Understanding Consumer Reactions to Assortment Unavailability

infrequently in the sample stores and therefore did not make a purchase before the delisting in
the studied time period.
Because the detergent sales of new buyers equal 0 before the delisting, we cannot apply
the same methodology as we did previously. Instead, we consider the detergent sales generated
by the new buyers relative to the total detergent sales. For all stores, we expect this percentage to
increase over time, because some households that bought detergent before the delisting may stop
purchasing detergents at the store but more and more new households will enter. The stores in
which the assortment reduction actually took place may attract more new buyers relative to the
control stores, which will lead to a larger percentage of purchases made by new buyers.
Figure 4.5: Sales effect due to entrance of new buyers
To quantify the difference, we again perform a regression analysis using spline functions (see
0.30
0.25
0.20
0.15
0.10
0.05
0.00
Baseline sales generated by new buyers (% of total detergent sales)
2003
Additional sales generated by new buyers in test stores (% of total detergent sales)
2003
Figure 4.5). We initiate our analysis two weeks after the start of the delisting. In Figure 4.5, the
top graph shows the estimated baseline effect, which demonstrates that, regardless of the
delisting, the percentage of sales attributed to new buyers tends to increase over time. The
bottom graph shows the additional effect in the test stores, namely, the additional share of sales
of the new buyers in the test store. Immediately following the delisting, there is no significant
difference between the control and the test stores. However, at the conclusion of our sample, the
111