Weather sensitivity analysis on grocery shopping

Abstract

Although weather is known to impact consumer behavior and, accordingly, businesses react to weather-influenced consumer behavior, marketing scholars have not examined weather marketing as intensively as its practical importance suggests. To fill this research void, we choose the grocery shopping setting, where weather influences shoppers’ shopping trip incidence and basket size. We theorize that the weather event (i.e., rain, snow, thunder, and fog) decreases both the shopping trip and the basket size. On the shopping trip, however, the negative weather impact is mitigated for less frequent shoppers because such shoppers have a higher basic shopping need than more frequent shoppers. Similarly, in terms of the basket size, store familiarity lessens such a negative weather impact because shoppers in a familiar environment are more comfortable about unfavorable weather. Besides, when sustained bad weather is expected, shoppers may turn proactive in determining when to go shopping. Our empirical application combines rich scanner panel data and daily weather data to explain both shopping trip and basket size simultaneously in the form of a Type 2 Tobit model. Our analysis results support our weather marketing hypotheses.

Keywords

grocery panel data Tobit model weather marketing

Introduction

Weather influences such consumer choices as store visit and spending amount across industries. Inclement weather often turns consumer mood negative and makes a shopping trip less likely. On the other hand, some weather stimulates consumers’ need for certain products (e.g., ice cream in hot weather). Based on these consumer reactions to various weather conditions, businesses, especially retailers focused on consumers’ store visits, develop and implement specific marketing strategies. For example, Walmart found that ideal weather for berry sales is associated with low wind and temperatures below 80°F. Accordingly, in such weather, Walmart enhances berry displays, while promoting berries through mobile advertisements (Neff, 2014). Similarly, weather influences new product development and consumer reviews. Therefore, weather consulting businesses are prospering using analytics for weather marketing. For example, WeatherTrends360° offers 11-month-ahead weather forecasting to its clients such as Anheuser-Busch and Unilever. The Met Office, the United Kingdom’s national weather service, emphasizes “Weather Sensitivity Analysis,” quantitative analytics designed to uncover and quantify the precise relationship between the weather and business performance (Agnew & Thornes, 1995).

Compared to such tremendous practice activities, weather marketing research has been limited. On one hand, most research has focused on a link between psychological constructs in consumer behavior caused by a weather event (WE) and consumers’ purchasing attitude (Simonsohn, 2007). On the other hand, most current empirical research in the area has explored the effects of WEs on an aggregate-level variable (e.g., total product sales). Since Steele’s (1951) pioneering study, where he developed a regression model to measure the extent of weather variables on the total sales of a department store, there have been limited studies devoted to weather marketing. For example, Murray, Di Muro, Finn, and Leszczyc (2010) found that as exposure to sunlight increases, negative affect decreases and, accordingly, consumer spending increases. Using both laboratory and field data, Zwebner, Lee, and Goldenberg (2014) revealed a temperature-premium effect: warm temperatures increase consumers’ product valuation because physical warmth generates positive feelings through the physiological homeostasis process. Primarily, these studies prove that weather conditions affect consumers’ store visit and spending at the aggregate level.

Compared to most weather marketing studies that examined consumers’ psychological processes or aggregate-level consumer data, our research explores the effects of WEs on individual consumer-level shopping behavior (specifically, the effects of a WE on their purchasing incidence and basket size), utilizing a combination of rich scanner panel data and daily weather data. While existing studies primarily illuminated the main effects of various weather conditions influencing retail sales, our study extends the current literature by examining the influence of consumer-related moderators that can mitigate the negative impact of bad weather on retail consumers (i.e., shopping trip frequency [TF] and store familiarity [SF]). These moderating effects would not be possible without an individual-level examination. In addition, we extend the current research stream to grocery shopping, one of the industries constantly affected by daily weather changes, as most households conduct a grocery shopping trip at least once a week. Bertrand, Brusset, and Fortin (2015) found that during the spring season, 52% of supermarkets sales anomalies can be explained by temperature. Two of the most important decisions of a grocery shopping trip are when to go shopping and how much to spend on the shopping trip. We look into how favorable or unfavorable weather influences these two decisions in the grocery shopping environment. Specifically, we focus on how two shopping environment variables, TF and SF, mitigate the impact of unfavorable WEs (i.e., rain, snow, thunder, or fog) on the two shopping decisions. When store managers can understand these mitigating factors, they can prepare effective marketing strategies for their stores, targeting the right shopper segment instead of treating all visiting shoppers in the same way during bad weather.

