confidence level at least. During COVID-19, respondents with children in the household were more likely to consume more fresh produce (19.1%), dairy (8.3%), and grains (11.0%). Individuals who identify as Black or African American (non-Hispanic) were more likely to consume more fresh produce (15.7%), dairy (17.0%), meat (15.1%), and bottled water (17.2%). Compared to Detroit residents, Phoenix residents were more likely to consume less fresh produce (3.8%) and grains (3.5%), but more likely to consume more frozen food (4.9%). 2

With respect to income and income shocks, we find little statistical significance across food categories, with the exception of fast food consumption. Those who were furloughed or lost their job due to COVID-19 were 11.7% more likely to indicate consuming less fresh produce. In addition, the association between the utilization of food assistance programs and consumption of dairy, meat, and grain consumption is statistically significant.

Respondents on SNAP were more likely to consume more of these three food groups (8.3%, 10.5%, and 11.5%, respectively). Additionally, respondents who visited a food pantry within the past thirty days were less likely to T A B L E 3 Change in grocery delivery and take-out during COVID-19

Change in

Get take-out meals or meal-delivery (e.g., from a restaurant)? The dummy variable for city (Phoenix vs. Detroit) represents a shift in the intercept of the regression equation. It should be noted that the dummy variable for city is intended to control for the two study sites, which might also be capturing differences in COVID regulations. That said, at the time of data collection, most states were adhering to similar shelter-in-place orders. For the remaining demographic variables, although they vary between the two cities, the estimates presented in the results tables represent the average marginal effects across the two cities, controlling for location. T A B L E 5 Determinants of changes in consumption since the beginning of COVID-19 by food group Coef.

Coef.

Coef.

Coef.

Coef. Coef.

Coef.

Coef.

Coef.

Coef. Coef.

Coef.

Coef.

Coef.

Coef. Coef.

Coef.

Coef.

Coef.

Coef. Coef.

Coef.

Coef.

Coef.

Coef. indicate consuming less canned food (−5.9%), prepped meals (−10.0%), and bottled water (−8.1%). These results are consistent with those from Model 4 (Appendix Table A3 ).

Due to concerns about potential common factors that might affect both purchase and consumption changes, we reserve Model 4 results only as a robustness check. However, when adding variables that account for shopping frequency and dietary patterns, the results from Models 1 and 4 are robust, as the sign, scale, and significance of the results obtained from Model 1 and discussed in Section 3.5.1 are largely the same. Nonetheless, additional covariates yield several statistically significant findings. Respondents who shopped more often were more likely to consume more dairy (8.2%), grains (16.0%), frozen food (8.9%), and bottled water (12.4%). Unsurprisingly, those who indicated buying more were more likely to consume more of each food category, except fast food (7.8% more likely to consume less). Finally, respondents who indicated stockpiling were more likely to consume more dairy (5.6%), meat (8.8%), frozen food (9.0%), and canned food (9.7%).

This study aimed to shed light on food purchasing behaviors, acquisition methods, and consumption during the pandemic. In this section, we synthesize our findings as they pertain to food retailers, food manufacturers, and other food industry stakeholders in navigating this new terrain.

First, based on the observed interest in grocery delivery and pick-up services, there is an opportunity for retailers to sustain market share if they offer these programs or partner with existing services like Instacart.

Retailers that have more robust logistics in place may be more prepared to seamlessly integrate grocery delivery, offering it as a permanent service, as delivery and pick-up services will continue to grow in popularity as a shopping mode for many consumers. Second, as many respondents indicated, they purchased "what they could" when shopping, so it would serve food manufacturers who are focused on capital efficiency to establish a presence in many types of stores to increase brand exposure. Because of channel agnosticism among consumers, brand continuity is crucial, hence, food manufacturers would need to build a flexible back end to offer products more ubiquitously. Finally, over half of respondents indicated some form of stockpiling food. This evidence of insecurity in the food supply chain should be noted and key considerations must be made to ensuring food safety and reestablishing trust in the food system.

As with any research conducted during this time, shopping patterns and consumption behaviors will continue to evolve, in some cases reverting to pre-COVID-19 norms. For instance, some customers might simply prefer to not go into the store during the pandemic but want to keep shopping at their preferred store, ensuring they receive their favorite products. Nevertheless, the commerce that surrounds the food industry is a crucial component of the economy whose revival is dependent on consumers' sentiments and trust in the food system. How consumers navigate food-purchasing decisions in times of uncertainty has important consequences for both food retailers and food manufacturers. Emerging from the pandemic, it will be critical to understand these behaviors to retain customers in the long-term. COVID-19 has already proven to be a disruptive event that will shape the future of the pandemic, we conducted an online survey in 2020 during the first wave of COVID-19 in two major metropolitan areas in the United States. The results that follow should be interpreted insofar as they pertain to the study sites, namely Detroit, MI, and Phoenix, AZ. With regard to food shopping, we found that about three-quarters of respondents were buying the food they could get due to out-of-stock situations, and about half the participants bought more food than usual even though the majority went to the food store less frequently. Consumers also tended to purchase more groceries than normal during their shopping trips while buying what was available due to stock-outs of commonly used and popular items. It comes as no surprise that consumers in the study areas attempted to avoid shopping during the beginning of the pandemic when no clear rules, such as wearing masks, having plastic shields for cashiers, and floor stickers that indicate six feet distance, were yet in place.

