Driver and Consequences of Multichannel Shopping

Previous research has investigated what happens when customers start utilizing more than a single sales channel (i.e., become multichannel). This research stream has identified two key consequences of multichannel usage. First, Shankar et al. (2003) and Hitt and Frei (2002) determine that customers using an internet channel in addition to the traditional brick-and-mortar channel are more loyal than customers who use a single channel. Sousa and Voss (2004) explain that these higher customer retention rates are because of increased coordination between channels; the coordination among channels increases customer satisfaction, which improves retention rates. Second, Neslin et al. (2006), Thomas and Sullivan (2005), Kumar and Venkatesan (2005), Venkatesan et al. (2007), Ansari et al. (2008), and Kushwaha and Shankar (2008) determine that on average multichannel customers spend more than single channel customers. Although plenty of research exists about multichannel customer management, there is relatively little known about the drivers that induce customers to adopt a new channel. Additionally, previous research has mainly focused on the short term effects and has not attempted to quantify, if any, the long-term effects of multichannel usage. This dissertation examines multichannel customers' decisions. Specifically, I address the following questions: (1) What factors lead customers to adopt new sales channels? and (2) What is the long-term effect of multichannel shopping on customers' spending? The first essay investigates the drivers of new sales channel adoption. In this essay, I propose a conceptual framework grounded in diffusion theory, and test this framework on longitudinal data from a major catalog company using a discrete-time, hazard model. This essay contributes to the marketing literature by providing empirical evidence that social influence impacts the timing of new channel adoption. I find that longer tenured customers are more eager to adopt a new channel and less impacted by social influence. I also find that customers adopt a physical store at a faster rate than an Internet store. Moreover, social influence and customer tenure play more important roles when customers adopt an Internet channel than a brick-and-mortar channel. In contrast, marketing activities play a more important role in customers' adoption of the physical store than in the customer's adoption of the internet channel. These new findings have implications for identifying early adopters and accelerating the diffusion of a new channel. The second essay is the first study to look at how multichannel shoppers' spending pattern changes over time, and is distinctive from past research which examines multichannel customers' spending only in the short term. For this study, I examine longitudinal data from a major U.S. retailer. My empirical analysis is likely to be affected by self-selection bias because heavy users may self-select themselves into using more than one channel. To control for such bias, I combine different panel data econometrics techniques with the propensity score matching method. The results provide empirical evidence that multichannel customers increase their spending when they initially start to use a new channel. In the long run, however, I find that the difference between multichannel and mono-channel customers' spending disappears. The findings have implications for predicting revenue streams from multichannel customers over time. Methodologically, this study is the first to combine dynamic panel data estimation with the propensity score matching. In addition, several papers in social sciences rely on aggregate level data (for example, zip code level demographics from U.S. Census), to create matched pairs. These papers are criticized as some scholars (Gensler et al., 2012) argue that zip code level data do not provide sufficient information to construct functional matched pairs. To address this issue, I create matched pairs based on U.S. Census data and household level data. The findings show that the estimates obtained by both matching procedures are exceptionally similar results