Abstract
Purpose
Little research on the influence of external factors, such as weather and holiday periods, on retail sales on alcoholic beverages is available. This study aims to investigate how weekly retail sales of different alcoholic beverages vary in association with daily maximum temperatures and annual federal holidays across selected US counties in the years 2013 to 2015. The research provides information, which can contribute to better sales forecasts.
Design/methodology/approach
Secondary data of weekly retail sales (volume) of alcoholic beverages from 37,346 stores in 651 counties in the USA are analysed. The data cover on average 21% of all existing US counties and 12% of the total US off-trade retail sales of alcoholic beverages in the period studied (Euromonitor, 2017). Additional data of federal holidays and meteorological data are collated for each county in the sample. Seasonal autoregressive integrated moving average models with exogenous regressors (SARIMAX) are applied to develop forecasting models and to investigate possible relationships and effects.
Findings
The results indicate that off-trade retail sales of beer, liquor, red and white wine are temperature sensitive throughout the year, while contrary to expectations rosé, sparkling and other wines are not. Sales sensitivities to temperature also differ by geography. In the warmest regions, liquor and white wine sales do not respond to temperature changes, as opposed to the coolest regions, where they are responsive. Public holidays, particularly Easter, Thanksgiving, Christmas and New Year holidays, represent a constant influencing factor on short-term sales increases for all investigated alcoholic beverage categories.
Originality/value
This is the first large-scale study of weather and holiday-related sales variations over time, across geographies and different alcoholic beverage categories. Seasonal and non-seasonal short-term sales variations are important for retailers and manufacturers alike. Accounting for expected changes in demand accommodates efficiencies along the supply chain and has implications for retail management, as well as adjusting marketing efforts in competing categories.
Keywords
Citation
Hirche, M., Haensch, J. and Lockshin, L. (2021), "Comparing the day temperature and holiday effects on retail sales of alcoholic beverages – a time-series analysis", International Journal of Wine Business Research, Vol. 33 No. 3, pp. 432-455. https://doi.org/10.1108/IJWBR-07-2020-0035
Publisher
:Emerald Publishing Limited
Copyright © 2020, Emerald Publishing Limited
1. Introduction
Retailing plays an important role for the distribution and sales of alcoholic beverages in the US. For the time period of this study (2013–2015), off-trade retailing contributed on average 75% of the total sales volume (30 billion L) in the category (Euromonitor, 2017). About 80% of the sales by volume is beer, 11% wine and 9% spirits. The retail industry in general is highly affected by climate and weather events. For 2008, the economic sensitivity to weather variability in the US is estimated to be 3.4% ($485bn) of the gross domestic product (Lazo et al., 2011). Business and academia are aware of the importance and potential of predicting consumer behaviour in relation to weather events (Brooke, 2017). There are clear indications that weather variability and seasonality affect the retail sales of alcoholic beverages. Holiday periods have also been shown to influence retailing (Agnew and Thornes, 1995).
The main objective of this study is to examine possible associations of daily maximum temperatures, as well as holiday periods with retail sales of different alcoholic beverages over time and across geographies, and how such information can contribute to better sales forecasts. We apply seasonal autoregressive (AR) integrated moving average models with exogenous regressors (SARIMAX) to test possible associations and develop forecast models. The data cover weekly sales of alcoholic beverages in 37,346 retail stores in 651 counties in the US over a period of three years. The comprehensive nature of this research approach is new and enriches understanding of consumer behaviour. It adds to the existing literature and has significant practical implications, as well as implications for future research. This study should be considered at the observational and exploratory stage of scientific discovery forming the basis for generating formal hypotheses that can be tested in future research.
This study examines potential seasonal patterns in off-trade retail sales of various alcoholic beverages and tests if any sales effects also differ by geography. We further investigate if public holidays are associated with changes in sales for all beverages. The findings represent an important contribution for marketers, retailers and the scientific community. New knowledge of the influence of climate and holiday periods on sales outcomes has implications for various marketing and management areas. Timing, as well as location-based decisions regarding, for example, sales promotions, supply and inventory planning can be informed by suitable demand forecasting models at the product category level in defined markets. Future research may consider demand fluctuations over time when planning and conducting related studies.
2. Literature review
External conditions, such as weather and felt temperature, affect consumers in their response to marketing stimuli (Bruno et al., 2017; Li et al., 2017). However, the dynamics and variability of consumer preferences and respective retail sales over time are not easy to detect, let alone to predict. The practical value of weather forecasts and the prediction of demand because of weather variability and demand sensitivity is well-acknowledged in the literature (Agnew and Thornes, 1995; Harrison, 1992; Curtis, 2003; Lazo et al., 2011; Steele, 1951; Arunraj et al., 2016). Previous research revealed clear indications that climate, seasonality and weather events can affect economic outcomes, for example, in the US (Bujisic et al., 2017; Lazo et al., 2011; Starr-McCluer, 2000), the UK (Agnew and Palutikof, 1999; Agnew and Thornes, 1995), Germany (Arunraj and Ahrens, 2016), South Korea (Bahng and Kincade, 2012) or Japan (Yohannes and Matsuda, 2016).
There are also clear indications that weather sensitive consumption varies geographically within markets (Govind et al., 2014). Weather events can influence consumers’ choice of products affecting total sales (Murray et al., 2010). For example, a study of restaurant visitors demonstrates how temperature changes correlate with variations in total sales and the selection of menu items (Bujisic et al., 2017). Research suggests that climate can have very contrasting effects on different consumer goods and services industries. For example, clothing and footwear see declining sales with increasing temperatures, whereas the drinks industry is a great beneficiary of hot weather (Agnew and Palutikof, 1999; Agnew and Thornes, 1995; Harrison, 1992; Bahng and Kincade, 2012). Within beverage categories, it is suggested that decreasing temperatures see sales of tea bags rise and those of soft drinks fall relative to the warmer seasons (Harrison, 1992). Sales of meat and meat products, fresh produce, soft drinks and ice cream are some of the product categories named to be specifically weather sensitive (Curtis, 2003). As opposed to product category level sales, total grocery store sales are not significantly affected by temperature changes (Maunder, 1973). This is likely because of the constant demand for food and beverages, and substitution across categories over time.
However, studying how weather events and climate affect economic outcomes is very complex. The nature of supply chains, logistics and infrastructure, as well as consumer decision-making along with other possible situational variables, can all directly or indirectly be affected by weather and ultimately influence sales performance (Agnew and Thornes, 1995; Maze et al., 2006; Arunraj and Ahrens, 2016; Maunder, 1973; Steele, 1951; Murray et al., 2010; Parsons, 2001; Roslow et al., 2000; Starr-McCluer, 2000).
It appears that the nature of sales fluctuations over time is multi-dimensional, and therefore, related to various concepts. In economic terms, besides production and consumption-related regulations, the demand-supply relationship is arguably the most profound dynamic behind sales outcomes. From a demand perspective, demographic and socio-economic factors (Zhang et al., 2016), trends related to health benefits and risks (Rimm et al., 1996), factors related to ethnicity, religion and customs (Chaturvedi et al., 2019), the temperature preferences for drinks in warm/hot conditions (Eccles et al., 2013; Gachons et al., 2016), the consumption of alcoholic beverages with food and event specific consumption during festive periods in the year (Kushnir and Cunningham, 2014), purchases for gift-giving occasions (Hirche and Bruwer, 2014), as well as seasonal tourism demand at destinations (Ridderstaat and Croes, 2020), could all be related to seasonal and non-seasonal sales fluctuation in a market.
From a supply perspective, the dependence of production on natural resources and related supply issues (Santos et al., 2020), general product availability and distribution (Thach and Olsen, 2006) and other marketing-mix factors (Lockshin and Knott, 2009).