Conceptual background and hypotheses development

Effects of weather and TF on store visit

Consumers’ decisions on when to go shopping have important managerial implications for retailers. Retailers need to be prepared for an expected number of shoppers for the day and keep the right product assortment according to their forecast of what shoppers will buy. Through an analysis of grocery panel data, Kahn and Schmittlein (1989) identified strong preferences for shopping to occur on a specific day of the week rather than based on when a product is used up. Meyer-Waarden and Benavent (2009) found that grocery shoppers typically visit multiple outlets and that more frequent shoppers are more likely to enroll in loyalty programs. Kim and Park (1997) found that shoppers are heterogeneous in terms of their shopping frequency and shopping trip regularity (random vs routine). In addition, Gilbride, Inman, and Stilley (2015) showed that the shopper’s trip budget explains the shopper’s unplanned versus planned purchase behavior.

Another major factor known to influence when to shop is weather conditions. Most weather marketing studies have focused on the direct relationships between weather conditions (e.g., temperature, precipitation) and product sales (Parker & Tavassoli, 2000; Steele, 1951). These studies show that favorable weather conditions (e.g., sunny day) increase consumer spending through a positive mood created by good weather, whereas unfavorable weather conditions decrease consumers’ spending through negative mood created by bad weather. Mood is described as a phenomenological property of a person’s subjectively perceived affective state. It is viewed as a mild, transient, generalized, and pervasive affective state, not an intense emotion, and not directed at specific target objects (Swinyard, 1993). Broadly speaking, individuals in positive mood states have been shown to evaluate stimuli more positively than individuals in neutral or negative mood states. Furthermore, the literature supports the same mechanism that mood caused by weather conditions influences active or generous behavior in various areas such as university admission decisions (Simonsohn, 2007) and tipping at hotel casinos and restaurants (Rind, 1996). Finally, some studies have reported better performance on the stock market during sunny and summer days versus rainy and winter days (Hirshleifer & Shumway, 2003).

Parker and Tavassoli (2000) explained the variation in consumption patterns in response to different stimuli levels created by weather conditions, specifically, temperatures and exposure to sunlight. According to the Wundt curve (Orth & Bourrain, 2005), psychological homeostasis, measured by psychological pleasantness, is the highest at the optimal stimulation level, which is summarized as the inverted U-shaped relationship between the intensity of stimulation and pleasantness. Notably, even before consumers decide to buy some product items, they have to decide whether or not to shop, which is influenced by weather conditions. Existing studies indicate the same patterns in the relationship between the weather and store traffic (Agnew & Thornes, 1995). Specifically, when there is a bad WE such as rain, snow, thunder, or fog, people’s mood turns negative, which reduces the probability that they go shopping. Because consumers consider the weather forecast in their relevant decision making and weather forecasts are easily obtained, the negative impact of a bad WE on store visit can take place on the following day (Guo, Wilson, & Rahbee, 2007). Furthermore, same-day and 1-day weather forecasting is quite accurate these days (Robbins & Robbins, 2015).

Contrary to the widely accepted finding that bad weather has a negative impact on various human activities, if sustained bad weather is expected (e.g., rain on two consecutive days), consumers may still decide to go shopping in spite of negative mood caused by the current bad weather. They may realize their need to buy some products to manage their consumption life efficiently. For example, if it is expected to rain both today and tomorrow, forward-looking shoppers may consider their shopping needs to cover the extended bad weather period more seriously (e.g., “Since it will rain both today and tomorrow, I may as well do my shopping today”). Therefore, we hypothesize that forward-looking consumers decide to go shopping proactively when inclement weather is sustained on multiple days. In other words, if the rain today is expected to continue tomorrow, shoppers may increase their probability of visiting a grocery store today in spite of the current bad weather. This takes place because bad weather tends to make them more prudent and more risk-averse (Bassi, Colacito, & Fulghieri, 2013). This knowledge can help grocery stores better prepare their product assortments and inventories.

According to Simonsohn’s (2007) study as to the influence of mood on information processing style, sad moods are linked to increased analytic processing, with a greater focus on detail. Specifically, negative mood states inform people that they may be facing a problem, which provokes systematic processing of information and motivates them to engage in effortful cognitive processing. Conversely, in benign states, positive mood motivates people to engage in less effortful heuristic processing (Schwarz & Clore, 1983). Similarly, people in negative mood tend to have a lower willingness to take risk than in neutral or positive mood (Yuen & Lee, 2003). This association of negative mood and risk aversion implies that shoppers can be more proactive in shopping to avoid getting forced into shopping in bad weather (e.g., running to a grocery store to buy baby milk in the midst of thunderstorm). In summary, we propose the following three hypotheses related to the direct impacts of weather on shoppers’ store visit.

H1(a): Consumers are less likely to visit a store on a day with a WE (i.e., rain, snow, thunder, and fog).

H1(b): Consumers are more likely to visit a store today if consecutive WEs on both today and tomorrow are expected.

