Emergency supplies purchase patterns during COVID-19 outbreak in the developing economy: frequency and stockpiling drivers’ assessment

The pandemic crisis caused by the spread of the virus severe acute respiratory syndrome coronavirus 2 (SARS-CoV-2) [coronavirus disease 2019 (COVID-19)] has influenced all areas of human activity globally. Cities suffered more significantly than rural areas of low population density did (Hamidi et al., 2020). Because of the difficulties associated with urban logistics, urban areas are highly vulnerable to people's shopping behavior changes, which cause changes in the supply chains and require new strategies to be developed (Wu and Chaipiyaphan, 2019; Chaudhuri et al., 2020; Altay and Narayanan, 2021; Thompson and Anderson, 2021). The pandemic, as is common among all crisis and disasters, affected vulnerable population groups first (Allahi et al., 2021; Babatunde et al., 2020; Breitbarth et al., 2021; Blank, 2021; Sokat and Altay, 2021). Older people, for example, had difficulty accessing their usual food supplies due to a risk of being infected during physical store visits. Given the possible severe consequences of infection for seniors (Jordan et al., 2020) the ability to safely access food supplies became crucial. As shopping activities during COVID-19 had substantial modifications, and became more online oriented, even for food items (Grashuis et al., 2020), home delivery–based supplies were considered the key option to provide older people with the food they needed (Breitbarth et al., 2021).

Along with food products, prompt access to medical supplies became a frontline concern for people's safety during the pandemic outbreak (Babatunde et al., 2020; Blank, 2021; Malmir and Zobel, 2021). The rapid demand for face masks (Worby and Chang, 2020) and personal hygiene products (Berardi et al., 2020) placed extra pressure on supply chain systems, inspiring studies to be conducted on distribution issues in adapting to the COVID-19 reality (Malmir and Zobel, 2021; Ivanov, 2021). People's purchase behaviors during the COVID-19 outbreak in regard to masks, personal hygiene products and food revealed t significant changes in consumption patterns in the face of the dangers and uncertainly of the pandemic crisis. A list of studies identified such shopping behaviors as impulsive and stockpiling (Hall et al., 2020, 2021; Zhang et al., 2020; Ahmed et al., 2021; Arafat et al., 2021; Barnes et al., 2021; Wang et al., 2020; Harahap et al., 2021; Islam et al., 2021; Kaur and Sharma, 2020; Li et al., 2020; Naeem, 2020; Xiao et al., 2020).

Lockdown measures brought further anxiety and uncertainty about the future of people's lives, provoking them to buy more than usual (Ahmed et al., 2021; Harahap et al., 2021; Kaur and Sharma, 2020; Naeem, 2020). Social media placed additional pressure on people, increasing levels of fear and concern (Arafat et al., 2021; Hall et al., 2020; Barnes et al., 2021; Islam et al., 2021). Moreover, the high density of information on the COVID-19 situation coming from local authorities also contributed to stockpiling behaviors (Xiao et al., 2020). An analysis of the amount of transactions made in retail sectors between January 2017 and December 2020 shows an increase in spending during COVID-19 greater than is normally seen for Christmas and Black Friday (Hall et al., 2021). The availability of mobile apps and online shops only facilitated people's increased shopping activity (Zhang et al., 2020; Ali Taha et al., 2021), acting as a sublimation process under the severe pandemic situation (Li et al., 2020).

Given the growing list of studies on impulsive shopping, it is worth noting that most have been implemented for countries with high levels of COVID-19 cases. The revealed roots of increased shopping have been considered in terms of social media and perceived danger. But it is worthwhile considering the roots of stockpiling behaviors, together with changes in frequency and purchase amounts, in countries with low numbers of COVID-19 cases. Such knowledge will be helpful in predicting future possible extraordinary demands for a specific list of commodities such as first-priority items and food, and to forecast a range of purchase behavior patterns. In this way suppliers will be able to take into account the sharp fluctuations in demand and meet city inhabitants' needs with an appropriate amount of food and first-priority items.

Based on the context described above, this study aimed to assess changes in frequency and amount of goods purchased before and during the COVID-19 outbreak. We assume that the government measures deployed to prevent the virus spread in regions with low case numbers negatively influenced shopping patterns of the end-consumers. Along with these measures the influence of international and local news media on shopping pattern changes during the COVID-19 outbreak will also be considered in this study. The paper is structured as follows. Section 2 is a literature review, which covers the current state of the problem revealing the major findings already obtained in regard to stockpiling behaviors along with determining the existing gaps in the studied object. The methodology chapter (Section 3) contains a description of the survey technique, the structure of the questionnaire used to collect the behavioral data and the components of the models proposed for assessing shopping frequency changes due to the COVID-19 pandemic. Section 4 provides the data collected, with a description of the formed sample, the pandemic crisis in Ukraine and the government measures deployed. Section 5 contains the results of the assessment of general shopping behavior changes and an evaluation of shopping volume changes with special consideration of the drivers involved, as well as regression analysis results on shopping frequency before and during the COVID-19 pandemic. Section 6 compares the results of this research with other studies and provides suggestions for future research directions. The last section, Section 7, contains a discussion of the major findings of this study.