In summary, the literature suggests that weather-related events and conditions can affect sales, and the sales response to weather variability appears to be product category specific. It is, therefore, advisable to investigate short-term sales variance at the product category-level. In addition, seasons and holiday periods are mentioned as external influencing market forces (Agnew and Thornes, 1995). Based on previous findings, we hypothesise that the sales of alcoholic beverages are associated with holiday periods and temperature changes over time. We also expect that the short-term (non-seasonal) sales response differs across different alcoholic beverages. Anecdotal evidence suggests that, for example, red wine is more preferred during colder periods of the year and white wine in hotter periods (Campbell, 2013).
3. Research methods
Temperature and holidays form the two independent variables tested for a relationship with the retail sales of different alcoholic beverages. Temperature is a common meteorological measure in related economic and retailing literature and, compared to other weather-related measures, is suggested to be the best predictor of sales variations (Agnew and Palutikof, 1999; Bujisic et al., 2017). Daily maximum temperatures are collected via the Global Historical Climatology Network-daily data set from the National Oceanic and Atmospheric Administration (Durre et al., 2010; Menne et al., 2012). Weather stations for each county are identified and the daily maximum temperatures (TMAX) for the periods studied are aggregated to weekly averages on county-level matching the exact range of weekdays from the sales data. Data of weekly store sales (volume) of alcoholic beverages from 37,346 retail stores for the years 2013–2015 is provided by Nielsen [1]. This sample offers enough observations for the SARIMAX modelling approach (Hyndman and Kostenko, 2007). Table 1 provides a summary of the sample stores by retail channel.
Over two-thirds of the stores belong to the drug and mass merchandise channels. The remaining third is represented by food, convenience and liquor stores. Alcoholic beverages are aggregated and categorised into beer, spirits and wine. Wine is sub-categorised into dry white, dry rosé, dry red, sparkling and other wines. All sub-categories comprise domestic and imported wines. Other wines include sweet dessert wine, vermouth, aperitifs, flavoured wine, sake, sangria, kosher table wine and non-alcoholic wine. The units for all products are standardised to volume in litres (L).
The overall sample consists of all counties whose average store sales of alcoholic beverages are greater than zero in every week during the studied period. State and county names are assigned to each store according to their given Federal Information Processing Standard county code. The weekly store data are further aggregated by county for each product category under investigation. The data are then merged with the available climate data resulting in 156 observations (weeks) for 651 counties and each of the 7 alcoholic beverage categories. This results in a total of 710,892 observations.
To identify any geographic relationship between sales and temperature, two sub-samples are analysed in addition to the total time-series of all counties included in this study. The first sub-sample represents the quartile of counties with the highest average temperature, and the second sub-sample is the quartile with the lowest average temperature for the studied period.
To provide more context, Table 2 presents the category shares for the total sample and respective sub-samples. As mentioned before, the time-series analysis comprised the sample of all counties (total), the 25% of all counties with the highest average temperature in the study period and the 25% with the lowest average temperature. As can be seen, approximately 70% of all alcoholic beverage off-trade retail sales (volume) are beer, 10% are liquor and 20% are in the wine category. There are minimal variations in these shares when comparing the sub-sampled counties. The share of liquor sales, for example, is slightly higher in the high and low temperature counties compared to the rest.
Time-series data require a careful investigation of any potential critical patterns before further analysis. The detection of possible seasonality in time-series data is important to draw conclusions on the proper methodology to use (Davey and Flores, 1993; Kapoor et al., 1981; Hanssens et al., 2002). In a first step, the time-series is visualised to gain a first impression of its composition (Figure 1).
The time-series plots in Figure 1 show regularly occurring sales peaks with different seasonal and non-seasonal variability across the categories. The significant sales shocks are assumed to be holiday-related. Figure 2 highlights the federal public holidays and the likely related sales shocks for the white wine category. A seasonal dummy variable (here for the week before and the week in which federal holidays occur) is used for the analysis (Chu and Zhang, 2003). The variable is based on the federal holiday list and dates from the US Office of Personnel Management (OPM, 2019). The visual examination of the data plots also reveals an underlying, near constant trend for each data series over time. All data series including temperature are constantly growing, except for dry rosé (decreasing trend). Further analysis needs to determine the nature of the trends to implement potential de-trending operations (Wu et al., 2007). Across the categories, the long-term trends of deterministic nature are eliminated by computing the least squares linear trend function (1) and dividing the original values by its trend estimates (2).
The variable
This operation will not just ensure the elimination of trends but also centres the data series, which results in the trend-stationarity of the time-series. The centred data will further facilitate the interpretation of positive and negative deviations from the base line. The constant term (mean = 0) in any further regression can be suppressed. After de-trending and centring the data, stationarity is examined using the Augmented Dickey-Fuller unit-root test (Elliott et al., 1996), Portmanteau (Q) test for white noise and the Phillips-Perron unit-root test (Phillips and Perron, 1988). The Dickey-Fuller and Phillips-Perron tests assess the null hypothesis that the time-series follows a unit root process. In all cases, the hypothesis can be significantly rejected. Importantly, the Portmanteau test examines possible autocorrelations in the residuals. Here the null hypothesis of no autocorrelation is rejected in all cases. This is likely explained by the exogenous changes (seasonality and weekly shocks), which are yet to be explained and incorporated in an adequate modelling approach. The test results are presented in Appendix Table A1.
When examining the autocorrelations (ACF) and partial autocorrelations (PACF) the pattern clearly indicates seasonality characteristics with, for example, the 1st and 52nd lag being outside the 95% confidence bands. This also confirms the visually appearing reoccurrence of similar sales and temperature patterns from season to season (year). In all categories’, data series ACF lags show a rapid decay after the first significant lag and PACF lags cut off after the first lag. This has implications regarding the order of a possible AR process and/or the moving-average process when modelling. Dynamic linear modelling is recommended to isolate seasonality and extreme weather events when forecasting sales of alcoholic beverages (Johnston and Harrison, 1980). In this regard, Kapoor et al. (1981) suggest a combined deterministic and stochastic model approach for time-series analysis of sales data. In this analysis, we follow a similar approach by eliminating the deterministic trend and further follow a stationary stochastic process accounting for seasonality effects (Arunraj et al., 2016). Therefore, a SARIMAX time-series modelling approach is considered, which exhibits a periodic seasonal component including independent variables. The statistical software version Stata/SE 13.1 is used.
The SARIMAX model is built on a linear regression Model (4) with two exogenous variables (holiday and temperature) and uses a multiplicative seasonal ARIMA process (5). The simplified model expression is ARIMA (p, d, q) * (P, D, Q)s. The non-seasonal AR lag polynomial
4. Results and discussion
The possible influence of weekly average maximum temperatures and public holidays on volume sales of different alcoholic beverage categories is analysed. Not only the general influence of temperature is of interest; in addition, this study compares possible sales variation in the 25% warmest and 25% coolest counties. Figure 3 illustrates the spatial distribution of all counties, which are categorised into the two sub-samples.
The counties with the lowest average temperature are highlighted mid-grey and are mainly located in the north of the US, whereas the warmest counties (dark grey) are primarily in the south. The counties highlighted in light grey are the remaining counties. The total sample of counties includes all highlighted counties.
As demonstrated in Table 3, the time-series analysis reveals a strong relationship between public holidays and the sales of all alcoholic beverage categories, regardless of the general climate of the region. Federal public holidays are generally a solid indicator of short-term increased sales. Specifically, sparkling wine and other wines (including sweet dessert wines) show the highest sensitivities. This is a satisfying outcome, as the holiday dummy variable is a reliable absorber of major short-term sales variances in the regression. This enhances a reliable isolation of the temperature variable and its influence on sales. The percentage change indicates how much more of each category is sold during the week of the public holiday or the week before. Figure 4 visualises the sales effects accordingly.
