Measuring the impact of alcohol multi-buy promotions on consumers' purchase behaviour

doi:10.1016/j.jocm.2016.05.001

Journal of Choice Modelling

Volume 24, September 2017, Pages 75-95

https://doi.org/10.1016/j.jocm.2016.05.001 Get rights and content

Abstract

The objective of this study was to understand the impact of alcohol multi-buy promotions on individual's purchasing behaviour. Our study deployed a Stated Preference survey to measure consumers' potential responses towards price changes and the introduction of promotions, as well as the resulting effects on demand. A series of econometric models were developed, ranging from simple selection models to advanced multiple discrete-continuous extreme value (MDCEV) models, to capture the discrete and continuous feature of alcohol purchasing choice behaviour. The model results were compared and then extrapolated to a series of policy scenario tests, to enable the evaluation of factors that underpin consumers' alcohol purchasing behaviour. This research contributes to evidence on the role of multi-buy promotions on alcohol purchasing behaviour, as well as adding to recent developments in the choice modelling literature by providing a comparison of results across a range of different model structures suitable for the analysis of data such as used here.

Introduction

Excessive alcohol consumption is a major cause of ill-health and mortality and is also associated with economic and social harm. The Department of Health in the UK has estimated that the harmful use of alcohol costs the National Health Service (NHS) approximately £3.5 billion per year and 7% of all hospital admissions were alcohol related¹ in 2009–2010. The Government Alcohol Strategy report, 2012, 2 stated that the cost of alcohol related harm is estimated to be £21 billion annually.

Alcohol pricing is considered by some to be a potential means of influencing levels of alcohol consumption (Anderson et al., 2009, Purshouse et al., 2010). However, alcohol pricing is a sensitive policy issue, with those in favour of price regulation arguing that it has the potential to reduce harms from overconsumption of alcohol, and those against emphasising the need to limit the impact on those who drink alcohol in moderation.

In this study, we examine the impact of one aspect of pricing – multi-buy promotions – on consumers' purchasing of alcohol for consumption off the premises where the purchase was made. By multi-buy promotions, we refer specifically to promotions where there is a link between the number of products purchased and the price of the product, for example ‘two for the price of one’, ‘three for the price of two’ or the purchase of more than one item for a fixed price discount. The study was carried out under considerable time pressure between September 2012 and January 2013 in order to inform the impacts of proposed policy to ban alcohol multi-buy promotions.

We contribute to the empirical literature on better understanding alcohol purchase behaviour under multi-buy promotion in three ways. First, we deploy a novel data collection approach to measure consumers' stated alcohol purchases under different market changes which are not easily observed in the real market. Second, a series of econometric techniques ranging from the Tobit and Heckman models to advanced MDCEV models are developed to explain the discrete – continuous nature of alcohol purchasing behaviour. This is, to the best of our knowledge, the first time that MDCEV models have been used for interpreting consumers' alcohol purchase behaviour. Third, the results are then used in a series of policy scenario tests, which help to gauge the potential impact of removing alcohol multi-buy promotions.

The remainder of the paper is organised as follows. Section 2 briefly describes the design of the stated preference survey and the data collection more widely. Section 3 summarises key literature regarding econometric models that predict discrete–continuous choices with a special emphasis on the comparison of these models. Section 4 discusses the estimation of the econometric models. Section 5 presents a series of policy scenario tests, followed by Section 6 which concludes the paper with the discussion of the policy implications and future work.

Section snippets

Stated preference survey design

Given the limitations of available retail measurement data (lack of detail on consumers, and limited information on promotions), an online survey was designed to collect self-reported information on existing patterns of alcohol consumption and purchasing. This included a stated preference component to examine potential responses to alcohol promotions under different market situations, including multi-buy promotions. Stated choice techniques have been widely used in marketing, environmental

Econometric models

The stated preference scenarios collected respondents' choices of both the type of alcohol they would purchase (a discrete choice) and the amount that they would purchase (a continuous value). When zero purchases are present, standard econometric models using ordinary least squares (OLS) based on all the positive observations would generate biased parameter estimates (Amemiya, 1984). In addition, excluding zero observations would cause a loss of efficiency.

Model estimation

In all the models, the dependent variable was the volume of alcohol purchased, reflected by the number of units, using average alcohol by volume (ABV) conversions (as in the methodology employed for the General Lifestyle Survey). This follows on from initial tests using expenditure as the dependent variable, which yielded worse fit to the data, as did a log formulation of the explanatory variables. We found that the pattern of model estimations from Moderate A and Moderate B alcohol consumption

Forecasting framework

To understand consumers' likely responses to alcohol policy interventions and their behaviour in a changing market, the Tobit, Heckman and MDCEV models were next used in a forecast framework. For this, the sample observations were re-weighted to achieve a nationally representative distribution of population characteristics and consumption level.

