International Choice Modelling Conference, International Choice Modelling Conference 2017

Font Size: 
In-store or online shopping of search and experience goods: A Hybrid Choice approach
Basil Christian Schmid, Simon Andreas Schmutz, Kay Werner Axhausen

Last modified: 28 March 2017

Abstract


Motivation and objectives

Information and communication technologies (ICT) have experienced a persistent increase in usage over the last 25 years, which, in the context of e-commerce, allow for a more flexible spatial and temporal accomplishment of shopping activities. Identifying the key drivers that affect individuals’ choices between in-store and online shopping is not only important for developing effective retailing policies, but also for predicting the responsiveness to specific attributes of heterogeneous consumer segments in the context of activity-based travel demand modeling: How do people value travel, delivery and shopping/ordering time when directly facing the trade-offs between these two alternative shopping channels?

Results provide new insights on purchasing channel preferences for two distinctly different types of goods by allowing attribute sensitivities to differ by shopping purpose: Search goods (e.g. standard electronic appliances) are more often purchased online, where search costs are substantially lower, while the main product characteristics of experience goods (e.g. groceries) can only be obtained in-store (Peterson et al., 1997).

Methodology

Past research has shown that apart from the shopping purpose, attitudes play a major role in explaining choice behavior (e.g. Mokhtarian and Tang, 2013). We estimate a Hybrid Choice model with alternative-specific attributes using stated preference (SP) data, applying an integrated choice and latent variable approach (e.g. Ben-Akiva et al., 2002): Two latent variables that are hypothesized to affect the choice of the purchasing channel are included, capturing the acceptance level of ICT and online shopping and the pleasure of shopping, which both are defined by various socio-economic characteristics (see supplementary file: modeling_framework.png). Using simulated maximum likelihood, we show that the structural part of the model exhibits unstable results when using Halton instead of MLHS draws (Hess et al., 2006) even for large (>1000) number of draws and yields more robust effects if the attitude indicator probabilities are averaged over subjects before entering the log-likelihood function.

Data

As part of a multi-stage travel survey with 467 respondents in Zurich, Switzerland, 2016, a SP experiment requested respondents to trade-off different attributes related to their choice between in-store and online shopping, including attitudinal questions of shopping related aspects. Personalized attribute levels were presented based on previously observed shopping behavior from the travel and expenditure diaries.

Results and implications

The average value of delivery time savings (VDTS) of about 6 CHF/day (1 CHF = 1.04 US$) for groceries (G) is about three times higher than for standard electronic appliances (E), being consistent with the hypothesis that buying E is usually done on a more irregular basis and exhibits a longer planning horizon, thus leading to a lower disutility of delivery time. Given the relatively low VDTS, the average values of travel time savings of about 60 CHF/hour for G and 45 CHF/hour for E indicate a potential for ICT services in both categories. It shows that similar to delivery time, people weight the time to travel for buying E lower than for G. Importantly, respondents did not react on travel costs, but were anchoring behavior with respect to shopping costs, which was not the case for delivery costs: Shopping costs were perceived as much less unpleasant than delivery costs, and online retailers should better incorporate them in the price when designing an effective retailing strategy.

A larger size/weight of the shopping basket was found to strongly increase the choice probability of online shopping; for older and female respondents in an almost deterministic pattern. Apart from a random-effects specification of the alternative-specific constant, a log-normal distribution of the size/weight coefficient was included to account for the heterogeneity in physical conditions.

A higher acceptance level of ICT and online shopping (LV1) shows a strong and positive effect on choosing the online alternative, being more than twice as large for G than for E: Buying groceries online – nowadays still unusual – is mainly driven by ICT affine traits. Supporting the findings by Rudolph et al. (2004) that price advantages are a main factor for doing online shopping in Switzerland, an increased LV1 implies a higher shopping cost sensitivity, which can be explained by the larger choice set when effectively considering both purchasing channels. Thus, given the strong positive correlation between income and LV1, the income elasticity of shopping cost is not significantly different from zero. In addition, younger, well-educated and male car users facing a low store accessibility tend to have a higher LV1.

The pleasure of shopping (LV2), being higher for well-educated, non-working unmarried women facing a high store accessibility, shows a negative effect on choosing the online alternative for G; for E the estimated coefficient is zero. Especially for experience goods, going to a store is more attractive in terms of shopping enjoyment than for search goods: Respondents exhibiting a high LV2 value the time of in-store shopping for G less negative than for E. To summarize, heterogeneity in cost sensitivity is mainly captured by differences in acceptance levels of online shopping, while time sensitivities differ by shopping purpose and the pleasure of shopping.

Literature

Ben-Akiva, M., D. McFadden, K. Train, J. Walker, C. Bhat, M. Bierlaire, D. Bolduc, A. Boersch-Supan, D. Brownstone, D. S. Bunch and others (2002) Hybrid choice models: Progress and challenges, Marketing Letters, 13 (3) 163–175.

Hess, S., K. E. Train and J. W. Polak (2006) On the use of a modified latin hypercube sampling (MLHS) method in the estimation of a mixed logit model for vehicle choice, Transportation Research Part B: Methodological40 (2) 147-163.

Mokhtarian, P. L. and W. L. Tang (2013) Trivariate Probit models of pre-purchase/purchase shopping channel choice: Clothing purchases in northern California, Choice Modelling: The State of the Art and the State of Practice - Selected Papers from the Second International Choice Modelling Conference, chapter 12, 243-273, edited by S. Hess and A. Daly, Edward Elgar Publishing Ltd.

Peterson, R. A., S. Balasubramanian and B. J. Bronnenberg (1997) Exploring the implications of the Internet for consumer marketing, Journal of the Academy of Marketing Science25 (4) 329-346.

Rudolph, T., B. Rosenbloom and T. Wagner (2004) Barriers to online shopping in Switzerland, Journal of International Consumer Marketing16 (3) 55-74.


Conference registration is required in order to view papers.