The general effect of public holidays on alcoholic beverage sales is positive. Sparkling and other wine show the largest effects. In the cooler regions, the effect for sparkling wine is considerably greater compared to the average. On the other hand, beer sales increase more in the warmer regions during and before public holidays. The percent sales effects can be translated into sales volume for the average store in the sample. During a week with public holiday(s), the average store, which actually sells the individual category would sell about 72 more litres of beer, 10 more litres of liquor, 9 more litres of white wine, 8 more litres of red wine, 5 more litres of sparkling wine, 4 more litres of other wine and 2 more litres of rosé. These are cautious estimates based on the sales fluctuation from the general base trend. A general positive sales trend over months and years may relativise such isolated marginal effect sizes.
Figure 2 also suggests that sales effects may differ by the type of public holiday. We apply the identified time-series models to test the effects of individual federal public holidays adjusted for seasonality and with day temperatures accounted for. The results in Table 4 below show that Easter, Thanksgiving, Christmas and New Year holidays are highly associated with sales boosts for all categories. The Easter holidays see increases in all categories, particularly in sparkling wine (+23.1%) and other wine (+18.2%). Beer sales consistently increase during the Memorial Day period (+15.4%). The tendency for other alcoholic beverages is positive, except for sparkling wine, which likely experiences slower sales in that time. Independence Day is related to sales increases of beer (+22.0%) and liquor (+23.7%). Labour Day weekends are known for parades, picnics and barbecues at the end of summer. Beer (+16.1%) and liquor (+10.5%) sales increase around Labour Day and so do the sales of rosé wine (+3.1%). The Thanksgiving holidays are traditionally very food-centric, and during this time particularly red (+26.3%), white (+17.3%), rosé (+19.4%) and other wine (+110.8%) sales increase significantly. The Christmas and New Year’s holidays see the most pronounced boost in sales, particularly in sparkling wine (+247.7%), as well as all other wine types, beer and liquor, which is likely linked to the holidays’ nature of culinary events, gift-giving occasions and festive celebrations.
Temperature is significantly associated with the sales of beer, liquor, red and white wine. With rising temperature, sales increase for beer, liquor and white wine and decrease for red wine. The negative sales sensitivity of red wine to temperature supports the anecdotal evidence that consumers may prefer red wine during colder periods of the year. However, during weeks of holidays, this effect could be neutralised or even reversed. In other words, even during seasonally warmer periods in the year, public holiday periods can have a short-term counter-effect seeing short boosts of red wine sales above the yearly average. This tendency is illustrated in Table 5.
Beer sales are positively associated with increasing temperatures regardless of geography and respective average temperatures in individual counties. When comparing the warmest with the coolest counties, liquor sales are not influenced by rising temperatures in the warmest regions, but they are overall and in the cooler counties. This makes sense when considering that rising temperatures in the already warmest counties may lose their effect. The considerable increase of liquor sales over summer could be explained by an increased consumption of chilled mixed drinks such as cocktails. Anecdotal evidence suggests that consumers associate liquor with relaxing and socializing, which are more dominant activities during the warmer summer months compared to winter (Dowdy, 2004).
Red wine sales drop with rising temperatures throughout the sampled counties. White wine sales are temperature sensitive overall and specifically in the cooler US counties. Similar to liquor, white wine sales are not significantly responsive to temperature changes in the warmest counties. Temperature does not seem to play an important role for rosé, sparkling and other wines. Figure 5 shows the sales effects of increasing temperature on the sales volume by the categories of alcoholic beverages.
Beer, liquor and white wine sales considerably increase with positive temperature changes from one week to another. In the total sample of US counties, an average temperature change of +10°C results in around +10.2% more beer sales, +4.6% more liquor sales and +1.4% more white wine sales per week. Red wine would see a −3.2% sales drop. The effect for beer is halved (+5.9%) in the cooler, northern counties. However, the positive white wine effect doubles (+3.0%) in these counties.
Again, the effects can be translated into volume sales effects for individual stores. With a +1°C increase in temperature the average store in the sample would sell 11.4 L more beer, 1 L more liquor and 0.3 L more white wine. The store would sell 0.6 L less red wine. In the warmer regions, the significant effect of liquor and white wine disappears, but the effect on beer and red wine is even stronger. The average store in the warmest counties of the US sells 53.3 L more beer and 2.8 L less red wine per +1°C rise in temperature. The percentage-change effects in the cooler regions (north) are stronger than those in the warmer regions (south). This confirms that the consumption of alcohol is more sensitive to weather changes in the north than in the south (Sun et al., 2009).
Finally, we apply our SARIMAX models to forecast sales for the out-of-sample year 2015. The models are based on 2013 and 2014 information to forecast the periodic sales during the unseen year 2015. Figure 6 presents the forecast results highlighting the predicted and actual sales deviations from the base trend of every beverage category in the total sample. This is a reasonable forecast result as the models generally cover major periodic sales and seasonal changes.
5. Conclusion
The perspective of how retail sales for various competing categories change over time is important for marketers of alcoholic beverages. Understanding sales variation as a result of holiday periods, seasonality in climate and weather, geography and the beverage categories, has potential strategic and managerial implications for all stakeholders along the supply chain of consumer packaged goods, including alcoholic beverages. This study examined the possible relationships and contributions to forecasting accuracy with SARIMAX time-series models. The analysis revealed that off-trade retail sales of different competing alcoholic beverages are temperature sensitive and vary during holiday periods. Specifically, the sales of beer, liquor, red and white wine are temperature sensitive throughout the year. This sales sensitivity may also differ in regions that are generally very warm or very cool. Public holidays represent a constant influencing factor for all investigated beverage categories. We also find that sales of various beverage types differ by individual holiday periods. We can further conclude that understanding the effects of temperature and holiday data can considerably inform and improve forecasting performance.
6. Implications, limitations and future research
The findings of this study add to existing literature and contribute to the knowledge in this field. The findings also have significant practical implications. Time-series modelling forms the basis for forecasting. In applying forecasting techniques marketers could, for example, better plan for short and long-term sales promotions according to forecasted changes in sales based on weather across different product categories at different times of the year. The same applies to inventory management and the distribution of products. By integrating meteorological, geographical and holiday/event data, distributors and retailers could forecast the expected demand for their products given the location, seasonal temperature changes and holiday periods. Also, researchers may consider those fluctuations overtime when conducting studies. Considering holiday seasons and weather events when, for example, surveying consumers or retail stores, can add value to research integrity and outcomes.
In this study, we focussed on the possible influence of day temperature and holiday periods on retail sales of alcoholic beverage categories within a universe of retail stores in the sampled market. From theory, we know that there are many more factors influencing the sales in retail stores, as outlined in the literature review. This study should be considered at the observational and exploratory stage of the scientific discovery forming the basis for generating formal hypotheses that can be tested in future research. We, therefore, invite researchers and analysts to develop more sophisticated forecasting approaches, which can be realistically applied by retailers and marketers of alcoholic beverages. Future studies may consider analysing geocoded panel or household data for causal inference.
Future research may better decompose the series into seasonal and non-seasonal components and incorporate additional variables in time-series or demand models, and thereby increasing the predictive power. Such additional variables could reflect out-of-stock occasions, temporary price reductions, feature and display occasions, geographic and temporal price sensitivities, seasonal tourism demand linked to holidays and other spatially dependent variables affecting supply and demand. Time-series-cross-section data require special attention when establishing models. For data similar to that used in this study, spatial modelling approaches may be advisable (Beck, 2001). Future research may consider these aspects with further spatial analyses of retail sales.
