A Comparative Empirical Study of Discrete Choice Models in Retail Operations

Choice-based demand estimation is a fundamental task in retail operations and revenue management, providing necessary input data for inventory control, assortment, and price-optimization models. The task is particularly difficult in operational contexts where product availability varies over time and customers may substitute into the available options. In addition to the classical multinomial logit (MNL) model and extensions (e.g., nested logit, mixed logit, and latent-class MNL), new demand models have been proposed (e.g., the Markov chain model), and others have been recently revisited (e.g., the rank list-based and exponomial models). At the same time, new computational approaches were developed to ease the estimation function (e.g., column-generation and expectation-maximization (EM) algorithms). In this paper, we conduct a systematic, empirical study of different choice-based demand models and estimation algorithms, including both maximum-likelihood and least-squares criteria. Through an exhaustive set of numerical experiments on synthetic, semisynthetic, and real data, we provide comparative statistics of the predictive power and derived revenue performance of an ample collection of choice models and characterize operational environments suitable for different model/estimation implementations. We also provide a survey of all the discrete choice models evaluated and share all our estimation codes and data sets as part of the online appendix. This paper was accepted by Vishal Gaur, operations management.

The evolution of choice set formation in dwelling and location with rising prices: A decadal panel analysis in the Greater Toronto Area

Home location choice is based on both the characteristics of the dwelling (e.g., size, style, number of bedrooms) and the location (e.g., proximity to work, quality of schools, accessibility). Recent years have seen a steep increase in the price of housing in many major cities. In this research, we examine how these price increases are affecting the types of dwelling and locations considered by households. A large sample of real estate listings from 2006 and 2016 from the Greater Toronto Area is used to develop the empirical models. Two recently developed discrete choice models are used in the study: a nested logit model with latent class feedback (LCF) and a semi-compensatory independent availability logit (SCIAL) model. A method of alternative aggregation is proposed to overcome the computational hurdle that often impedes the estimation of choice set models. We find a significant increase in the probability of larger households considering townhouses and apartments over detached single-family dwellings between 2006 and 2016.

Why do SMEs switch suppliers?

Purpose Combining a conceptual framework with empirical evidence, this study aims to offer insights into why small and medium-sized enterprises (SMEs) in the business-to-business beauty sector switch suppliers, due to pricing considerations. Design/methodology/approach Data gathered from 475 telephone surveys of Spanish hairdressers provide the input for discrete choice models for testing the proposed hypotheses. Findings The SMEs that change suppliers tend to be sensitive to promotions, express less satisfaction with a current supplier’s offerings and serve fewer customers who buy professional products for their in-home use. If SMEs are satisfied with the supplier’s services though, they are less likely to change and more prone to negotiate with that supplier. Research limitations/implications This study does not address why dissatisfied SMEs might remain with their current suppliers. Further research might replicate this study using additional pricing data from suppliers. Practical implications Suppliers in business-to-business (B2B) sectors can leverage these findings to allocate their marketing budgets optimally and establish service strategies that will enable them to retain buyers and reduce their switching risk. Originality/value As an extension of extant literature, this study specifies switching drivers for SMEs in the B2B beauty sector. The findings should apply throughout this worldwide service sector, as well as to similar markets such as health, beauty and personal care and well-being services.

Freight shipper’s mode choice preference for sustainable inland transportation

Modelling of transport mode choice preferencehas beencurrently regarded essential to identify the preference of transport agents towards a transport policy. This research proposes a model of the freight mode choice preference by using Stated Preference Methods for the data collection from industrial freight shippers in Gresik, Indonesia. This research examines how truck as an existing mode and rail as a sustainable mode alternative compete for goods movement. As the common feature of discrete choice models, the Binary Logit was utilized to analyze the data. The sensitivity of mode preferences was investigated by changing the shipping cost and hauling time. The results indicated that the respondents were sensitive to haul time and shipping cost. When the shipping cost and hauling time was similar, the probability of choosing a truck was 77%. However, the industrial freight shippers changed the preference when the truck’s shipping cost and haul time was higher than that in the train. The train had a 65% choosing probability when the cost difference was IDR 500,000 lower and the hauling time difference was two days faster than the truck. This study assisted the policymakers to correctly design the variables of shipping cost and hauling time for the future sustainable inland transportation based on freight shipper’s preferences.

Independent Retailer Restocking Choices in Urban Goods Movement and Interaction Effects with Traditional Markets

Traditional markets play a key role in local supply chains in many countries, often influencing retailer decisions due to their inherent attractiveness. In contrast to restocking choices for retailers as part of large chains, choices of independent retailers driven by local traditional markets have not been widely researched and are not well understood. This paper analyses the factors influencing independent retailer restocking choices and investigates the interplay between the presence of traditional markets and retailer choices. Bandung city in Indonesia is chosen for the study where independent retailers are prevalent, and where a number of traditional markets are thriving. A retrospective questionnaire was used to capture independent retailer restocking behaviour and generation models were calibrated to arrive at the trip propensity. Discrete choice models were estimated to explain the retailer preferences for supplier location and transport service choice. Results indicate that trips generated by independent retailers are explained by the presence of traditional markets and retailers’ vehicle ownership, in addition to the standard variables such as number of persons employed, weekly goods demand and average shipment weight. As for restocking location choice, retailers are more likely to choose suppliers within a traditional market where the number of wholesaler units is larger. Furthermore, the choice of traditional markets has a positive influence on whether retailers choose to use their own vehicle to restock their shops.

On Stochastic-User-Equilibrium-Based Day-to-Day Dynamics

Researchers have proposed many different concepts and models to study day-to-day dynamics. Some models explicitly model travelers’ perceiving and learning on travel costs, and some other models do not explicitly consider the travel cost perception but instead formulate the dynamics of flows as the functions of flows and measured travel costs (which are determined by flows). This paper investigates the interconnection between these two types of day-to-day models, in particular, those models whose fixed points are a stochastic user equilibrium. Specifically, a widely used day-to-day model that combines exponential-smoothing learning and logit stochastic network loading (called the logit-ESL model in this paper) is proved to be equivalent to a model based purely on flows, which is the logit-based extension of the first-in-first-out dynamic of Jin [Jin W (2007) A dynamical system model of the traffic assignment problem. Transportation Res. Part B Methodological 41(1):32–48]. Via this equivalent form, the logit-ESL model is proved to be globally stable under nonseparable and monotone travel cost functions. Moreover, the model of Cantarella and Cascetta is shown to be equivalent to a second-order dynamic incorporating purely flows and is proved to be globally stable under separable link cost functions [Cantarella GE, Cascetta E (1995) Dynamic processes and equilibrium in transportation networks: Towards a unifying theory. Transportation Sci. 29(4):305–329]. Further, other discrete choice models, such as C-logit, path-size logit, and weibit, are introduced into the logit-ESL model, leading to several new day-to-day models, which are also proved to be globally stable under different conditions.