Shoppers react to weather conditions differently. On a bad weather day, some shoppers can delay their shopping with ease if they have a low need to shop (e.g., “We are stocked up on almost everything”). By contrast, other shoppers may still have to go shopping despite the bad weather because of an urgent need (e.g., “We ran out of baby formula last night”). Studies examining when to shop indicate the household’s shopping frequency as a way to estimate their need to shop (Kim & Park, 1997). Bawa and Ghosh (1999) proposed a model of grocery shopping behavior, based on the view that households are rational inventory managers who seek to minimize both travel costs and inventory costs. Their empirical results support the contention that households’ shopping behavior reflects how they balance the costs while satisfying their consumption needs. Because infrequent shoppers are likely to have a higher need to shop than frequent shoppers on the given day in terms of their shopping transaction cost and opportunity test (Campo, Gijsbrechts, & Nisol, 2000), we predict that infrequent shoppers are more likely to go shopping in bad weather than frequent shoppers. For example, there will be a higher proportion of random shoppers who need to buy urgent items such as milk, diapers, and sandwich buns in bad weather than during normal weather (Kahn & Schmittlein, 1989). This knowledge can help retailers adjust their store marketing strategy, targeting less frequent shoppers on bad weather days. As a result, store managers can promote these products proactively to lessen the negative impact of bad weather on their business using mobile advertising and smart shopping carts in real time (Hui, Inman, Huang, & Suher, 2013). Thus, we propose the following hypothesis:

H2: The negative impact of a WE on consumers’ store visit is mitigated for less frequent shoppers.

Effects of weather and SF on shopping basket size

The amount a shopper spends on a given shopping trip is of primary interest to retailers. Shopping basket size is determined by consumption needs dictated by shoppers’ demographic characteristics and shopping trip frequency dictated by the balance of travel costs and inventory costs (Bawa & Ghosh, 1999). Shopping basket size is also associated with store price format (Desai & Talukdar, 2003). For example, large-basket shoppers prefer everyday low-priced (EDLP) stores because of the lower expected basket price. That is, large-basket shoppers purchase many product categories and, accordingly, pay attention to pricing across a wide range of categories. Moreover, they are relatively less flexible in terms of taking advantage of occasional price deals (Bell & Lattin, 1998). As previously indicated, weather marketing studies have found that favorable weather increases consumers’ spending through positive mood, whereas unfavorable weather decreases consumers’ spending through negative mood, which is the same weather-mood-shopping behavior mechanism we used to develop H1 above (Murray et al., 2010; Parker & Tavassoli, 2000; Steele, 1951).

Furthermore, weather has the potential to reduce consumers’ self-control and increase impulsiveness. Consumers’ impulsive choice is traditionally attributed to contextual factors, such as the store environment (Mishra & Mishra, 2010). Gailliot (2014) found that environmental temperatures can affect a person’s complex cognition and attention control. Mood caused by weather is also associated with consumers’ purchase risk-taking (Spies, Hesse, & Loesch, 1997). That is, sunny weather puts shoppers in a good mood, which they may misinterpret as optimism about the market, thereby making them likely to take more risk; rainy weather, on the other hand, puts shoppers in a negative mood, which they misinterpret as pessimism, thereby making them likely to take less risk (Bassi et al., 2013). Accordingly, the negative mood and pessimism in bad weather make shoppers more risk-averse, resulting in less spending by avoiding less important spending and impulsive spending (Isen, Nygren, & Ashby, 1978). Therefore, we propose the following hypothesis:

H3: Consumers are likely to reduce their shopping basket size on a day with a WE.

When consumers encounter bad weather, their reaction to reduce spending can be mitigated in a more familiar environment because a negative mood caused by bad weather can be lessened by environmental familiarity. Simonsohn (2007) found that individual judgments vary when examining the same data on different occasions, suggesting that situational factors (e.g., weather) can influence decision making. Bettman (1973) maintained that consumers’ perceived risk decreases with their product familiarity. In a similar vein, Miranda (2011) empirically found that extra daylight during daylight saving time allows shoppers to have additional grocery trips to their familiar grocery stores. Simonin and Ruth (1998) concluded that brand familiarity moderates the strength of relations between constructs in a manner consistent with information integration and attitude accessibility theories. Although people use their moods as the basis for forming evaluations of objects (Schwarz & Clore, 1983), mood states do not influence evaluation when the object being evaluated is highly familiar (Salovey & Birnbaum, 1989). Thus, increased familiarity can reduce perceived risk related to decision making (Hawes & Lumpkin, 1986). In examining the impact of weather on consumer spending, this mechanism remains effective because SF can reduce the negative mood and associated degree of information processing created by bad weather. Oppositely, an unfamiliar store environment will make shoppers more involved in processing information stimuli with respect to the amount of information and the difficulty of processing information. Then, the effect of mood on shopping intentions will be greater in more involved (i.e., less familiar) shopping situations (Swinyard, 1993).