Studies may expand this research to other product categories and their sensitivity to external factors similar to the ones used in this study. Forecasting with hierarchical or grouped time-series models in a big data environment is an option that allows state-space and space-time-related analysis (Hyndman and Athanasopoulos, 2018; Murray et al., 2018). Combining SARIMAX time-series analysis with machine-learning techniques, such as neural networks, can also be considered (Aburto and Weber, 2007). This would allow for new methodological discoveries, more accurate forecasting predictions and ultimately could lead to further practical applications and new knowledge.
Figures
Number of stores by channel
| Channel | Stores | (%) |
|---|---|---|
| Drug | 12,778 | 34.2 |
| Mass merchandise | 12,671 | 33.9 |
| Food | 10,599 | 28.4 |
| Convenience | 1,079 | 2.9 |
| Liquor | 219 | 0.6 |
| Total | 37,346 | 100.0 |
Volume share by category and county sample
| Counties | Total | 25% high temperature | 25% low temperature | 50% medium temperature |
|---|---|---|---|---|
| Category | ||||
| Beer | 70.4 | 70.1 | 69.6 | 71.1 |
| Liquor | 9.7 | 10.4 | 11.0 | 8.3 |
| Red wine | 7.6 | 7.6 | 7.3 | 7.8 |
| White wine | 7.6 | 7.4 | 7.4 | 7.8 |
| Rosé wine | 1.4 | 1.4 | 1.5 | 1.4 |
| Sparkling wine | 0.9 | 1.0 | 0.8 | 0.9 |
| Other wine | 2.4 | 2.1 | 2.5 | 2.6 |
| Total | 100.0 | 100.0 | 100.0 | 100.0 |
Relationship of holidays with sales by category
| Counties | Total | 25% high temp. | 25% low temp. | |||||||||
|---|---|---|---|---|---|---|---|---|---|---|---|---|
| Seasonality | With seasonal autoregressive term | With seasonal differencing | With seasonal autoregressive term | With seasonal differencing | With seasonal autoregressive term | With seasonal differencing | ||||||
| Category | Coeff. | %-change | Coeff. | %-change | Coeff. | %-change | Coeff. | %-change | Coeff. | %-change | Coeff. | %-change |
| Beer | 0.062*** | 6.374 | 0.074*** | 7.607 | 0.091*** | 9.527 | 0.109*** | 11.411 | 0.058*** | 5.824 | 0.051*** | 5.121 |
| Liquor | 0.047*** | 4.888 | 0.040*** | 4.160 | 0.054*** | 5.564 | 0.052*** | 5.358 | 0.061*** | 6.319 | 0.045** | 4.662 |
| Red wine | 0.037*** | 3.910 | 0.036*** | 3.804 | 0.052*** | 5.639 | 0.049*** | 5.313 | 0.049*** | 5.049 | 0.048*** | 4.946 |
| White wine | 0.042*** | 4.390 | 0.046*** | 4.808 | 0.069*** | 7.207 | 0.072*** | 7.520 | 0.049*** | 5.038 | 0.061*** | 6.272 |
| Rosé wine | 0.048*** | 4.594 | 0.048*** | 4.594 | 0.039*** | 3.743 | 0.038*** | 3.647 | 0.064*** | 6.025 | 0.070*** | 6.589 |
| Sparkling | 0.205*** | 22.914 | 0.212*** | 23.697 | 0.197*** | 21.999 | 0.208*** | 23.228 | 0.258*** | 28.043 | 0.250*** | 27.173 |
| Other wine | 0.138*** | 17.838 | 0.137*** | 17.708 | 0.137*** | 17.635 | 0.136*** | 17.507 | 0.150*** | 18.525 | 0.151*** | 18.648 |
Significance levels:
1% significance level.
5% significance level.
10% significance level. %-change is the percent increase from the base trend line based on the coefficient for 1 unit change of the independent variable; the marginal effects are obtained by rescaling the centred data using the ratios of the mean absolute deviations of both, the original and transformed data series
Effects of individual holidays on sales from the base trend line (%)
| Category | Beer | Liquor | Red wine | White wine | Rosé wine | Sparkl. wine | Other wine |
|---|---|---|---|---|---|---|---|
| Public holiday | |||||||
| Martin Luther King Day | −12.2 | −12.4 | −3.1 | −6.4 | −5.1 | −41.0 | −26.1 |
| Washington’s Birthday | −11.3 | −8.5 | +4.1 | −1.0 | +3.7 | −2.2 | −7.0 |
| Easter Holidays | +4.8*** | +2.5** | +4.2*** | +5.0*** | +4.8*** | +23.1*** | +18.2*** |
| Memorial Day | +15.4** | +16.3 | +1.5 | +1.3 | +3.9 | −7.7 | +11.7 |
| Independence Day | +22.0*** | +23.7*** | −2.0 | +3.0 | +4.6 | −6.5 | +16.3 |
| Labour Day | +16.1*** | +10.5*** | +2.0 | +1.9 | +3.1** | +6.2 | +12.3 |
| Columbus Day | −6.4 | −10.3 | +1.6 | −4.5 | −0.9 | −25.9 | +0.8 |
| Veterans Day | −8.4 | −8.2 | +6.7 | −1.7 | +0.1 | −22.0 | +12.1 |
| Thanksgiving Day | +7.9** | +5.7 | +26.3*** | +17.3*** | +19.4*** | +41.2 | +110.8*** |
| Christmas and New Year Holidays | +15.9*** | +24.8*** | +24.1*** | +13.4*** | +22.6*** | +247.7*** | +136.3*** |
Significance levels:
1% significance level.
5% significance level.
10% significance level
Relationship of temperature (°C) with sales by category
| Counties | Total | 25% high temp. | 25% low temp. | |||||||||
|---|---|---|---|---|---|---|---|---|---|---|---|---|
| Seasonal adjustment | With seasonal autoregressive term | With seasonal differencing | With seasonal autoregressive term | With seasonal differencing | With seasonal autoregressive term | With seasonal differencing | ||||||
| Category | Coeff. | %-change | Coeff. | %-change | Coeff. | %-change | Coeff. | %-change | Coeff. | %-change | Coeff. | %-change |
| Beer | 0.185*** | 1.016 | 0.113*** | 0.621 | 0.270*** | 1.103 | 0.183** | 0.748 | 0.073*** | 0.588 | 0.072*** | 0.580 |
| Liquor | 0.083*** | 0.461 | 0.057* | 0.317 | 0.038 | 0.153 | 0.048 | 0.193 | 0.072*** | 0.598 | 0.065*** | 0.540 |
| Red wine | −0.057*** | −0.322 | −0.033* | −0.186 | −0.131** | −0.554 | −0.096** | −0.406 | −0.043*** | −0.355 | −0.033* | −0.273 |
| White wine | 0.025** | 0.140 | 0.029* | 0.162 | 0.004 | 0.016 | 0.028 | 0.114 | 0.036*** | 0.297 | −0.005 | −0.041 |
| Rosé wine | −0.020 | −0.102 | −0.018 | −0.092 | −0.016 | −0.060 | −0.005 | −0.019 | −0.011 | −0.083 | −0.019 | −0.144 |
| Sparkling | −0.190 | −1.135 | −0.042 | −0.251 | −0.241 | −1.050 | −0.060 | −0.261 | −0.104 | −0.907 | −0.055 | −0.480 |
| Other wine | −0.097 | −0.670 | −0.086 | −0.594 | −0.158 | −0.794 | −0.119 | −0.598 | −0.026 | −0.258 | −0.031 | −0.307 |
Significance levels:
1% significance level.
5% significance level.
