Abstract
In the sales literature it is standard practice for researchers to collect cross-sectional data from multiple salespeople, and to compare those salespeople on the data obtained. This between-person approach is suitable for research aiming to draw conclusions between salespeople. However, many salesperson processes are dynamic and vary within salespeople over time, requiring datasets with repeated-measures. This article highlights the need to adopt a within-person theoretical perspective in sales research. Critically, the article shows how our present understanding of boundary conditions may change depending on whether a between-person or within-person level of analysis is adopted. Using examples from the sales literature, we show how the practical implications from between-persons research designs do not necessarily generalize to the within-person level. Further, we explain the methodological and analytical considerations that researchers must account for when undertaking within-person research. Furthermore, the article provides decision criteria that help to identify when within-person analysis should be conducted, outlining analysis tools that are capable of correctly estimating within-person effects without bias. Examples of how within-person research can enhance theory within future sales research, and how within-person research may influence management implications are also discussed. Finally, potential remedies to within-person research barriers are given.
Notes
1 Within- and between-person levels of analysis are not the same as those discussed in Johnson, Friend, and Horn (Citation2014). In their article, a “within” analysis refers to relationships within organizations, and not within-individuals over time.
2 Importantly, within-person research must not to be confused with research examining differences both within and between companies (e.g., Briggs, Jaramillo, and Weeks Citation2012), which is multilevel cross-sectional research, not dealing with intra-individual changes.
3 This is analagous to the “groups” discussed in cross-sectional multilevel modeling literature.
4 In a multilevel analysis, individual change (intra-individual) becomes level 1 and individual differences (inter-individual), level 2.
5 In this analytic model the residuals become the isolated within-person estimates, successfully achieving disaggregation.
6 Variables can also be grand-mean centered; however, this form of centering does not correctly disaggregate within- and between- variance (Curran and Bauer Citation2011).
7 Interested readers should see McArdle (Citation2009), Hamaker, Kuiper, and Grasman (Citation2015), Curran et al. (Citation2014), and Tate (Citation2004), for further reading material on the change score model, random-intercept cross-lagged panel model, latent growth model with structured residuals, and slopes-as-outcomes models, respectively.
8 How to treat missing data is beyond the scope of this paper