Changes in people's shopping behaviors, particularly with strongly marked stockpiling during the first wave of COVID-19, have inspired a range of academic studies. Researchers have used different approaches and methodologies to reveal the roots and drivers that provoked the changes in people's shopping behaviors. Among the most widely used we can distinguish descriptive analyses (Hall et al., 2020; Harahap et al., 2021), regression analyses (Jezewska-Zychowicz et al., 2020; Kaur and Sharma, 2020) and structural equation modeling (Laato et al., 2020; Ahmed et al., 2021; Li et al., 2020; Xiao et al., 2020). The full list of methodologies and research techniques used in the reviewed studies is presented in Table 1, where the main findings of the papers devoted to stockpiling behavior in the COVID-19 pandemic time are summarized.

For most of the reviewed studies stockpiling behavior was considered as a motivation at the root of this process, without assessing the precise change in purchased amounts of products (Hall et al., 2020; Clemens et al., 2020; Laato et al., 2020; Ahmed et al., 2021; Barnes et al., 2021; Islam et al., 2021; Kaur and Sharma, 2020; Li et al., 2020; Naeem, 2020; Xiao et al., 2020). Only a few studies have focused, along with revealing the stockpiling roots, on also estimating food reserve changes (Wang et al., 2020; Jezewska-Zychowicz et al., 2020; Harahap et al., 2021). For instance, Wang et al. (2020) assessed the changes in reserves for food in China, where the pandemic started and the first lockdown measures were deployed. Based on a sample of 1,188 people from several Chinese regions that were suffering mostly due to the COVID-19 pandemic, Wang et al. revealed strong stockpiling behavior. They estimated changes in fresh food purchase frequency from 3.37 (before COVID-19) to 7.37 days (during the first wave of the pandemic), indicating an increase in purchase amount. Using the ordered logit model, Wang et al. studied the perceived utility due to food reserves in the pandemic. Along with unobserved (latent) parameters, the positive influence of the respondents' socioeconomic attributes (gender, income and education) on stockpiling patterns was also determined.

The studies by Jezewska-Zychowicz et al. (2020) and Harahap et al. (2021) estimated changes in purchase volume for a list of products. Thus, Harahap et al. analyzed the chart for most impulsive pandemic products that were bought in the USA. The leading commodities that were stockpiled by US residents were food and grocery items, with 47% increases in purchase volume. In turn, Jezewska-Zychowicz et al. revealed for Poland a slight increase in purchase volume for fruits and vegetables (+18%), and a significant stockpile effect for nonperishable goods such as bottled water (+47%), flour and sugar (+63.4%), pasta, groats and rice (+75%). Along with descriptive statistics of the shopping volume change Jezewska-Zychowicz et al. used a logistic regression analysis to assess the likelihood of fear of limited access to food during the first wave of COVID-19. The results indicated that perceived limited access to food provoked fear, forcing people to buy more products to stockpile them. In addition, Jezewska-Zychowicz et al. (2020) revealed that people became more mass media oriented during the pandemic, which was also noted by Hall et al. (2020), Kaur and Sharma (2020).

In several cases, social media usage during the pandemic was estimated as a trigger for stockpiling behavior, provoking increased levels of fear, anxiety and uncertainty in people (Ahmed et al., 2021; Harahap et al., 2021; Naeem, 2020; Islam et al., 2021). Moreover, an information overload was observed in the pandemic to cause negative effect in regard to shopping activity. Thus, Laato et al. (2020) estimated that information overload, along with mobility restrictions inclined people to make extra purchases online. This can be considered as a way that people could sublimate the perceived mix of anxiety and boredom; a phenomenon termed by Laato et al. (2020) as “cyberchondria.”

As the root of the stockpiling behavior has an emotional nature, structural equation modeling (SEM) was used in several studies (Ahmed et al., 2021; Laato et al., 2020; Li et al., 2020; Xiao et al., 2020) to trace the interconnections between different negative feelings perceived by the people and their intentions to buy more products than usual. For instance, the study made by Ahmed et al. (2021) revealed a list of events that contributed to the emergence of stockpiling behavior in the USA. Thus, the perceived fear of a complete lockdown, US stimulus checks and an increase in shopping activity all formed a snowball effect, resulting in stockpiling behavior. Li et al. (2020) also used SEM to estimate similar drivers of extra-shopping behavior in China. The perceived severity of the pandemic, along with a reduction of individuals' perceived control, has influenced people's intentions to buy extra amounts of food and health-care products. The findings from the study by Xiao et al. (2020) correlate with the results in Li et al. (2020). Also, it should be mentioned that SEM-based studies have revealed that latent fear and uncertainty not only gave rise to stockpiling behaviors but also panic shopping according to Clemens et al. (2020), who obtained it using ANOVA, as well as Ahmed et al. (2021). The US was the case study for both studies, where the number of COVID-19 cases was very high. Considering Italy, where the pandemic crisis was very severe, Barnes et al. (2021) revealed that lack of perceived control and uncertainty forced people to stockpile products.