10% significance level. %-change is the percent increase from the base trend line based on the coefficient for 1 unit change of the independent variable; the marginal effects are obtained by rescaling the centred data using the ratios of the mean absolute deviations of both, the original and transformed data series
Stationarity tests
| Augmented Dickey-Fuller2 | Q-test | Phillips-Perron3 | |||||
|---|---|---|---|---|---|---|---|
| All counties | Z(t) | L1 coef. | Z(rho) | Z(t) | L1 coef. | ||
| Beer | −6.098*** | −0.389*** | 619.540*** | −57.592*** | −5.990*** | 0.611*** | |
| Liquor | −5.018*** | −0.289*** | 433.051*** | −43.553*** | −4.954*** | 0.711*** | |
| Red wine | −3.835*** | −0.214*** | 804.276*** | −28.012*** | −3.488*** | 0.786*** | |
| White wine | −7.848*** | −0.611*** | 77.725*** | −95.865*** | −7.882*** | 0.389*** | |
| Rosé wine | −7.251*** | −0.569*** | 80.792*** | −91.456*** | −7.354*** | 0.430*** | |
| Sparkling wine | −9.091*** | −0.573*** | 59.384** | −104.487*** | −9.277*** | 0.427*** | |
| Other wine | −7.355*** | −0.478*** | 145.997*** | −81.577*** | −7.560*** | 0.522*** | |
| Temperature | −2.031** | −0.046** | 2,456.908*** | −7.977** | −2.126** | 0.954*** | |
| High average temperature quartile of counties | |||||||
| Beer | −8.145*** | −0.599*** | 250.956*** | −93.678*** | −8.166*** | 0.401*** | |
| Liquor | −7.109*** | −0.516*** | 155.683*** | −82.227*** | −7.181*** | 0.484*** | |
| Red wine | −3.568*** | −0.179*** | 1,050.228*** | −22.913*** | −3.215*** | 0.821*** | |
| White wine | −7.353*** | −0.548*** | 130.444*** | −90.276*** | −7.520*** | 0.452*** | |
| Rosé wine | −7.051*** | −0.536*** | 136.919*** | −88.320*** | −7.227*** | 0.464*** | |
| Sparkling wine | −9.119*** | −0.581*** | 62.631** | −104.858*** | −9.297*** | 0.419*** | |
| Other wine | −7.684*** | −0.507*** | 130.125*** | −85.977*** | −7.867*** | 0.493*** | |
| Temperature | −2.277** | −0.057** | 2,360.511*** | −9.237** | −2.318** | 0.943*** | |
| Low average temperature quartile of counties | |||||||
| Beer | −4.775*** | −0.260*** | 841.377*** | −34.741*** | −4.480*** | 0.740*** | |
| Liquor | −3.885*** | −0.182*** | 771.364*** | −25.542*** | −3.715*** | 0.818*** | |
| Red wine | −4.385*** | −0.267*** | 513.170*** | −37.109*** | −4.134*** | 0.733*** | |
| White wine | −5.874*** | −0.393*** | 179.345*** | −60.089*** | −5.842*** | 0.607*** | |
| Rosé wine | −6.626*** | −0.503*** | 144.176*** | −79.759*** | −6.692*** | 0.497*** | |
| Sparkling wine | −8.782*** | −0.545*** | 61.721** | −101.082*** | −9.011*** | 0.455*** | |
| Other wine | −7.103*** | −4.463*** | 151.850*** | −78.720*** | −7.308*** | 0.537*** | |
| Temperature | −2.102** | −0.050** | 2,419.544*** | −8.095** | −2.136** | 0.950*** | |
Significance levels:
1% significance level.
5% significance level.
10% significance level. 2 3 α = 0, δ = 0, no constant and no trend term included, as data is de-trended and centred (averages = 0). Process under null hypothesis is random walk without drift, 3Uses the number of Newey–West lags for the calculation of the standard error
SARIMAX models (example for beer all counties)
| (p, d, q) x (P,D,Q)S |
(0,0,0)x (1,0,0)52 |
(1,0,0)x (1,0,0)52 |
(1,0,1)x (1,0,0)52 |
(0,0,0)x (1,0,1)52 |
(1,0,0)x (1,0,1)52 |
(1,0,1)x (1,0,1)52 |
(0,0,0)x (0,1,0)52 |
| Holiday | 0.062*** | 0.063*** | 0.073*** | 0.063*** | 0.064*** | 0.074*** | 0.074*** |
| Temperature | 0.185*** | 0.182*** | 0.137*** | 0.186*** | 0.183*** | 0.137*** | 0.113*** |
| L1.ar | 0.061 | 0.911*** | 0.060 | 0.915*** | |||
| SL1.ar | 0.904*** | 0.900*** | 0.906*** | 0.925*** | 0.921*** | 0.929*** | |
| L1.ma | −0.801*** | −0.805*** | |||||
| SL1.ma | −0.116 | −0.115 | −0.132 | ||||
| AIC | −406.104 | −404.586 | −407.928 | −404.673 | −403.147 | −406.645 | −329.227 |
| BIC | −393.905 | −389.337 | −389.629 | −389.424 | −384.848 | −385.296 | −321.294 |
| (p, d, q) x (P,D,Q)S |
(1,0,0)x (0,1,0)52 |
(1,0,1)x (0,1,0)52 |
(0,0,0)x (1,1,0)52 |
(1,0,0)x (1,1,0)52 |
(1,0,1)x (1,1,0)52 |
||
| Holiday | 0.072*** | 0.076*** | 0.075*** | 0.073*** | 0.077*** | ||
| Temperature | 0.118*** | 0.112*** | 0.117*** | 0.122*** | 0.116*** | ||
| L1.ar | −0.109 | −0.962*** | −0.107 | −0.963*** | |||
| SL1.ar | −0.099 | −0.092 | −0.099 | ||||
| L1.ma | 0.918*** | 0.919*** | |||||
| SL1.ma | |||||||
| AIC | −328.411 | −327.684 | −327.723 | −326.835 | −326.174 | ||
| BIC | −317.833 | −314.462 | −317.145 | −313.613 | −310.307 |
Significance levels:
1% significance level.
5% significance level.
10% significance level
SARIMAX models (example for liquor all counties)
| (p, d, q) x (P,D,Q)S |
(0,0,0)x (1,0,0)52 |
(1,0,0)x (1,0,0)52 |
(1,0,1)x (1,0,0)52 |
(0,0,0)x (1,0,1)52 |
(1,0,0)x (1,0,1)52 |
(1,0,1)x (1,0,1)52 |
(0,0,0)x (0,1,0)52 |
| Holiday | 0.0380*** | 0.047*** | 0.050*** | 0.038*** | 0.047*** | 0.050*** | 0.040*** |
| Temperature | 0.093*** | 0.083*** | 0.060* | 0.093*** | 0.082*** | 0.060* | 0.057* |
| L1.ar | 0.373*** | 0.763*** | 0.373*** | 0.766*** | |||
| SL1.ar | 0.958*** | 0.931*** | 0.934*** | 0.955*** | 0.929*** | 0.938*** | |
| L1.ma | −0.456*** | −0.460*** | |||||
| SL1.ma | 0.035 | 0.014 | −0.032 | ||||
| AIC | −445.714 | −460.554 | −464.159 | −443.770 | −458.562 | −462.199 | −382.799 |
| BIC | −433.514 | −445.304 | −445.860 | −428.520 | −440.263 | −440.850 | −374.865 |
| (p, d, q) x (P,D,Q)S |
(1,0,0)x (0,1,0)52 |
(1,0,1)x (0,1,0)52 |
(0,0,0)x (1,1,0)52 |
(1,0,0)x (1,1,0)52 |
(1,0,1)x (1,1,0)52 |
||
| Holiday | 0.042*** | 0.041*** | 0.040*** | 0.042*** | 0.041*** | ||
| Temperature | 0.056* | 0.041 | 0.056* | 0.056* | 0.041 | ||
| L1.ar | 0.078 | 0.995*** | 0.076 | 0.995*** | |||
| SL1.ar | 0.035 | 0.021 | −0.016 | ||||
| L1.ma | −0.973*** | −0.972*** | |||||
| SL1.ma | |||||||
| AIC | −381.403 | −382.122 | −380.860 | −379.426 | −380.135 | ||
| BIC | −370.826 | −368.900 | −370.283 | −366.204 | −364.269 |
Significance levels:
1% significance level.