The forecast models are applied to the first record of the SP survey for each respondent in the main survey, weighted to reflect population and

Discussion and conclusions

The aim of this study was to measure the impact of multi-buy promotions on consumers' off-trade alcohol purchasing. Our analysis complements the existing evidence by deploying a stated preference approach to measure and to explain consumers' responses to changing price and multi-buy promotions which might not be easily observed in the real market. A range of econometric models were developed to interpret the consumers' behaviour and the impact of multi-buy promotion. The Tobit and Heckman

Acknowledgement

The work reported in this paper is based on a study commissioned jointly by Her Majesty's Revenue and Customs (HMRC), the UK Department of Health (DH) and the UK Home Office. The interpretation of the results as well as any mistakes, remains the responsibility of the authors. The second author acknowledges the financial support by the European Research Council through the consolidator grant 615596‐DECISIONS.

References (45)

T. Amemiya
Tobit models: a survey
J. Econ.
(1984)
P. Anderson et al.
Effectiveness and cost-effectiveness of policies and programmes to reduce the harm caused by alcohol
Lancet
(2009)
C. Bernardo et al.
An empirical investigation into the time-use and activity patterns of dual-earner couples with and without young children
Transp. Res. Part A: Policy Pract.
(2015)
C.R. Bhat
A multiple discrete-continuous extreme value model: formulation and application to discretionary time-use decisions
Transp. Res. Part B
(2005)
C.R. Bhat
The Multiple Discrete-Continuous Extreme Value (MDCEV) Model: Role of Utility Function Parameters, Identification Considerations, and Model Extensions
Transp. Res. Part B
(2008)
C.R. Bhat et al.
A new estimation approach for the multiple discrete-continuous probit (MDCP) choice model
Transp. Res. Part B
(2013)
C.R. Bhat et al.
The impact of demographics, built environment attributes, vehicle characteristics, and gasoline prices on household vehicle holdings and use
Transp. Res. Part B
(2009)
M. Castro et al.
Accommodating multiple constraints in the multiple discrete-continuous extreme value (MDCEV) choice model
Transp. Res. Part B,
(2012)
C.A. Gallet et al.
Elasticities of beer demand revisited
Econ. Lett.
(1998)
A.R. Pinjari et al.
A Multiple discrete-continuous nested extreme value (mdcnev) model: formulation and application to non-worker activity time-use and timing behavior on weekdays
Transp. Res. Part B
(2010)

R. Purshouse et al.

Estimated effect of alcohol pricing policies on health and health economic outcomes in England: an epidemiological model

Lancet

(2010)

B. Yu et al.

Modeling household energy consumption behavior: a comparative analysis

Transp. Res. Part D: Transp. Environ.

(2015)

B. Yu et al.

Representing in-home and out-of-home energy consumption behavior in Beijing

Energy Policy

(2011)

Accent and RAND Europe, 2002. Terrestrial Digital Telephone Services: RP/SP Survey Methodology. Internal research...

Augustin, B., Pinjari, A.R., Eluru, N., Pendyala, R.M., 2015. Alternative Approaches for Estimating Annual Mileage...

C. Bhat et al.

Multiple discrete-continuous choice models: a reflective analysis and a prospective view

J.R. Blaylock et al.

Wine consumption by US men

Appl. Econ.

(1993)

Bolduc, D., Khalaf, L., Moyneur, E., 2001. Joint discrete/continuous models with possibly weak identification. In:...

R. Breen

Regression Models: Censored, Sample Selected, or Truncated Data

(1996)

Collis, J., Grayson, A., Johal, S., 2010. Econometric Analysis of Alcohol Consumption in the UK, HMRC Working Paper 10....

R.T. Carson et al.

The tobit model with a non-zero threshold

Econ. J.

(2007)

A. Daly

Forecasting choice

Cited by (22)

