To further examine the robustness of the results obtained from the HLM analyses, we tested the hypotheses pooling respondents across stores using two additional methods: (a) OLS regressions, and (b) regressions with a cluster correction of the error covariance matrix (Rogers, 1993). Although OLS ignores the nesting nature of the data and thus may produce biased estimators of standard errors, OLS might be more stable in small samples and more robust against model misspecification than HLM (James & Williams, 2000) and therefore useful for checking purposes. The cluster method adjusts the estimated variance-covariance structure of the error terms to account for the interdependence among observations from the same store and heterogeneous errors across stores (see Glomb & Liao, 2003; Liao, Arvey, Butler, & Nutting, 2001; Milton & West- phal, 2005). We found that the pattern of results from the OLS regressions and the regressions with the cluster correction for both the longitudinal sample and the combined sample was highly consistent with that from the HLM analyses, providing additional confidence in our statistical inferences.