5% significance level.
10% significance level
SARIMAX models (example for red wine all counties)
| (p, d, q) x (P,D,Q)S |
(0,0,0)x (1,0,0)52 |
(1,0,0)x (1,0,0)52 |
(1,0,1)x (1,0,0)52 |
(0,0,0)x (1,0,1)52 |
(1,0,0)x (1,0,1)52 |
(1,0,1)x (1,0,1)52 |
(0,0,0)x (0,1,0)52 |
| Holiday | 0.038*** | 0.030*** | 0.037*** | 0.038*** | 0.029*** | 0.036*** | 0.035*** |
| Temperature | −0.048*** | −0.049** | −0.057*** | −0.048*** | −0.048** | −0.056*** | 0.001 |
| L1.ar | 0.385*** | 0.923*** | 0.386*** | 0.930*** | |||
| SL1.ar | 0.963*** | 0.940*** | 0.948*** | 0.962*** | 0.945*** | 0.963*** | |
| L1.ma | −0.710*** | −0.723*** | |||||
| SL1.ma | 0.005 | −0.044 | −0.142 | ||||
| AIC | −535.394 | −552.531 | −568.537 | −533.396 | −550.613 | −567.340 | −452.537 |
| BIC | −523.195 | −537.281 | −550.237 | −518.147 | −532.314 | −545.991 | −444.603 |
| (p, d, q) x (P,D,Q)S |
(1,0,0)x (0,1,0)52 |
(1,0,1)x (0,1,0)52 |
(0,0,0)x (1,1,0)52 |
(1,0,0)x (1,1,0)52 |
(1,0,1)x (1,1,0)52 |
||
| Holiday | 0.034*** | 0.036*** | 0.035*** | 0.033*** | 0.035*** | ||
| Temperature | −0.001 | −0.033* | 0.005 | 0.003 | −0.029 | ||
| L1.ar | 0.057 | 0.998*** | 0.062 | 0.999*** | |||
| SL1.ar | −0.109 | −0.113 | −0.261*** | ||||
| L1.ma | −0.965*** | −0.975*** | |||||
| SL1.ma | |||||||
| AIC | −450.835 | −464.608 | −451.107 | −449.450 | −465.949 | ||
| BIC | −440.257 | −451.386 | −440.530 | −436.228 | −450.083 |
Significance levels:
1% significance level.
5% significance level.
10% significance level
SARIMAX models (example for white wine all counties)
| (p, d, q) x (P,D,Q)S |
(0,0,0)x (1,0,0)52 |
(1,0,0)x (1,0,0)52 |
(1,0,1)x (1,0,0)52 |
(0,0,0)x (1,0,1)52 |
(1,0,0)x (1,0,1)52 |
(1,0,1)x (1,0,1)52 |
(0,0,0)x (0,1,0)52 |
| Holiday | 0.042*** | 0.039*** | 0.046*** | 0.042*** | 0.039*** | 0.045*** | 0.046*** |
| Temperature | 0.025** | 0.023 | 0.014 | 0.024* | 0.023 | 0.014 | 0.029* |
| L1.ar | 0.160** | 0.995*** | 0.160** | 0.995*** | |||
| SL1.ar | 0.930*** | 0.919 | 0.935*** | 0.923 | 0.918*** | 0.948*** | |
| L1.ma | −0.970*** | −0.967*** | |||||
| SL1.ma | 0.045 | 0.005 | −0.097 | ||||
| AIC | −617.087 | −618.590 | −624.773 | −615.167 | −616.591 | −623.157 | −476.494 |
| BIC | −604.887 | −603.340 | −606.473 | −599.917 | −598.292 | −601.808 | −468.561 |
| (p, d, q) x (P,D,Q)S |
(1,0,0)x (0,1,0)52 |
(1,0,1)x (0,1,0)52 |
(0,0,0)x (1,1,0)52 |
(1,0,0)x (1,1,0)52 |
(1,0,1)x (1,1,0)52 |
||
| Holiday | 0.047*** | 0.047*** | 0.046*** | 0.046*** | 0.047*** | ||
| Temperature | 0.029* | 0.028 | 0.029* | 0.030* | 0.028 | ||
| L1.ar | −0.030 | −0.970*** | −0.027 | −0.972*** | |||
| SL1.ar | −0.029 | −0.022 | −0.023 | ||||
| L1.ma | 0.940*** | 0.944*** | |||||
| SL1.ma | |||||||
| AIC | −474.575 | −473.945 | −474.536 | −472.598 | −471.971 | ||
| BIC | −463.997 | −460.723 | −463.959 | −459.376 | −456.104 |
Notes: Significance levels:
1% significance level.
5% significance level.
10% significance level
SARIMAX models (example for rosé wine all counties)
| (p, d, q) x (P,D,Q)S |
(0,0,0)x (1,0,0)52 |
(1,0,0)x (1,0,0)52 |
(1,0,1)x (1,0,0)52 |
(0,0,0)x (1,0,1)52 |
(1,0,0)x (1,0,1)52 |
(1,0,1)x (1,0,1)52 |
(0,0,0)x (0,1,0)52 |
| Holiday | 0.047*** | 0.043*** | 0.048*** | 0.047*** | 0.043*** | 0.047*** | 0.049*** |
| Temperature | −0.017 | −0.012 | −0.020 | −0.018 | −0.012 | −0.020 | −0.002 |
| L1.ar | 0.335*** | 0.896*** | 0.337*** | 0.900*** | |||
| SL1.ar | 0.946*** | 0.941*** | 0.955*** | 0.946*** | 0.946*** | 0.962*** | |
| L1.ma | −0.676*** | −0.680*** | |||||
| SL1.ma | 0.003 | −0.043 | −0.075 | ||||
| AIC | −554.636 | −570.662 | −582.722 | −552.637 | −568.740 | −580.967 | −445.128 |
| BIC | −542.437 | −555.412 | −564.423 | −537.388 | −550.441 | −559.618 | −437.194 |
| (p, d, q) x (P,D,Q)S |
(1,0,0)x (0,1,0)52 |
(1,0,1)x (0,1,0)52 |
(0,0,0)x (1,1,0)52 |
(1,0,0)x (1,1,0)52 |
(1,0,1)x (1,1,0)52 |
||
| Holiday | 0.044*** | 0.048*** | 0.049*** | 0.043*** | 0.047*** | ||
| Temperature | −0.002 | −0.018 | −0.001 | 0.001 | −0.019 | ||
| L1.ar | 0.291*** | 0.945*** | 0.298*** | 0.954*** | |||
| SL1.ar | −0.047 | −0.096 | −0.150 | ||||
| L1.ma | −0.745*** | −0.758*** | |||||
| SL1.ma | |||||||
| AIC | −451.742 | −469.455 | −443.238 | −450.191 | −468.558 | ||
| BIC | −441.165 | −456.233 | −432.660 | −436.969 | −452.691 |
Notes: Significance levels:
1% significance level.
5% significance level.