Due to stress, fear and uncertainty about how long the pandemic would last, shopping behaviors tended to become stockpile oriented. Such behaviors are similar to the crisis patterns that occur, for instance, after disasters and hazards (Holguín-Veras et al., 2014, 2016). But most of the studies reviewed in this chapter assessed the drivers of the stockpiling behavior from the qualitative point of view, revealing the latent drivers and events that influence people's shopping activity. Only a few of the studies tried to assess the changes in shopping patterns from a quantitative point of view (Wang et al., 2020; Jezewska-Zychowicz et al., 2020; Harahap et al., 2021), without deep statistical analysis of the phenomena. Also, it should be noted that reviewed papers considered the countries with more or less severe pandemic conditions which obviously influenced people's behavior. But in addition to these countries, stockpiling behavior has been observed in countries with low penetrations of cases during the first global wave of COVID-19. In these cases the drivers of shopping behavior changes can be studied within entirely different conditions, whereby people did not observe the pandemic through a comparable lens of personal danger.

As most of the reviewed studies focused on assessing the roots of stockpiling behaviors due to the COVID-19 pandemic without estimating the quantitative changes of the purchase process, we aim to reveal the roots of the stockpiling behavior by analyzing specific quantitative attributes of the shopping process. Thus, shopping frequency change is a very important parameter, with ramifications that influence entire supply chains. The definition of shopping frequency predictors will allow us to forecast changes in shopping frequency that can occur due to a pandemic crisis, such as COVID-19. Such analyses would enable forecasting that would allow for more prompt adjustments in the supply chain in response to demand changes, and would reduce the probability of commodity shortages in the retail system during this and possible future crises. Taking into account the findings in Wang et al. (2020) obtained for China where the COVID-19 crisis was severe, we would like to test the hypothesis that shopping frequency changes in those countries with low numbers of COVID-19 cases also had socioeconomic roots. As such, we propose to use a regression analysis methodology to assess the predictors of this process. So, the shopping frequency can be presented as the following function:

where $SFi$ is the shopping frequency for a person i; $βeconomic$ is a vector of economic parameters; $βsocial$ is a vector of social parameters; $Xeconomic(i)$ is a vector of economic attributes of a person i; $Xsocial(i)$ is a vector of social attributed to a person i; $εi$ is a random error representing the discrepancy in the approximation (Chatterjee and Hadi, 2015).

Assuming that function $f(βeconomic,βsocial,Xeconomic(i),Xsocial(i))$ has a linear type, $SFi$ can be presented as follows:

where $βincome$ is a personal income parameter; $βgender$ is a gender parameter; $βchildren$ is a parameter reflecting the influence of children's availability in a household; $βadults65$ is a parameter reflecting the influence the number of senior people over 65 years old in a household; $βeducation(j)$ is a parameter for j education level; J is the total number of education levels considered in the study; $βemployment(k)$ is a parameter for k employment status of the people from the sample; K is the total number of employees levels considered in the study; $βage(h)$ is a parameter for h age group; H is the total number of age groups of the people from the formed sample; $Incomei$ is a personal monthly income (UAH); $Genderi$ is a binary attribute for a gender which equals to one if person i is female and zero if male; $Childreni$ is a binary attribute which equals to one if there are persons younger than 18 years old in a household; $Adults65i$ is the number of senior people over 65 years old in a household; $δeducation(j)$ is a binary attribute for the education level which equals to one if person i has the education level j; $δemployment(k)$ is a binary attribute for an employment status of a person i which equal to one if person i has an employment status k; $δage(h)$ is a binary attribute for age which equals to one if person i belongs to age group h.

As we can see, model (2) considers the economic and social attributes of the consumers. We paid special attention to age and employment status as the COVID-19 pandemic influenced differently people from different age groups and working activities. Also, we emphasized senior people availability in households as this age group was most vulnerable to the COVID-19 virus (Breitbarth et al., 2021; Sokat and Altay, 2021).

Taking into account the changes in shopping behavior during the pandemic with stockpiling patterns we can assume the following:

where $SFipre-COVID−19$ is the shopping frequency of a person I in pre-COVID-19 time; $SFiCOVID−19$ is the shopping frequency of a person I during the COVID-19 pandemic.

Given that $SFipre-COVID−19$ and $SFiCOVID−19$ are determined by the same predictors $Xeconomic(i),Xsocial(i)$ we can trace the influence of every $xi$ on shopping frequency. In this case, every change in the statistical significance level for $β$ values will allow us to reveal the predictors of the shopping frequency in pre-COVID-19 and during the pandemic times.