Modelling activity patterns of wild animals - An application of the multiple discrete-continuous extreme value (MDCEV) model
2023, Journal of Choice Modelling
Advanced econometric models used in the field of transport or marketing are becoming increasingly sophisticated and able to capture complex decision making and outcomes. In this paper, we apply state-of-the-art discrete-continuous choice models to the field of Ecology, in particular to model activity engagement of the population of Swedish Brown bears. Using data from GPS collars that track wild animals over time, we estimate a Multiple Discrete-Continuous Extreme Value (MDCEV) model to understand activity engagement and duration as a function of both bear characteristics and other external factors. We show that the methodology is not only suitable to address this aim, but also allows us to produce insights into the connection between the animal's age and gender and activity engagement as well as the links with climate variables (temperature and precipitation) and human activity (hunting).
Extending the Multiple Discrete Continuous (MDC) modelling framework to consider complementarity, substitution, and an unobserved budget
2022, Transportation Research Part B: Methodological
Many decisions can be represented as interrelated discrete and continuous choices, i.e. what and how much to choose from a set of finite alternatives (incidence and quantity of consumption). In the last twenty years, several models of Karush–Kuhn–Tucker demand systems have been developed and used to study these kinds of decisions. While strongly grounded in economic theory, most of these models have two limitations: they require specifying a budget, and usually omit any complementarity effects. In this paper, we propose two extensions to the Multiple Discrete Continuous (MDC) modelling framework: (i) an MDC model including explicit complementarity and substitution effects, and (ii) an MDC model with complementarity, substitution that requires no budget definition. Model (ii) relies on the hypothesis that total expenditure on the alternatives under consideration is small compared to the overall budget. This allows using a linear utility function for the numeraire good, leading to a likelihood function without the budget or numeraire good in it. The lack of a budget is specially useful when forecasting, as it avoids cascading errors due to an inaccurate budget specifications. The inclusion of complementarity and substitution effects enriches the interpretability of the models, while the resulting functional form avoids theoretical issues present in previous formulations. Alongside the derivation of the models, we discuss their main properties and propose an efficient forecasting algorithm for (ii). We also report four applications to datasets about time use, household expenditure, supermarket scanner data, and trip generation. Free estimation code for both models is made available online.
Volumetric choice experiments (VCEs)
2022, Journal of Choice Modelling
Citation Excerpt :
Finding the quantities picked to be almost always zero, one, or two, they estimate an interesting variant of an ordered logit model. In a choice experiment involving beer, wine, and spirits, Lu et al. (2017) demonstrated the versatility of the DCE approach by bundling individual items to create new goods. For example, buying three bottles of wine described by price and attributes as displaying the type of alcohol (e.g., Chardonnay of a particular quality) vs. buying other bundles that could differ by quantity, type, and quality (e.g., four bottles of a particular brand of vodka each with a different flavour added at a specified price).
Volumetric Choice Experiments (VCEs) are designed to capture purchase quantities rather than a single, discrete choice. They can be seen as an extension of Discrete Choice Experiments (DCEs) where individuals decide how many units of a specific good or service to buy/use rather than deciding whether to buy/use it or not. There is different information in such integer count data than is contained in traditional binary or multinomial discrete choices, which presents new opportunities and interesting challenges. Like DCEs, VCEs have different components ranging from experimental design to modelling and our focus is on the overall process of implementation rather than detailed analysis of components. Our empirical examples come from large-scale VCEs embedded in surveys administered to samples drawn from Information Resources, Inc. (IRI) consumer panel for two product categories: single serve-coffee K-pods and canned tuna. The response for each alternative is a planned purchase count, possibly zero. These counts are fit using a negative binomial regression with a multilevel mixed-effects specification. Our VCE design allows for statistical identification of own- (brand by size) and cross-price elasticities, plus the effects of other attributes and demographics and their interactions with prices. The external validity of our approach is compared to results on actual canned tuna data purchases from the same IRI panelists. Advantages and limitations of VCEs as well as many unresolved research issues are discussed.
Modelling the effects of COVID-19 on travel mode choice behaviour in India
2020, Transportation Research Interdisciplinary Perspectives
Citation Excerpt :
It may be noted that MDCEV models have been applied in many different empirical contexts, both in transport and beyond. Examples include applications to the choice of vehicle type and mileage (Bhat and Sen, 2006), time-use (Bhat, 2005; Calastri et al., 2020; Enam et al., 2018; Pendyala and Bhat, 2004; Srinivasan and Bhat, 2005); multi-buy alcohol promotions (Lu et al., 2017), patterns of social interaction between people and their social contacts (Calastri et al., 2017). But, to the best of our knowledge, this is the first application of the MDCEV framework in modelling mode and trip frequencies.
