Consumer Demand for Labor Standards, Part II 

  Michael Hiscox and Nicholas Smyth , guest bloggers 

 We continue yesterday's entry discussing questions that arose during our recent presentation of our paper on consumer demand and labor standards labeling. 

 Another excellent question that was raised in the discussions concerned the evidence that sales of our labeled items actually rose (relative to sales of unlabeled control products) when their prices were raised. We have been interpreting this as evidence that consumers regarded the label as more credible when the product was more expensive relative to alternatives, since they expect to pay  more for higher labor standards. One question was whether relative sales would have risen with price increases for any good (labeled or unlabeled) just because higher prices can signal better quality. Since we did not raise the price of unlabeled items, we cannot address this concern directly. It is not critical to one of our main findings: sales of labeled items increased markedly relative to sales of unlabeled alternatives when the labels were put in place (before prices were adjusted). But we will try to track down the research on the price-quality issue in the literature on consumer psychology. Our basic assumption is that the existing (equilibrium) product prices and sales levels at ABC (in the "baseline" period) accurately reflected the relative quality of treatment and control products. 

 Other questions raised concerned the evidence we discussed in the paper about the marked increase in sales of Fair Trade Certified coffee. It was pointed out that, to the extent that retailers like Starbucks are marketing only fair trade coffee as the brewed "coffee of the day" this seems more like a general CSR strategy by the firm and not a sign of demand for improved standards. We were really talking about sales of certified coffee beans, rather than brewed coffee. The labeled beans are sold in direct competition with similar (unlabeled) beans at both Starbucks and Peets. But it is important that we check the data and see if we can discriminate clearly between sales in different categories. 

 In general, we felt we have to do better in accounting for seasonal patterns in demand for home furnishings at ABC and how they might bear on our findings. This is obviously not a problem for our core results that hinge on the ratio of sales of labeled brands to unlabeled brands during each phase of the experiment. But for measuring price elasticities using changes in absolute sales of labeled items over time we would like to allow for the fact that sales of home furnishings were expected to dip during the summer months. To do this, we will probably need to estimate weekly sales for each brand using all the data we have from ABC prior to the start of our experiment (covering sales in 2004 and the first half of 2005). The relevant covariates would probably include recorded levels of total foot traffic in the store, total sales of other store products, some national or regional measures of economic activity and consumer confidence, variables accounting for any special sales and promotional campaigns, and seasonal dummies. We can then compare actual (absolute) sales of labeled brands with out-of-sample predictions based upon the estimations and thereby gauge the impact of our experimental treatments. 

 We will conclude our discussion in tomorrow's post.