Regarding how mood influences information processing style, happy moods facilitate increased heuristic processing, broader categorizations, and the consideration of a wider range of inputs, whereas sad moods are linked to increased analytic processing with a greater focus on detail (Simonsohn, 2007). Therefore, this mechanism predicts that on cloudier days, shoppers tend to place additional weight on attributes more closely related to the decision (e.g., product price and quality), which makes shoppers more prudent and more conservative in making a purchase. By contrast, on sunny days, shoppers tend to increase their attention to situational factors (e.g., weather, mood, background music). On the other hand, favorable store information that can arise from frequenting a familiar store is known to positively influence perceptions of quality and value, as well as subjects’ willingness to buy (Dodds, Monroe, & Grewal, 1991).

To conclude, as shoppers become more risk-averse in bad weather, they tend to become more conservative in purchasing products, which results in less spending. However, SF can decrease such risk aversion because of shoppers’ enhanced positive perceptions and, accordingly, can mitigate the negative impact of bad weather on consumer spending. Therefore, we propose the following hypothesis:

H4: The negative impact of a WE on consumers’ shopping basket size is mitigated by their SF.

Weather sensitivity analysis and empirical results

Data and key variable operationalization

We combined grocery panel data (Rubio & Yagüe, 2009) and weather data to test the hypotheses proposed. To the best of our knowledge, this study is the first to combine the two types of data to examine weather marketing. Specifically, we obtained extensive grocery panel data from Symphony IRI. These data contained seven years of store information (2001–2007) for Eau Claire, Wisconsin. The market data provide complete information on both the store and panel dimensions. We obtained daily weather data from weatherunderground.com for the city during the corresponding period. The weather data included daily temperatures and WEs (i.e., rain, snow, thunder, or fog) (see Table 1).

Table 1.

Variable summary.

Continuous variable	Mean	Standard deviation
Basket size	US$37.39	37.69
Temperature discrepancy (TD)	24.2°F	20.6
Trip frequency (TF)	363 times	207
Store familiarity (SF)	44.9%	31.1
Family size	2.4 people	1.2
Categorical variable	Frequencies
Shopping trip	24.0%
Weather event (WE)	45.5%
Store price format	High mean and high variance (HILO) = 29.0%, High mean and low variance = 35.4%, Low mean and high variance = 25.1%, Low mean and low variance (EDLP) = 10.5%
Residential type	Owner = 84.9%, Renter = 15.1%
Income (per year)	US$0 to US$9,999 = 3.5%; US$10,000 to US$11,999 = 3.1%; US$12,000 to US$14,999 = 4.1%; US$15,000 to US$19,999 = 5.0%; US$20,000 to US$24,999 = 8.3%; US$25,000 to US$34,999 = 14.5%; US$35,000 to US$44,999 = 12.5%; US$45,000 to US$54,999 = 12.6%; US$55,000 to US$64,999 = 9.5%; US$65,000 to US$74,999 = 8.5%; US$75,000 to US$99,999 = 11.0%; US$100,000 and greater = 7.4%
Marital status	Married = 69.0%
Zip code	54701 = 36.4%; 54703 = 49.6%; 54720 = 8.8%; Others = 5.2%

There are two WE variables—Weather Event: Today (WETD) and Weather Event: Today and Tomorrow (WETT).

To capture the inverse U-shaped relationship between temperature and pleasantness, TD is defined to be the absolute degree difference between the daily mean temperature and the most comfortable temperature (68°F = 20°C).

We divided the combined data with a total of 2,868,873 shopping trip observations (i.e., both shopping days and non-shopping days on a daily basis) from 2,352 households into three periods: (a) initialization (1 year = 2001), (b) estimation (5 years = 2002 ~ 2006), and (c) validation (1 year = 2007). We used the initialization-period data to compute the household-specific TF variable. TF was computed as the natural log of the number of shopping trips during the initialization period for each household. SF was the proportion of the shopper’s visit to the particular store out of all the shopper’s store visits up to the shopping day during the data period. Furthermore, the store price format is considered to be a primary variable in defining the store environment, as implied in retailing terms such as HILO (a promotional pricing store with a high price variation) and EDLP (with a low price variation) (Bell & Lattin, 1998). We used this store price format measure to account for price format heterogeneity across stores by dividing the stores in the city market into the four groups specified in Table 1. To determine high versus low, we used the median of the average and the standard deviation of the prices of the available stores in the initialization period, respectively. The key weather variable in our analysis is WE. This binary variable indicates whether there is one of the four WEs—rain, snow, thunder, and fog—on the given day. People recognize such WEs clearly because they create unfavorable driving conditions and psychological barriers. We combined the four types of WEs into one variable for two reasons. First, we theorized our hypotheses at the WE construct level instead of measuring the magnitude of weather effect by the WE type. Second, snow and fog occur much less frequently than rain and thunder, which makes the empirical measurement of the weather impact of these two weather effects more difficult.