The COVID-19 pandemic has resulted in unprecedented changes in the activity patterns and travel behaviour around the world. Some of these behavioural changes are in response to restrictive measures imposed by the Government (e.g. full or partial lock-downs), while others are driven by perceptions of own safety and/or commitment to slow down the spread (e.g. during the preceding and following period of a lock-down). Travel behaviour amidst the stricter of these measures is quite straightforward to predict as people have very limited choices, but it is more challenging to predict the behavioural changes in the absence of restrictive measures. The limited research so far has demonstrated that different socio-demographic groups of different countries have changed travel behaviour in response to COVID-19 in different ways. However, no studies to date have either (a) investigated the changes in travel behaviour in the context of the Global South, or (b) modelled the relationship between changes in transport mode usage and traveller characteristics in order to quantify the associated heterogeneity. In this paper, we address these two gaps by developing mathematical models to quantify the effect of the socio-demographic characteristics of the travellers on the mode-specific trip frequencies before (January 2020) and during the early stages of COVID-19 spread in India (March 2020). Primary data collected from 498 respondents participating in online surveys have been used to estimate multiple discrete choice extreme value (MDCEV) models in this regard. Results indicate – a) significant inertia to continue using the pre-COVID modes, and b) high propensity to shift to virtual (e.g. work from home, online shopping, etc.) and private modes (e.g. car, motorcycle) from shared ones (e.g. bus and ride-share options). The extent of inertia varies with the trip purpose (commute and discretionary) and trip lengths. The results also demonstrate significant heterogeneity based on age, income, and working status of the respondents. The findings will be directly useful for planners and policy-makers in India as well as some other countries of the Global South in better predicting the mode-specific demand levels and subsequently, making better investment and operational decisions during similar disruptions.
Investigating passenger behavior in airport terminals with multisource data: Recall bias and time budget effects
2020, Transportation Research Part A: Policy and Practice
Passenger activities in terminals can be understood better in the current era, when sensors are pervasive throughout airports. This study demonstrated the usefulness of combining multiple-source data to investigate passenger behavior in airports. Objective time-use and terminal activity participation data of 266 air passengers’ behavior at Taipei Songshan International Airport were collected from beacons, a self-developed mobile application, and questionnaires. The study first investigated recall errors generated from self-report questionnaires, an approach commonly used in previous studies with a retrospective design. The study then assessed the participation and duration of terminal activities conducted by passengers with standard, nested, and mixed multiple discrete-continuous extreme value models. The analysis results revealed that recall errors were associated with activity choice and duration and passenger characteristics. These errors were not random but systematic and could potentially lead to biased results in retrospective studies solely based on self-report questionnaires. The estimation results of extreme value models generally confirmed the expected association of terminal activity choice with passenger characteristics, such as more frequent retail store visits in female passengers than in male passengers. However, the associations of passenger characteristics with activity duration could be consistent with or opposite to those with activity choice; for example, frequent flyers tended to consume food and beverages (F&B) more frequently and spend a longer time in restaurants whereas passengers having a long free dwell time were less likely than those having a short free dwell time to consume F&B but the duration was relatively long on average. Managerial implications for airports and retailers and recommendations for future air passenger behavioral studies were provided.
A multiple discrete extreme value choice model with grouped consumption data and unobserved budgets
2020, Transportation Research Part B: Methodological
In this paper, we propose, for the first time, a closed-form multiple discrete-grouped extreme value model that accommodates grouped observations on consumptions rather than continuous consumptions. For example, in a time-use context, respondents tend to report their activity durations in bins of time (for example, 15-minute intervals or 30-minute intervals, depending on the duration of an activity). Or when reporting annual mileages driven for each vehicle owned by a household, it is unlikely that households will be able to provide an accurate continuous mileage value, and so it is not uncommon to solicit mileages in grouped categories such as 0–4,999 miles, 5000–9,999 miles, 10,000–14,999 miles, and so on. Similarly, when reporting expenditures on different types of commodities/services, individuals may round up or down to a convenient dollar value of multiples of 10 or 100 (depending on the length of time in which expenditures are sought). In some other cases, a product itself may be available only in specific package sizes (such as say, instant coffee, which is typically packaged in fixed sizes). In this paper, we use the so-called linear outside good utility MDCEV structure of Bhat (2018) to show how the model can be used for grouped consumption observations. Of course, this is also possible because the linear outside good utility does not need a continuous budget value, and allows for unobserved budgets. We discuss an important identification issue associated with this linear outside good utility model, and proceed to demonstrate applications of the proposed model to the case of weekend time-use choices of individuals and vehicle type/use choices of households.

View all citing articles on Scopus

View full text

Measuring the impact of alcohol multi-buy promotions on consumers' purchase behaviour

Abstract

Introduction

Section snippets

Stated preference survey design

Econometric models

Model estimation

Forecasting framework

Discussion and conclusions

Acknowledgement

J. Econ.

Lancet

Transp. Res. Part A: Policy Pract.

Transp. Res. Part B

Transp. Res. Part B

Transp. Res. Part B

Transp. Res. Part B

Transp. Res. Part B,

Econ. Lett.

Transp. Res. Part B

Lancet

Transp. Res. Part D: Transp. Environ.

Energy Policy

Multiple discrete-continuous choice models: a reflective analysis and a prospective view

Wine consumption by US men

Appl. Econ.

Regression Models: Censored, Sample Selected, or Truncated Data

The tobit model with a non-zero threshold

Econ. J.

Forecasting choice