As the regression model allows us only to determine the predictors of shopping frequency, we should also reveal the roots of stockpiling behavior in the context of low cases of COVID-19. In this case we would like to know what source of information or event inclined people to feel concerns for their personal safety and then stimulated them to make their emergency supply purchases. To reveal this, we proposed using a five-point Likert scale (Likert, 1932) allowing us to evaluate what influenced people's emergency supply purchases during the pandemic outbreak. So, having a list of sources (local news, national news, etc.), we can reveal which made a significant impact on people's purchasing behavior, increasing their levels of fear and uncertainty. According to the five-point Likert scale methodology we assign five points, with 5 as having a “strong impact,” and 1 as having no impact, or “did not taken into account.” Hence, given the sample of I persons with their evaluation of S number of sources, we can assess the average value of the estimated score for every s:

where $EP¯s$ is the average score for an information source s; $EPs(i)$ is the estimated score for s information source by person I; $I$ is the sample size.

Having S number of $EP¯s$ we can range them from maximum to minimum to evaluate the most and least influencing sources of the information on people's intentions to stockpile emergency goods during the COVID-19 outbreak. It should be mentioned that along with maximum value of $EP¯s$ every source of information should have the minimum standard deviation for $EP¯s$⁠.

Along with the above-described tasks, we also aim to reveal the behavioral roots of stockpiling, for instance, did people buy extra items to sell them, as was observed in the USA (The New York Times, 2020)? Or, were the extra amount purchases made due to personal health and safety concerns? In this case, we would like to know why and for what purposes people started to stockpile emergency goods during the first wave of the pandemic. To reveal these roots, we propose using a five-point Likert scale to assess the list of possible drivers for people's intentions to buy more than usual.

Given the formalized aims of this study we can form the research frame to show the general interconnections of the issues studied in Figure 1.

To achieve the aims described within this section, the survey must be developed to collect the necessary data. The major steps of survey development are described in the next section of the chapter.

The polling was based on the Rensselaer Polytechnic Institute (RPI) Institutional Review Board Protocol (Troy, NY, USA) “Purchasing Behavior During Times of Crisis,” which was developed under the leadership of Dr. José Holguín-Veras and was adapted to the Ukrainian language and economic attributes by the authors. The protocol was presented as a questionnaire aimed at “gaining the insight into the behavioral determinants that explain the purchases of supplies by individuals before and after the crisis.” It should be noted that this paper's research hypotheses were developed by the authors, and RPI's protocol was used only as a research frame to collect the necessary data on shopping behaviors during the COVID-19 crisis in Ukraine. The questionnaire was made on a Survey Monkey platform. The snowball sampling technique was used to collect the data. The chain referral sampling allows us to conduct the survey with a small initial number of respondents given the big list of surveyed units under the final stages of the data collection (Biernacki and Waldorf, 1981). We used social media channels to disseminate the questionnaire within the first round of polling, as personal contact was not possible due to lockdown measures. Within the first-round the participants were asked to disseminate the questionnaire among their contacts to form the chain with the research group. In such a way under every next polling round people had to involve other people in the polling, ultimately providing the snowball effect.

The questionnaire was divided into four sections. The first contained the questions about normal purchases of basic supplies, namely, food items, personal hygiene items, cleaning supplies and medical supplies. The questions were aimed at revealing purchase frequency, the duration of goods usage, store type used for purchase (convenience store, pharmacy, supermarket etc.), and the availability of basic supplies at home in normal times. To evaluate whether the shopping behaviors had changed temporarily, only for the duration of the crisis, or whether they had become permanent, people were asked to assess their shopping channel usages in the after COVID-19 era. This section allows us to understand the respondents' shopping behaviors in normal times, when they do not have any concerns about their safety, or their ability to access goods in the future.

The second part of the questionnaire was focused on assessing the emergency basic supply purchases made in response to the COVID-19 crisis. Given the pandemic situation, the questions in this chapter were targeted on evaluating possible changes in shopping behaviors. The basic attributes under consideration were the following: when respondents started purchasing emergency supplies in response to COVID-19, the frequency of those purchases, and the reasons and factors that fueled these purchases during the crisis. Also, within the second chapter, the respondents were asked to point to which goods they could not purchase during COVID-19 because of the supply deficits. This gave us important information about which goods the respondents considered high-priority and vital in the pandemic. Special attention in this section was given to the shopping frequency assessment and the drivers that caused people to buy the emergency supplies in the outbreak. These questions allowed us to evaluate why people changed their shopping patterns. Participants were also asked to evaluate what sources influenced them to purchase emergency supplies in the pandemic. The five-point Likert scale used allowed us to define the influence of the external drivers for stockpiling behaviors.

The third section had a list of questions about relief organizations, which allowed us to define people's attitudes and trust toward local and national authorities in the COVID-19 crisis. The list of organizations covered the different levels of authorities, from international relief organizations to local level authorities such as national/state/local health officials and local relief organizations. The score scale from the previous section was used to evaluate the level of trust that the respondents associated with each organization from the list.

The final part of the questionnaire was devoted to socioeconomic data collection, allowing us to connect each participant with his/her shopping behaviors before and during the COVID-19 pandemic.