10% significance level
SARIMAX models (example for sparkling wine all counties)
| (p, d, q) x (P,D,Q)S |
(0,0,0)x (1,0,0)52 |
(1,0,0)x (1,0,0)52 |
(1,0,1)x (1,0,0)52 |
(0,0,0)x (1,0,1)52 |
(1,0,0)x (1,0,1)52 |
(1,0,1)x (1,0,1)52 |
(0,0,0)x (0,1,0)52 |
| Holiday | 0.205*** | 0.203*** | 0.203*** | 0.206*** | 0.204*** | 0.204*** | 0.199*** |
| Temperature | −0.190 | −0.187 | −0.187 | −0.175 | −0.167 | −0.167 | −0.063 |
| L1.ar | 0.050 | 0.012 | 0.060 | 0.019 | |||
| SL1.ar | 0.967*** | 0.964*** | 0.964*** | 0.961*** | 0.956*** | 0.956*** | |
| L1.ma | 0.038 | 0.041 | |||||
| SL1.ma | 12.693 | 10.258 | 10.249 | ||||
| AIC | 29.892 | 31.614 | 33.611 | 31.654 | 33.265 | 35.261 | −73.365 |
| BIC | 42.091 | 46.863 | 51.910 | 46.903 | 51.564 | 56.610 | −65.432 |
| (p, d, q) x (P,D,Q)S |
(1,0,0)x (0,1,0)52 |
(1,0,1)x (0,1,0)52 |
(0,0,0)x (1,1,0)52 |
(1,0,0)x (1,1,0)52 |
(1,0,1)x (1,1,0)52 |
||
| Holiday | 0.209*** | 0.212*** | 0.199*** | 0.209*** | 0.212*** | ||
| Temperature | −0.086 | −0.042 | −0.048 | −0.079 | −0.043 | ||
| L1.ar | −0.223*** | 0.395 | −0.221*** | 0.395 | |||
| SL1.ar | 0.054 | 0.028 | −0.002 | ||||
| L1.ma | −0.715** | −0.716** | |||||
| SL1.ma | |||||||
| AIC | −76.570 | −81.095 | −71.498 | −74.601 | −79.096 | ||
| BIC | −65.992 | −67.874 | −60.920 | −61.379 | −63.229 |
Notes: Significance levels:
1% significance level.
5% significance level.
10% significance level
SARIMAX models (example for other wine all counties)
| (p, d, q) x (P,D,Q)S |
(0,0,0)x (1,0,0)52 |
(1,0,0)x (1,0,0)52 |
(1,0,1)x (1,0,0)52 |
(0,0,0)x (1,0,1)52 |
(1,0,0)x (1,0,1)52 |
(1,0,1)x (1,0,1)52 |
(0,0,0)x (0,1,0)52 |
| Holiday | 0.140*** | 0.124*** | 0.138*** | 0.142*** | 0.127*** | 0.141*** | 0.145*** |
| Temperature | −0.203** | −0.111 | −0.097 | −0.203** | −0.098 | −0.074 | −0.146 |
| L1.ar | 0.567*** | 0.948*** | 0.567*** | 0.947*** | |||
| SL1.ar | 0.935*** | 0.934*** | 0.958*** | 0.917*** | 0.913*** | 0.937*** | |
| L1.ma | −0.658*** | −0.665*** | |||||
| SL1.ma | 0.154 | 0.170 | 0.257 | ||||
| AIC | −101.278 | −154.136 | −164.833 | −100.328 | −153.374 | −165.201 | −135.965 |
| BIC | −89.078 | −138.886 | −146.534 | −85.079 | −135.075 | −143.852 | −128.0.32 |
| (p, d, q) x (P,D,Q)S |
(1,0,0)x (0,1,0)52 |
(1,0,1)x (0,1,0)52 |
(0,0,0)x (1,1,0)52 |
(1,0,0)x (1,1,0)52 |
(1,0,1)x (1,1,0)52 |
||
| Holiday | 0.115*** | 0.137*** | 0.147*** | 0.118*** | 0.139*** | ||
| Temperature | −0.067 | −0.086 | −0.140 | −0.055 | −0.059 | ||
| L1.ar | 0.575*** | 0.990*** | 0.571*** | 0.988*** | |||
| SL1.ar | 0.120 | 0.108 | 0.206 | ||||
| L1.ma | −0.779*** | −0.786*** | |||||
| SL1.ma | |||||||
| AIC | −169.362 | −194.848 | −134.710 | −167.746 | −194.579 | ||
| BIC | −158.784 | −181.626 | −124.132 | −154.724 | −178.713 |
Notes: Significance levels:
1% significance level.
5% significance level.
10% significance level
Note
© This study is based on data from The Nielsen Company (US), LLC and marketing databases provided by the Kilts Centre for Marketing Data Centre at The University of Chicago Booth School of Business. The conclusions drawn from the Nielsen data are those of the researchers and do not reflect the views of Nielsen. Nielsen is not responsible for, had no role in and was not involved in analysing and preparing the results reported herein.
References
Aburto, L. and Weber, R. (2007), “Improved supply chain management based on hybrid demand forecasts”, Applied Soft Computing Journal, Vol. 7 No. 1, pp. 136-144.
Agnew, M.D. and Palutikof, J.P. (1999), “The impacts of climate on retailing in the UK with particular reference to the anomalously hot summer of 1995”, International Journal of Climatology, Vol. 19 No. 13, pp. 1493-1507.
Agnew, M.D. and Thornes, J.E. (1995), “The weather sensitivity of the UK food retail and distribution industry”, Meteorological Applications, Vol. 2 No. 2, pp. 137-147.
Arunraj, N.S. and Ahrens, D. (2016), “Estimation of non-catastrophic weather impacts for retail industry”, International Journal of Retail and Distribution Management, Vol. 44 No. 7, pp. 731-753.
Arunraj, N.S., Ahrens, D. and Fernandes, M. (2016), “Application of SARIMAX model to forecast daily sales in food retail industry”, International Journal of Operations Research and Information Systems (IJORIS), Vol. 7 No. 2, pp. 1-21.
Bahng, Y. and Kincade, D.H. (2012), “The relationship between temperature and sales: sales data analysis of a retailer of branded women’s business wear”, International Journal of Retail and Distribution Management, Vol. 40 No. 6, pp. 410-426.
Beck, N. (2001), “Time-series-cross-section data: what have we learned in the past few years?”, Annual Review of Political Science, Vol. 4 No. 1, pp. 271-293.
Brooke, Z. (2017), “Marketers are using the weather to predict buyer behavior”, available at: www.ama.org/publications/MarketingNews/Pages/how-marketers-are-using-the-weather-to-predict-buyer-behavior.aspx (accessed 26 June 2018).
Bruno, P., Melnyk, V. and Völckner, F. (2017), “Temperature and emotions: effects of physical temperature on responses to emotional advertising”, International Journal of Research in Marketing, Vol. 34 No. 1, pp. 302-320.
Bujisic, M., Bogicevic, V. and Parsa, H.G. (2017), “The effect of weather factors on restaurant sales”, Journal of Foodservice Business Research, Vol. 20 No. 3, pp. 350-370.
Campbell, B. (2013), “Wine on ice”, available at: www.therealreview.com/2013/12/01/wine-on-ice-2/ (accessed 12 June 2018).
Chaturvedi, H.K., Bajpai, R.C. and Tiwari, P. (2019), “Association of religion and cultural tradition with alcohol use among some tribal communities of Arunachal Pradesh, India”, Journal of Ethnicity in Substance Abuse, Vol. 18 No. 2, pp. 296-308.
Chu, C.W. and Zhang, G.P. (2003), “A comparative study of linear and nonlinear models for aggregate retail sales forecasting”, International Journal of Production Economics, Vol. 86 No. 3, pp. 217-231.
Curtis, J. (2003), “The weather effect”, Marketing, Haymarket Business Publications Ltd., London, pp. 22-23.
Davey, A.M. and Flores, B.E. (1993), “Identification of seasonality in time series: a note”, Mathematical and Computer Modelling, Vol. 18 No. 6, pp. 73-81.
Dowdy, C. (2004), “Competition hots up for summer”, in Marketing: Haymarket Media Group Ltd, pp. 18-19.