Model specification

Our weather sensitivity analysis (WSA) model has two successive outcome variables: whether household h goes shopping on the given day t and, if the shopping trip happens, how much the household spends. This structure falls into the Tobit model (Baltas, 1998), particularly Type 2. Besides, this is a censored regression model, as opposed to a truncated regression model, because the exogenous variables are observed for the day even when the household does not go shopping.

Let $y_{h t}^{1}$ be a binary variable indicating whether or not household h goes shopping, given day t. It is assumed that the probability that household h goes shopping ( $y_{h t}^{1} = 1$ ) is provided by the following logit model:

\Pr (y_{h t}^{1} = 1 | x_{h t}^{1}) = Φ (x_{h t}^{1} β^{1})

(1)

where $x_{h t}^{1}$ indicates regressors, and β¹ are corresponding parameters. Φ is the cumulative distribution function of the standard normal distribution. Once we determine when household h goes shopping, we want to know how much the household spends ( $y_{h t}^{2}$ ).

\Pr (y_{h t}^{2} | x_{h t}^{2}) = x_{h t}^{2} β^{2} + ε_{h t}^{2}

(2)

The $y_{h t}^{2}$ indicates the shopping basket size in the natural log. The $ε_{h t}^{2}$ follows the normal distribution. The $x_{h t}^{2}$ indicates regressors. The β² are the corresponding parameters. The two components need to be combined into a general modeling framework, given that they are two successive decisions that cannot be separated. In other words, the shopper’s basket size ( $y_{h t}^{2}$ ) is conditional on the shopper’s binary shopping decision ( $y_{h t}^{1}$ ). This problem nicely fits into the Tobit Type 2 model framework.

Empirical results

Table 2 shows that the weather model with interactions is consistently the best model in three different indices. It should be noted that the differences among the three models in the three comparison indices are relatively small, given that the only differences in the three models arose from changing the set of the independent variables with the same Tobit Type 2 model structure. However, we emphasize the fact that the full model consistently improves the explanatory power and predictive validity, which means that the weather interaction term explains households’ shopping behavior in a meaningful way. Notably, the same pattern supporting the superiority of the weather model with interactions is demonstrated in three different samples: frequent shoppers only, infrequent shoppers only, and the general sample with both groups of shoppers. The overall empirical results of the full model are summarized in Table 3.

Table 2.

Model comparison.

	Base model without weather variables	Weather model without interactions	Weather model with interactions
General case (for both frequent shoppers and infrequent shoppers)
Estimation for both shopping trip and basket size
Log likelihood	−1,372,002	−1,365,210	−1,358,418
BIC	2,744,842	2,731,314	2,717,776
Validation for shopping trip
Hit rate	68.0%	68.9%	70.4%
Validation for basket size
Mean absolute percentage error	68.6%	66.7%	65.6%
Frequent shoppers only
Estimation for both shopping trip and basket size
Log likelihood	−707,217.5	−707,162.7	−703,263
BIC	1,425,892	1,418,864	1,407,418
Validation for shopping trip
Hit rate	68.5%	68.9%	70.6%
Validation for basket size
Mean absolute percentage error	67.2%	66.2%	65.1%
Infrequent shoppers only
Estimation for both shopping trip and basket size
Log likelihood	−660,251	−656,982	−653,583
BIC	1,321,541	1,315,028	1,308,055
Validation for shopping trip
Hit rate	68.8%	69.3%	70.8%
Validation for basket size
Mean absolute percentage error	67.4%	65.9%	65.2%

BIC: Bayesian Information Criterion.

Boldfaced values indicate that they are focal variables.

Table 3.

Estimation results of the weather sensitivity analysis (WSA) model.

Variable	Shopping trip	Basket size
Intercept	−3.9533***	−4.4661***
Weather event: Today (WETD)	−.0606***	−.2610***
Weather event: Today and Tomorrow (WETT)	.0489***	–
Trip frequency (TF)	.5491***	–
WETD × TF	−.0039***	–
Store familiarity (SF)	–	−.1106***
WETD × SF	–	.1288***
Store price: High mean and high variance	–	.1771***
Store price: High mean and low variance	–	.1792***
Store price: Low mean and high variance	–	.0944***
House owner (vs renter)	−.0174***	.0913***
Family size	.0088***	.0711***
Income	.0018***	.0346***
Household head age	−.0027***	.0044***
Married	.0016	.0795***
Zip code 54701	−.0370***	.0167**
Zip code 54703	−.0451***	.0584***
Zip code 54720	−.0213***	−.0097
Monday	.0372***	−.1394***
Tuesday	.0197***	−.1762***
Wednesday	.0570***	−.1511***
Thursday	.0412***	−.0837***
Friday	.1051***	.0038
Saturday	.0582***	.0516***
February	.0691***	.0038
March	.1041***	.0116
April	.0748***	−.0277***
May	.1059***	.0312***
June	.1247***	.0274***
July	.1367***	−.0055
August	.1215***	−.0096
September	.1086***	−.0304***
October	.1307***	−.0138*
November	.1519***	.0471***
December	.0656***	.0854***
Variances	σ² = 1.0455*; σ¹² = −0.2863*

TD: temperature discrepancy.