The polling was conducted from June 2 to June 24, 2020 in three Ukrainian cities: Kharkiv, Zaporizhzhia and Kyiv. Kharkiv is the second largest city of Ukraine, with 1.44 million inhabitants. Kharkiv's economy is based on industrial and trade sectors, with thirteen national universities and a host of professional and private higher education institutions. The second city surveyed, Zaporizhzhia, has a similar economic profile as Kharkiv as Zaporizhzhia is a large Ukrainian industrial center, known for the fields of metallurgy, automobile and motor manufacturing. The city is located in the south of Ukraine, with a population of 750,000 people. The city has eight national universities and 16 professional and technical institutions. The third city under study was Kyiv, the capital of Ukraine. It has the largest population in the country: 2.9 million people. Kyiv's economy is based on utilities (electricity, gas and water supply), the manufacture of food, beverages, and tobacco products, and is known also for mechanical and chemical industries. The education sphere is also well developed; Kyiv has 30 national universities, 13 academies and a host of professional and technical schools. With the similarities among the cities, we can consider them as one polygon for the data collection.

The polling was administered through the Survey Monkey survey platform. At the first stage, the authors initiated the questioning through their contact lists, requesting that personal contacts fill in the questionnaire. At the subsequent stages, participants were asked to expand the polling process using their own contact lists. In this way, the research team received responses from 208 Ukrainian households, but after data cleaning the final sample became 136 units. We excluded from the sample 72 questionnaires because of insufficient or missing answers and in cases where the households were located outside of the cities of focus. As survey participation was voluntary, and the questionnaire was quite long, some of the participants avoided answering several questions within the sections. This necessitated data cleaning and sample reduction. The final sample comprised the following numbers of households within each city: Kharkiv – 42 units, Zaporizhzhia – 75 units and Kyiv – 19 units. The breakdown of this sample is presented in Table 2.

As we can see in Table 1, age groups are well represented in the sample. In terms of gender, females are more represented, (66.91%), which can be explained by the cultural features of Ukrainian households in regard to shopping activity. According to the Ukrainian Statistics Bureau, females make up 54% of the total population (State Statistics Service of Ukraine, 2020). The sample stratification on education level shows the biggest share for people with master's and Ph.D. degrees (49.36%), which t can be explained by the well-developed education infrastructure in the surveyed cites and a larger number of people with university degrees living in the large cities (McCormick and Wahba, 2005). That is why the sample has a slight dominance in the higher education level than is seen in the total population of Ukraine. According to the State Statistics Service of Ukraine the weight of people with “high school” education level is the largest (61.5%) within the total population.

In terms of household size, the sample shows all shares of the total population but with a slight predominance of three-person households. This can be explained by the features of Ukrainian households' size consisting of an average of three persons (mean is 2.6 persons based on State Statistics Service of Ukraine data). Within the sample we covered mostly employed people. Their attitude to perceived danger under COVID-19 is highly relevant since during the lockdown measures economic activity was reduced, and people could lose their jobs.

The first case of COVID-19 in Ukraine was revealed on March 3, 2020. After that for ten days no cases of infection were detected, but on March 11, a 23-day quarantine was deployed across Ukraine. The measures consisted of closing schools, a ban on mass events (more than 200 people) and closing flight connections with countries affected by the COVID-19 pandemic (based on infection dynamics at the time). On March 13, the first fatal case in the country due to COVID-19 was identified (Zhytomyr region, Radomyshl town). On March 16th the government announced the closing of the Ukrainian borders for foreign entrants. On the next day, measures for subway closings were deployed for Kyiv, Kharkiv and Dnipro. Besides that, public transit was allowed to service people with low vehicle occupation rates, which caused congestion and crowds of people at the stops. The additional number of road transport vehicles was not provided by the carriers.

Given the potential danger of COVID-19, but owing to the low number of actual cases, the Ukrainian government introduced a 30-day emergency regime across Ukraine on March 25. In the time period from March 25 to 30, the average number of daily cases number was 78. After April 6, the government deployed additional quarantine measures such as a stay-at-home order for senior people, a ban for visiting parks, and an order to wear a face mask in public places. The dynamics for cases revealed is presented in Figure 2, based on the Johns Hopkins Coronavirus Resource Center (2020). Since March 3 until August 1, the cumulative number of deaths in Ukraine was 1,733 people. In terms of the studied cities, the cumulative death cases over the same time period were as follows: Kharkiv region – 128 people, Zaporizhzhia region – 20 people and Kyiv (capital region) – 134 people.

As shown in Figure 2, the quarantine measures were deployed chaotically. For instance, since April 6, when it was introduced, the ban on visiting parks took place when the average daily cases number was 220, but growing. But, since May 11, within the first stage of easing the quarantine measures, this ban was lifted when the number of infected people was still growing, with an average of 440 daily cases. The same patterns can be seen in the next stages of easing quarantine measures. The timelines for easing quarantine measures are presented with green arrows in Figure 2, and described in detail in Table 3.