Durre, I., Menne, M.J., Gleason, B.E., Houston, T.G. and Vose, R.S. (2010), “Comprehensive automated quality assurance of daily surface observations”, Journal of Applied Meteorology and Climatology, Vol. 49 No. 8, pp. 1615-1633.
Eccles, R., Du-Plessis, L., Dommels, Y. and Wilkinson, J.E. (2013), “Cold pleasure. Why we like ice drinks, ice-lollies and ice cream”, Appetite, Vol. 71, pp. 357-360.
Elliott, G., Rothenberg, T.J. and Stock, J.H. (1996), “Efficient tests for an autoregressive unit root”, Econometrica, Vol. 64 No. 4, pp. 813-836.
Euromonitor (2017), Report: Statistics Alcoholic Drinks US, January 2017 ed, Euromonitor International.
Gachons, C.P., Des Avrillier, J., Gleason, M., Algarra, L., Zhang, S., Mura, E., Nagai, H. and Breslin, P.A.S. (2016), “Oral cooling and carbonation increase the perception of drinking and thirst quenching in thirsty adults”, PLoS ONE, Vol. 11 No. 9, pp. 1-12.
Govind, R., Garg, N. and Sun, W. (2014), “Geographically varying effects of weather on tobacco consumption: implications for health marketing initiatives”, Health Marketing Quarterly, Vol. 31 No. 1, pp. 46-64.
Hanssens, D.M., Parsons, L.J. and Schultz, R.L. (2002), Market Response Models Econometric and Time Series Analysis, Kluwer Academic Publishers, New York, NY.
Harrison, K. (1992), “Whether the weather be good”, Super Marketing, Vol. 13, pp. 15-17.
Hirche, M. and Bruwer, J. (2014), “Buying a product for an anticipated consumption situation – observation of high and low involved wine buyers in a retail store”, International Journal of Wine Business Research, Vol. 26 No. 4, pp. 295-318.
Hyndman, R.J. and Athanasopoulos, G. (2018), Forecasting: Principles and Practice, 2nd ed., OText, Melbourne, available at: www.otexts.org/book/fpp (accessed 13 January 2020).
Hyndman, R.J. and Kostenko, A.V. (2007), “Minimum sample size requirements for seasonal forecasting models”, Foresight, Vol. 6 No. Spring, pp. 12-15.
Johnston, F.R. and Harrison, P.J. (1980), “An application of forecasting in the alcoholic drinks industry”, Journal of the Operational Research Society, Vol. 31 No. 8, pp. 699-709.
Kapoor, S.G., Madhok, P. and Wu, S.M. (1981), “Modeling and forecasting sales data by time series analysis”, Journal of Marketing Research, Vol. 18 No. 1, pp. 94-100.
Kushnir, V. and Cunningham, J.A. (2014), “Event-specific drinking in the general population”, Journal of Studies on Alcohol and Drugs, Vol. 75 No. 6, pp. 968-972.
Lazo, J.K., Lawson, M., Larsen, P.H. and Waldman, D.M. (2011), “U.S. economic sensitivity to weather variability”, Bulletin of the American Meteorological Society, Vol. 92 No. 6, pp. 709-720.
Li, C., Luo, X., Zhang, C. and Wang, X. (2017), “Sunny, rainy, and cloudy with a chance of mobile promotion effectiveness”, Marketing Science, Vol. 36 No. 5, pp. 762-779.
Lockshin, L. and Knott, D. (2009), “Boozing or branding? Measuring the effects of free wine tastings at wine shops”, International Journal of Wine Business Research, Vol. 21 No. 4, pp. 312-324.
Maunder, W.J. (1973), “Weekly weather and economic activities on a national scale: an example using United States retail trade data”, Weather, Vol. 28 No. 1, pp. 2-19.
Maze, T., Agarwal, M. and Burchett, G. (2006), “Whether weather matters to traffic demand, traffic safety, and traffic operations and flow”, Transportation Research Record: Journal of the Transportation Research Board, Vol. 1948 No. 1, pp. 170-176.
Menne, M.J., Durre, I., Vose, R.S., Gleason, B.E. and Houston, T.G. (2012), “An overview of the global historical climatology network-daily database”, Journal of Atmospheric and Oceanic Technology, Vol. 29 No. 7, pp. 897-910.
Murray, P.W., Agard, B. and Barajas, M.A. (2018), “Forecast of individual customer’s demand from a large and noisy dataset”, Computers and Industrial Engineering, Vol. 118, pp. 33-43.
Murray, K.B., Di Muro, F., Finn, A. and Popkowski Leszczyc, P. (2010), “The effect of weather on consumer spending”, Journal of Retailing and Consumer Services, Vol. 17 No. 6, pp. 512-520.
OPM (2019), Federal Holidays Schedule - US Office of Personnel Management, U.S. Office of Personnel Management, available at: www.opm.gov/policy-data-oversight/snow-dismissal-procedures/federal-holidays/#url=2013 (accessed 19 December 2019).
Parsons, A.G. (2001), “The association between daily weather and daily shopping patterns”, Australasian Marketing Journal (AMJ), Vol. 9 No. 2, pp. 78-84.
Phillips, P.C.B. and Perron, P. (1988), “Testing for a unit root in time series regression”, Biometrika, Vol. 75 No. 2, pp. 335-346.
Ridderstaat, J. and Croes, R. (2020), “A framework for classifying causal factors of tourism demand seasonality: an interseason and intraseason approach”, Journal of Hospitality and Tourism Research, Vol. 44 No. 5, pp. 733-760.
Rimm, E.B., Klatsky, A., Grobbee, D. and Stampfer, M.J. (1996), “Review of moderate alcohol consumption and reduced risk of coronary heart disease: is the effect due to beer, wine, or spirits?”, BMJ, Vol. 312 No. 7033, pp. 731-736.
Roslow, S., Li, T. and Nicholls, J.A.F. (2000), “Impact of situational variables and demographic attributes in two seasons on purchase behaviour”, European Journal of Marketing, Vol. 34 Nos 9/10, pp. 1167-1180.
Santos, J.A., Ceglar, A., Toreti, A. and Prodhomme, C. (2020), “Performance of seasonal forecasts of douro and port wine production”, Agricultural and Forest Meteorology, Vol. 291, pp. 1-10.
Starr-McCluer, M. (2000), The Effects of Weather on Retail Sales, Report, St. Louis, Federal Reserve Bank of St Louis, January 2000.
Steele, A.T. (1951), “Weather's effect on the sales of a department store”, Journal of Marketing, Vol. 15 No. 4, pp. 436-443.
Sun, W., Govind, R. and Garg, N. (2009), “Geographic differences in the consumption of hedonic products: What the weather tells the marketer!”, Advances in Consumer Research, Vol. 36, pp. 710-711.
Thach, E.C. and Olsen, J. (2006), “Building strategic partnerships in wine marketing: implications for wine distribution”, Journal of Food Products Marketing, Vol. 12 No. 3, pp. 71-86.
Wu, Z., Huang, N.E., Long, S.R. and Peng, C.K. (2007), “On the trend, detrending, and variability of nonlinear and nonstationary time series”, Proceedings of the National Academy of Sciences of Sciences, Vol. 104 No. 38, pp. 14889-14894.
Yohannes, M.F. and Matsuda, T. (2016), “Weather effects on household demand for coffee and tea in Japan”, Agribusiness, Vol. 32 No. 1, pp. 33-44.
Zhang, X., Pei, J. and Ye, X. (2016), “Demographic transformation and clustering of transactional data for sales prediction of convenience stores”, in IEEE (Eds), 2016 IEEE International Conference on Cloud Computing and Big Data Analysis (ICCCBDA), Chengdu, China, pp. 102-108.