The reference categories for the categorical variables used are the following: low mean and low variance (for store price format), others (for zip code), Sunday (for day-of-the-week), and January (for month). σ¹ is set to be 1 for model identification. We did not include both WETD and TD variables in the same model because of their high correlation.

Significant at 10%; **significant at 5%; and ***significant at 1%.

Our WSA model in Table 3 estimates two outcome variables: shopping trip occasion and basket size. First, the Weather Event: Today (WETD) variable shows a significantly negative impact on both outcome variables, thus supporting the proposed H1(a) and H3 hypotheses. That is to say, unfavorable weather conditions discourage consumers from going shopping, H1(a). Even when they do go shopping, bad weather discourages them from spending, thus resulting in a smaller basket size (H3). (Notably, we also tested the same model in Table 3 with the WE variables replaced with the corresponding temperature discrepancy (TD) variables and obtained practically the same results. We did not include both WE and TD variables in the same model because of their high correlations.) By contrast, the Weather Event: Today and Tomorrow (WETT) variable shows a significantly positive impact, thus supporting H1(b) (Bassi et al., 2013; Yuen & Lee, 2003).

Next, we find that the TF and WETD interaction term (WETD × TF) has a significantly negative impact on the shopping trip, while the main TF term has a significantly positive impact. This means that although frequent shoppers are more likely to go shopping on any given day, their shopping trip probability decreases further on a bad weather day than less frequent shoppers, thereby supporting H2. Infrequent shoppers have on average more need to shop on any given day, given that they keep less grocery stock at home and are more likely to be forced into going to a store to buy some urgent items. Now, we observe that the SF and WETD interaction term (WETD × SF) has a significantly positive impact on basket size, while the main SF term has a significantly negative impact. A familiar store signifies a store that shoppers frequently patronize. Because they go to the familiar store frequently, they have fewer items to purchase for each trip and, therefore, end up having a smaller basket size. However, on a bad weather day, their spending is reduced less in familiar stores than in unfamiliar stores, insofar as SF can reduce the negative mood and associated degree of information processing by bad weather, thus supporting H4. For example, on a very cold day, some shoppers may feel pressured to go home quickly, which will reduce their basket size. However, in a familiar store, even in such a high-pressure situation, they can manage to find what they want quickly because they are already familiar with the store environment.

Table 3 also shows how other control factors, such as demographics, day-of-the-week, and month, influence shopping trip occasion and basket size. There are several notable results. First, high-income households tend to go shopping more frequently and generate larger baskets, pointing to their increased ability to spend more on grocery store items. Similarly, increased family size is positively associated with more frequent shopping trips and bigger basket size, which reflects that a larger household has a greater need to shop.

Second, in the day-of-the-week influence (Kahn & Schmittlein, 1989), Friday and Saturday turned out to be the most active shopping day in both shopping trip and basket size, as people prepare for their weekend activities including parties and social gatherings. By contrast, shoppers have the lowest likelihood to go grocery shopping on Sunday but, once they do, they tend to spend a lot (see Figure 1). Finally, in monthly effects, whereas the Eau Claire residents visit grocery store less often during the winter, they tend to spend more perhaps in order to adjust to longer intervals between shopping trips.

Discussion

Compared to weather-related research originating in various disciplines such as meteorology and finance, marketing scholars have generated limited research on weather marketing within the past few decades. This undesirable status has motivated our study. Whereas the existing weather marketing literature has mostly focused on the direct impact of weather on consumer behaviors, such as store visits and product purchases, we extend the findings by investigating the marketing factors that can mitigate the negative impact of bad weather (TF and SF) in the context of grocery shopping because marketers such as store managers can use such information to deal with unfavorable weather.

Toward that end, our empirical analysis combined weather data with grocery panel data, which makes this study the first to combine the two sources of data for weather marketing. Our empirical findings include the following. First, we found that bad weather produces a negative impact on consumers’ shopping behavior, specifically, by reducing both shopping trip occasion and shopping spending, H1(a) and H3. Second, we found that such a negative impact is lessened by infrequent (versus frequent) shoppers because infrequent shoppers tend to have a higher and more urgent need to shop on any given day (H2). Third, by the same token, shoppers tend to be less affected by bad weather in a familiar versus an unfamiliar store because a familiar environment can assuage their negative mood (H4). Finally, when sustained bad weather is expected, shoppers may turn proactive in determining when to go shopping, H1(b).