Such disproportionate government policies resulted in continued growth of infections. As a result, at the time this paper is being written (November 5, 2020), the number of confirmed daily cases in the country had risen to 10,138 people.

Since August 1, the government introduced adaptive quarantine measures that had to be deployed only in regions with high infection rates. The measures were related to mobility restrictions, the wearing of masks, social distancing orders and a ban on mass events. After the restrictions were introduced, a great number of violations were detected. People demonstrated a weariness and distrust of the government, as severe restriction measures in March and April were deployed for all regions of Ukraine while some regions had no cases of COVID-19 infection at all. Thus, the early intimidation and warning strategy failed and garnered an inverse effect.

According to hypotheses formed for the survey, we focused on an assessment of consumers' shopping patterns before and during the first wave of COVID-19 in Ukraine. For the first wave we made an estimation of changes in shopping preferences in regard to food items, cleaning products, personal hygiene products and medical supplies. Within each one of these categories we specified a list of products, for instance, for food items we chose water, meat, vegetables, rice etc. (grocery), canned goods and other (which can be determined by the consumer). The same lists were identified within all categories. The analysis results for this data are presented in Figure 3. In this way, we revealed the changes in end-consumers' shopping patterns as they responded to the COVID-19 crisis. Thus, in the case of food items, we observed an increase in long shelf-life food shopping (grocery and canned goods) and a reduction in expenditures on vegetables and bottled water. The same shopping strategy was noted in Spain, where the COVID-19 crisis was pronounced and people perceived a strong sense of danger and fear (Laguna et al., 2020). To protect the home space from possible infection, people increased purchases of bleach (+9.62%), cleaning wipes and disinfectant spays (+2.19%). The self-protection was implemented through stockpiling purchases of hand sanitizers (+10.34%) from the “personal hygiene products” group, face masks (+6.76%), N95 respirators (+3.8%) and gloves (+7.96%) from the “medical supplies” category.

The analysis of purchase priorities within the four described goods categories revealed the predominance of personal hygiene products and medical supplies versus foods items and cleaning products. Based on that, we can draw conclusions about changes in shopping behaviors unrelated to personal safety. Within the usual life rhythm, the average person mostly concentrates his/her shopping activity on food, considering the buying of personal hygiene products as secondary. As we can see, during the pandemic crisis that shopping paradigm changed. Along with goods shopping patterns, we have to assess changes in the average purchase volumes made during the COVID-19 crisis.

The consumers were asked to respond to four framing statements in terms of their shopping behavior changes: “I purchase more quantity of goods,” “I purchase more frequently,” “I changed retailers,” and “I switched from buying in-store to buying online.” The results of these statements show that 88.0% of respondents started purchasing in greater quantities, and 12.0% switched to online purchasing. Nobody noted that they had increased their shopping frequency or changed retailers. The breakdown of respondents with stockpiling behaviors who shifted to online shopping are presented in Figure 4 according to income groups.

Figure 4 shows the prevalence of stockpiling behaviors among people with middle incomes. This correlates with the sample stratification based on household income, where the middle-income households presented the biggest share. Nevertheless, we can trace stockpiling behaviors among almost all income groups. Also, we should point out the shift to online shopping, which can be explained by the fear of being infected in physical stores. These findings both correlate with the shopping behaviors seen during the first wave of COVID-19 in China (Li et al., 2020; Xiao et al., 2020; Wang et al., 2020), USA (Ahmed et al., 2021; Clemens et al., 2020; Harahap et al., 2021; Unnikrishnan and Figliozzi, 2020; Grashuis et al., 2020), New Zealand (Hall et al., 2020; Martin-Neuninger and Ruby, 2020) and Poland (Jezewska-Zychowicz et al., 2020).

To assess the key sources revealed stockpiling patterns, we used a five-point Likert scale where five points reflect a strong impact, and one point reflects no impact. The list of the sources assessed that influenced consumers’ decisions to purchase emergency supplies during the COVID-19 pandemic is presented in Figure 5.

As shown in Figure 5, a high importance has been estimated for factor #1 “International news,” #2 “National news,” #4 “Electronic social media,” and #6 “I have to work from home.” We can conclude from this the high impact of social media, together with lockdown restrictions, in influencing stockpiling behaviors. The wide range of points seen for source #8 “Reports of shortage” indicates an absence of hysteria during stockpiling shopping. This is supported with estimation results for source #10 “Family, friends and neighbors were purchasing.” So, the stockpiling effect can be explained, with its potentially dangerous consequences within the Ukraine territory, by the mass media coverage of worldwide pandemic cases. The introduction of lockdown measures even under small case numbers, increased the effect of a potential threat.

In this case, the personal drivers and motives that pushed people to stockpile goods are highly important, and valuable for future planning purposes. People could have bought some sanitizers, for instance, for resale, as was detected in New York City in March 2020 (The New York Times, 2020). The results for roots assessment in the frame of stockpiling behaviors are presented in Figure 6.