With regard to online grocery shopping since the COVID-19 pandemic, we found a 255% increase (from 4.5% to 11.5%) in the number of respondents that use grocery pickup as a shopping method. At the same time, there was a 158% increase in the number of households that utilize grocery delivery services. The surge in grocery pickup and delivery program participation could be explained mainly by consumers fearing COVID-19 (74.9%) and feeling unsafe (66.3%). However, while participants in Phoenix and Detroit were almost equally likely to order their groceries being delivered by Instacart, Amazon Fresh or grocery stores, the most popular outlet for grocery pick-up was directly from the supermarket. Therefore, to maintain such drastic growth in grocery pick-up after the threat of a virus diminishes, grocery stores have to ensure the quality and reliability of the services provided during the coronavirus outbreak. As a result, consumers' preference for a safe and reliable mode of grocery acquisition may be sustained.

Food consumption patterns were also of interest to us, as they offer key insights for food retailers and manufacturers who must adapt their inventory to satisfy imminent consumer demand. While food consumption patterns seemed to stay the same for the majority of our participants, some indicated that they had been consuming more food since the beginning of the pandemic. On the extensive margin, our results confirm an overwhelming shift away from consumption away from home (e.g., fast food) to snack food consumption. On the intensive margin, we found an increase in the consumption of fresh produce and dairy among households with children, and an increase in frozen foods and bottled water among households who were shopping more and purchasing more.

At the time of writing, the world is preparing for a "second wave" of the virus. With both improved surveillance policies (e.g., virus and antibody testing) and more stringent monitoring policies (e.g., temperature checks, local mandates to wear masks) in place, cities are better prepared not only to withstand subsequent outbreaks of the virus but also to endure strains placed on the food system at its expense. Many hope, and even take comfort, that life will return to the pre-COVID-19 norms, and, in the short-run, some reports indicate that consumer shopping is normalizing (CFI, NPD). However, other factors, such as rising food prices, may affect shopping behaviors, as agricultural labor markets face shortages and distribution channels remain tightened. In addition, the crisis has exposed significant inequities in the landscape for food retail. Future research could examine how the changes incurred would exacerbate some of the already heavy burdens on areas with food access challenges and a high proportion of food-insecure households.

To close, this study is not without limitations. First, we focused on two major metropolitan areas in the United States, and thus the results are only generalizable to the extent that the sample is representative of the U.S.

population. Future research could expand this to a nationwide survey and could also investigate food shopping and consumption in rural areas. Also, behavior in other countries likely differs, which is of interest to the food supply chain in a global marketplace. Second, we conducted our survey at a time when stay-at-home orders were being lifted. Hence, behavior might have been affected by this. Finally, as with all survey work, we rely on the recall ability of our participants which may lack accuracy. Future studies could employ other methods, such as analysis using revealed preference, that is, scanner data to supplement the findings of this study. Coef.

Coef.

Coef.

Coef.

Coef. Coef.

Coef.

Coef.

Coef.

Coef. Coef.

Coef.

Coef.

Coef.

Coef. Coef.

Coef.

Coef.

Coef.

Coef. Coef.

Coef.

Coef.

Coef.

Coef. Coef.

Coef.

Coef.

Coef.

Coef. Coef.

Coef.

Coef.

Coef.

Coef. Note: Bold indicates significance at the 90% confidence level at least.

Diagnostic and statistical manual of mental disorders

Data dashboard

Store choice and consumer behavior in food deserts: An empirical application of the distance metric method

Online grocery sales set to surge, grabbing 20 percent of market by 2025

Shopping behavior of rural consumer migrated to urban area in the Indian context: An emerging market

Hurricane katrina shakes consumers

Protecting people now, helping the economy rebound later, Mitigating the COVID Economic Crisis: Act Fast and Do Whatever

Big-name food brands lose battle of the grocery aisle

Grocery shopping preferences during the COVID-19 pandemic

Agriculture, transportation, and the COVID-19 crisis

Effects of TV advertisement on consumer buying behaviour: A comparative study of rural-urban and male-female consumers

Amazon's big breakdown

Food supply chains during the COVID-19 pandemic

Big food companies are trying to reverse the curse of the center aisle

Rural consumers' online shopping for food and fiber products as a form of outshopping

On-line shopping behavior: Cross-country empirical research

Marketers see uptick in online deliveries amid COVID-19 restrictions

Antecedents of rural and urban consumers' propensity to outshop and product specific outshopping behaviour

How the coronavirus crisis is changing grocery shopping

This is how deeply the coronavirus changed our behavior

As Amazon orders surge, coronavirus delays deliveries and threatens to put sellers out of business

Buying behaviour of rural and urban consumers in India: The impact of packaging

Coping with a natural disaster: Losses, emotions, and impulsive and compulsive buying

In grocery stores, the perimeters take center stage. The New York Times

Back to normal may not be best

Center aisle of the future. BrandPackaging

Where Do Americans Usually Shop for Food and How Do They Travel To Get There? Initial Findings From the National Household Food Acquisition and Purchase Survey