From a managerial perspective, store managers can use the results of this weather research in managing and merchandising their stores. Specifically, store managers can use TF and SF to mitigate a negative weather impact. For example, because store managers expect to meet more low-frequency shoppers on bad weather days, they can be more proactive in increasing their store visit frequency in the long term through their loyalty program campaigns using mobile advertising on bad weather days (Felgate, Fearne, DiFalco, & Martinez, 2012). Managers can encourage the installation of mobile applications for their stores and issue promotional coupons for shoppers to increase their store visits in bad weather (Mistry & Samant, 2012). Similarly, they can identify preference differences between loyal shoppers who are more familiar with their stores and store switchers who are less familiar with their stores. In bad weather, they can adjust their store displays toward store switchers because they are more affected by bad weather. In a broad sense, when store managers understand how weather influences their customers’ behaviors and decisions, they can manage product assortment and promotional tactics more effectively.

With the emergence of Big Data, public weather data are becoming more detailed, and weather forecasting is becoming more accurate. Accordingly, weather marketing companies are increasing their services by using technological advances, such as mobile devices and social media. For example, consumers frequently use smartphones to find nice restaurants. The current weather conditions can influence consumers’ biological and psychological conditions, which can also influence where and what they want to eat. Because this stream of Big Data-based weather marketing research is feasible these days, we suggest this stream of research for future study.

While this study addresses in-store shopping, consumers’ reactions to changing weather may be different in online shopping. Therefore, future research on the effects weather has on website traffic and online shopping conversions will be instrumental in establishing a successful digital marketing strategy (Otzasek, 2015; Weather Unlocked, 2014). Interestingly, Marshall and Pires (2017) found that weather can affect consumers’ use of online shopping outlets. Similarly, Chu, Chintagunta, and Cebollada (2008) suggest that consumers’ online product selection can vary based on weather conditions. Based on these studies, it would be worthwhile investigating how various weather conditions will influence consumers’ choices between online and in-store shopping.

Footnotes

Funding

The author(s) received no financial support for the research, authorship, and/or publication of this article.

References

Agnew

M. D.

Thornes

J. E.

(1995). The weather sensitivity of the UK food retail and distribution industry. Meteorological Applications, 2, 137–147.

Baltas

(1998). An integrated model of category demand and brand choice. International Journal of Market Research, 40, 295–306.

Bassi

Colacito

Fulghieri

(2013). O Sole Mio: An experimental analysis of weather and risk attitudes in financial decisions. Review of Financial Studies, 26, 1824–1852.

Bawa

Ghosh

(1999). A model of household grocery shopping behavior. Marketing Letters, 10, 149–160.

Bell

D. R.

Lattin

J. M.

(1998). “Shopping behavior and consumer preference for store price format: Why “large basket” shoppers prefer EDLP. Marketing Science, 17, 66–88.

Bertrand

J.-L.

Brusset

Fortin

(2015). Assessing and hedging the cost of unseasonal weather: Case of the apparel sector. European Journal of Operational Research, 244, 261–276.

Bettman

J. R.

(1973). Perceived risk and its components: A model and empirical test. Journal of Marketing Research, 10, 184–190.

Campo

Gijsbrechts

Nisol

(2000). Towards understanding consumer response to stock-outs. Journal of Retailing, 76, 219–242.

Chu

Chintagunta

Cebollada

(2008), A Comparison of within-household price sensitivity across online and offline channels. Marketing Science, 27, 283–299.

10.

Desai

K. K.

Talukdar

(2003). Relationship between product groups’ price perceptions, shopper’s basket size, and grocery store’s overall store price image. Psychology & Marketing, 20, 903–933.

11.

Dodds

W. B.

Monroe

K. B.

Grewal

(1991). Effects of price, brand, and store information on buyers’ product evaluations. Journal of Marketing Research, 28, 307–319.

12.

Felgate

Fearne

DiFalco

Martinez

M. G.

(2012). Using supermarket loyalty card data to analyze the impact of promotions. International Journal of Market Research, 54, 221–240.

13.

Gailliot

M. T.

(2014). An assessment of the relationship between self-control and ambient temperature: A reasonable conclusion is that both heat and cold reduce self-control. International Review of Social Sciences and Humanities, 8, 149–193.

14.

Gilbride

T. J.

Inman

J. J.

Stilley

K. M.

(2015). The role of within-trip dynamics in unplanned versus planned purchase behavior. Journal of Marketing, 79, 57–73.

15.

Guo

Wilson

Rahbee

(2007). Impact of weather on transit ridership in Chicago, Illinois. Transportation Research Record: Journal of the Transportation Research Board, 2034, 3–10.

16.

Hawes

J. M.

Lumpkin

J. R.

(1986). Perceived risk and the selection of a retail patronage mode. Journal of the Academy of Marketing Science, 14(4), 37–42.

17.

Hirshleifer

Shumway

(2003). Good day sunshine: Stock returns and the weather. Journal of Finance, 58, 1009–1032.

18.