According to the data in Figure 6, we can see the pro-safety intentions for stockpiling behaviors. For instance, the statement “Concern for myself and my family” was estimated as important and highly important by 87% of the respondents. Also, we observed the time limit factor based on the statement “I will need them this week.” Besides that, we determined some fear in regard to out of stock situation. The evaluation of statements “I was afraid they would run out,” “I was afraid I would not be able to buy them later” along with “I was afraid the stores would close” reveals that about 44% of the respondents were highly concerned about possible difficulties associated with purchasing goods in the future. As for purchases made for commercial purposes, we did not determine this intention as more than 84% of the respondents rejected such behavior.

The lockdown measures with mobility restrictions affected end-consumers' shopping frequency. In Ukraine, people were allowed to go outside, but not more than two persons in one group. As for the two weeks of self-isolation for people who crossed the border, these restrictions were quite severe. They were allowed to implement shopping activity only within two kilometers of their dwelling (Ukraine Government Portal, 2020). For people under mobility restrictions, we suppose a reduction in shopping frequency. The detected increase in shopping volume should complement a frequency change. Given this correlation, we used a multiple regression analysis to determine shopping frequency pattern changes due to the COVID-19 pandemic. The attributes we used in the socioeconomic data breakdown are presented in Table 2. Shopping frequency was reflected by the number of first-priority goods purchased during a month. This data was collected for two periods of time: before and during the COVID-19 pandemic. The regression analysis results are presented in Table 4.

As the results of the conducted regression analysis show, we determined a reduction in shopping frequency due to the COVID-19 pandemic. The sample size influenced significant levels for some socio-demographic attributes. Nevertheless, the conducted analysis revealed shopping frequency patterns. Of particular note are changes in significance levels for attributes “Income” and “Number of adults with age 65 years old and older in a household.” These indicate the increase in the importance of people's economic level in the pandemic. In addition, the inclusion of adults aged 65 years and older in households forced a reduction in shopping frequency. We can explain that with COVID-19's higher risk for seniors. Under such conditions we see that people tried to reduce the possibility of bringing the virus home by decreasing the frequency of their in-store attendance.

Taking into account the shopping volume analysis, we formed a general frame for end-consumers' purchase behaviors before and during the pandemic crisis, the results of which are presented in Figure 7.

The analysis of average stockpile volumes before and during the COVID-19 pandemic indicates an almost twofold increase in the duration of purchased goods usage. This is due to the significant growth of purchased goods volumes. In addition, we determined a decline in shopping frequency by an average of two days. Given the low number of cases in Ukraine during the survey period, we can conclude that severe restriction measures provoked high concerns among people about COVID-19's potential danger. This in turn forced them to anticipate worst-case scenarios and stockpile extra reserves of emergency supplies. Based on this, we can evaluate the revealed shopping behavior as panic stockpiling. In such situations the mass media played a key role in creating an environment of fear that was perceived by end-consumers during the first stages of the COVID-19 pandemic.

The conducted study shows that along with estimating the roots of stockpiling behaviors seen due to COVID-19, special attention should be given to an assessment of the quantitative attributes of the purchasing process, such as shopping frequency and purchased amounts of goods. Combining the quantitative and qualitative estimations allows us to reveal the external factors that influenced changes in shopping behaviors, and to trace their impact on people's perceived need to stockpile emergency supplies during the pandemic. Considering Ukraine as a case study, where the pandemic was not severe in the spring of 2020, we revealed similar stockpiling patterns as seen with consumers in the US (Ahmed et al., 2021; Clemens et al., 2020; Harahap et al., 2021), China (Li et al., 2020; Xiao et al., 2020; Wang et al., 2020) and Italy (Barnes et al., 2021). The obtained findings suggest that deployed lockdown measures, along with imposed mobility restrictions, formed a strong stockpiling effect in Ukraine. The consumers' orientation to international mass media, which had been broadcasting the pandemic crisis around the globe along with very low COVID-19 cases in the Ukrainian regions, inclined people to purchase greater quantities of goods to be prepared for worst-case scenarios. A similar effect was determined in Finland (Laato et al., 2020), New Zealand (Hall et al., 2020; Martin-Neuninger and Ruby, 2020) and Poland (Jezewska-Zychowicz et al., 2020) where the number of COVID-19 cases was quite low during the first stages of the global pandemic, and where lockdown measures were still introduced.

The deep analysis of shopping frequencies for emergency supplies allowed us to reveal a slight reduction in shopping activity from seven times in a month (pre-COVID-19 time) to five times in a month during the pandemic outbreak. But together with that, we revealed an almost twofold increase in average stockpile volume, namely from 11 days to 21 days. Such disproportional increase in purchase volumes, along with a slight reduction in shopping frequency, indicates strong stockpile patterns during the pandemic. Even under low numbers of cases, which could have been perceived as low risk, Ukrainian consumers showed deep concern about their families and themselves. The regression analysis results for shopping frequency confirmed these findings. Thus, the statistical significance of an attribute “number of adults with age 65 years old and older in a household” denotes that households with senior people reduced their shopping frequency to minimize the risk of senior household members being infected through visits to shops during the outbreak. This finding confirms results seen in studies by Babatunde et al. (2020), Breitbarth et al. (2021), Sokat and Altay (2021), where humanitarian issues during the pandemic crisis suggested that the needs of vulnerable populations should be considered first.

The findings revealed in this study are a good basis for policymaking in the field of supply chain management. Having models for the prediction of shopping frequencies would enable retailers to forecast the amount of emergency goods that should be supplied to the retail network to meet demand changes. Besides that, the purchase process should be controlled by local authorities to prevent the extra stockpiling of emergency supplies by households. The snowball effect among consumers produced by panic over possible f goods shortages would be controlled and leveled. Information about changes of consumed amounts can act as a foundation to deploy restrictions for selling items to provide people with necessary quantities of goods, but without facilitating extra volume shopping. This will allow for controlling the level of shortages and reduce the level of the panic buying. Effective information management is a crucial element, not only for providing an efficient response to a crisis such as the pandemic (Altay and Labonte, 2014; Dubey et al., 2021) but also for preventing the spread of panic and fear among people during the pandemic. The government mass media should balance the information coming from the international mass media to reduce panic and to more effectively control peoples' purchasing behaviors.

As further steps, we expect to conduct a behavioral study based on random utility modeling (RUM). With the stated-preference methodology, we are able to develop a list of the possible scenarios for people's shopping behaviors and mobility under different elements of the restriction measures deployed by the authorities. Significant efforts in the field of RUM usage for humanitarian issues were made by Cantillo et al. (2018) in assessing deprivation cost (Holguín-Veras et al., 2013). But this research direction is in its initial stages. We assume that the perceived utility of choice in the pandemic will differ from the utility in normal times. This forms a huge opportunity for academia to reveal changes in peoples' behavior patterns caused by the COVID-19 pandemic.

The research presented in this paper reveals significant findings on shopping behaviors during the pandemic crisis. Many cities and megalopolises around the globe have faced the problem of citizens' extra shopping activity during this crisis. All of them had common, specific roots or causes for this stockpiling behavior. Within this research, we considered Ukraine, where the number of the COVID-19 cases during the first wave was quite small, but the study revealed similar shopping behaviors and features as were detected in the USA, China and Italy, where the pandemic situation was very bad.

The three Ukrainian cities considered for the case study have a similar economic basis and cultural–educational environment, which allowed us to collect the survey-based data on consumption behavior changes due to the COVID-19 pandemic crisis in one sample. During the analysis of the consumption behaviors, we determined a change in purchased goods and volumes. Thus, Ukrainian consumers shifted to long shelf-life foods (rise, pasta, noodles and canned goods), bleach, hand sanitizers, face masks and gloves. The analysis of purchase volumes revealed an increase in shopping volume by 88.0% of the respondents from the sample. The assessment of the key drivers for the stockpiling behavior, determined from responses made on a five-point Likert scale, showed the high impact of mass media (both international and national news) and electronic social media as influencers. A big contribution to stockpiling behaviors has been the result of changes in job activity, as most of the business and education sectors were shifted to teleworking. The defined roots for consumers’ behavior reflected concerns and fears about personal and family safety during the pandemic crisis.

The shopping behavior changes also concerned purchase frequency changes. We observed an average reduction in shopping frequency from seven times in a month (before the COVID-19 pandemic) to five times in a month (during the first wave of COVID-19). Along with that, the average stockpile volume of emergency supplies changed from 11 to 21 days. An almost twofold change in supply usage indicates a high level of fear and concerns about goods availability and accessibility in the pandemic. A comparison of the obtained results with other studies for several countries confirm the same panic shopping patterns that were determined for the US, China, Italy, Finland, New Zealand and Poland.

The human behavior aspect became crucial for global and national supply chains during the pandemic. We have shown that, amid uncertainty about disease progression and potential infection rates, policy measures deployed to prevent the spread could have a reverse effect and increase panic among people. As a result, along with promoting behavior changes in people's buying activities, we should pay special attention to preparing efficient local initiatives (Frennesson et al., 2021) that should be appropriate to conditions of specific areas. Under such conditions, having the COVID-19 case statistics for specific areas, local authorities would be able to predict changes in people's shopping behaviors and deploy appropriate measures that would not severely affect other areas of the country.

The authors would like to acknowledge Prof. José Holguín-Veras and all members of the Center for Infrastructure, Transportation, and the Environment at Rensselaer Polytechnic Institute for sharing the questionnaire template and polling administration.

Funding: This research was supported by J. William Fulbright Foreign Scholarship Board and administered by the Bureau of Educational and Cultural Affairs, United States Department of State with the cooperation of the Institute of International Education in case of Fulbright Visiting Scholar Grant “Seamless last mile logistics for sustainable cities (SMILE Logistics)”.

Conflicts of interest: The authors have no conflicts of interest.