Hui

S. K.

Inman

J. J.

Huang

Suher

(2013). The effect of in-store travel distance on unplanned spending: Applications to mobile promotion strategies. Journal of Marketing, 77, 1–16.

19.

Isen

A. M.

Nygren

T. E.

Ashby

F. G.

(1978). Influence of positive affect on the subjective utility of gains and losses: It is just not worth the risk. Journal of Personality and Social Psychology, 55, 710–717.

20.

Kahn

B. E.

Schmittlein

D. C.

(1989). Shopping trip behavior: An empirical investigation. Marketing Letters, 1(1), 55–69.

21.

Kim

B.-D.

Park

(1997). Studying patterns of consumer’s grocery shopping trip. Journal of Retailing, 73, 501–517.

22.

Marshall

Pires

(2017). Measuring the impact of travel costs on grocery shopping. The Economic Journal. Advance online publication. doi:10.1111/ecoj.12523

23.

Meyer-Waarden

Benavent

(2009). Grocery retail loyalty program effects: Self-selection or purchase behavior change? Journal of the Academy of Marketing Science, 37, 345–358.

24.

Miranda

(2011). Motivating shoppers to expand purchases during daylight-saving time. Journal of Services Marketing, 25, 202–214.

25.

Mishra

(2010). We are what we consume: The influence of food consumption on impulsive choice. Journal of Marketing Research, 47, 1129–1137.

26.

Mistry

Samant

(2012). How research assisted the rollout of a mobile agriculture information service: The day Peepli went [live]. International Journal of Market Research, 54, 589–602.

27.

Murray

K. B.

Di Muro

Finn

Leszczyc

P. P.

(2010). The effect of weather on consumer spending. Journal of Retailing and Consumer Services, 17, 512–520.

28.

Neff

(2014, October 27). Cloudy with a chance of meatballs: How weather forecast predicts Walmart’s Sales Outlook. Ad Age. Retrieved from http://adage.com/article/dataworks/weather-forecast-predicts-sales-outlook-walmart/295544/

29.

Orth

U. R.

Bourrain

(2005). Ambient scent and consumer exploratory behaviour: A causal analysis. Journal of Wine Research, 16, 137–150.

30.

Otzasek

(2015, August 26). How weather influences the conversion rate and Google shopping purchasing behavior. Whoop! Retrieved from whoopapp.com/blog/weather-influences-google-shopping-purchasing-behavior

31.

Parker

P. M.

Tavassoli

N. T.

(2000). Homeostasis and consumer behavior across cultures. International Journal of Research in Marketing, 17, 33–53.

32.

Rind

(1996). Effect of beliefs about weather conditions on tipping. Journal of Applied Social Psychology, 26, 137–147.

33.

Robbins

N. B.

(2015, January 29). How accurate are weather forecasts? The Huffington Post. Retrieved from https://www.huffingtonpost.com/richard-robbins/how-accurate-are-weather-_b_6558770.html

34.

Rubio

Yagüe

M. J.

(2009). The determinants of store brand market share: A temporal and cross-sectional analysis. International Journal of Market Research, 51, 501–519.

35.

Salovey

Birnbaum

(1989). Influence of mood on health-related cognitions? Journal of Personality and Social Psychology, 57, 539–551.

36.

Schwarz

Clore

G. L.

(1983). Mood, misattribution, and judgments of well-being: Informative and directive functions of affective states. Journal of Personality and Social Psychology, 45, 513–523.

37.

Simonin

B. L.

Ruth

J. A.

(1998). Is a company known by the company it keeps? Assessing the spillover effects of brand alliances on consumer brand attitudes. Journal of Marketing Research, 35, 30–42.

38.

Simonsohn

(2007). Clouds make nerds look good: Field evidence of the impact of incidental factors on decision making. Journal of Behavioral Decision Making, 20, 143–152.

39.

Spies

Hesse

Loesch

(1997). Store atmosphere, mood, and purchasing behavior. International Journal of Research in Marketing, 14, 1–17.

40.

Steele

A. T.

(1951). Weather’s effect on the sales of a department store. Journal of Marketing, 15, 436–443.

41.

Swinyard

W. R.

(1993). The effects of mood, involvement, and quality of store experience on shopping intentions. Journal of Consumer Research, 20, 271–280.

42.

Weather Unlocked. (2014, August 21). Weather and eCommerce: How weather impacts retail website traffic and online sales. Weather Unlocked. Retrieved from www.weatherunlocked.com/blog/2014/august/weather-and-ecommerce-how-weather-impacts-retail-website-traffic-and-online-sales

43.

Yuen

K. S. L.

Lee

T. M. C.

(2003). Could mood state affect risk-taking decisions? Journal of Affective Disorders, 75, 11–18.

44.

Zwebner

Lee

Goldenberg

(2014). The temperature premium: Warm temperatures increase product valuation. Journal of Consumer Psychology, 24, 251–259.