What Explains Superior Retail Performance?

We analyze the performance of retail firms for the period 1978-97 using public financial data. Our performance measures are long-term stock returns and whether the firm filed for bankruptcy in the period of study. We assume that over a long time period of at least five years, stock returns are a reasonable measure of the overall success of a firm. We have found a very wide disparity in performance between firms. On the one hand, retailers like Wal-Mart, the Gap and Circuit City have had phenomenal success (nineteen year compounded stock returns of 31.2%, 29.5%, and 34.5%, respectively), while on the other, 17% of the public retail firms filed for bankruptcy. We investigate how the following levers managed by the CEO of a retail firm affect performance: return on assets, sales growth, inventory turns, gross margin, financial leverage, and selling, general, and administrative expenses. The nature of the analysis is contemporaneous, providing insights into managerial actions that correlate with success as measured by stock returns, but not a prediction of future stock returns. We find that (1) return on assets, sales growth, standard deviation of return on assets and financial leverage explain more than 50% of the variation in stock returns for periods of ten years or more; (2) retailers in different segments–apparel, department stores, grocery and convenience stores, drugs and pharmaceuticals, jewelry, consumer electronics, home furnishings, toys, and variety stores–achieve similar return on assets and return on equity by following very different strategies with respect to their gross margins and inventory turns; (3) even within the same segment, high gross margin correlates with low inventory turns, and with high selling, general, and administrative expenses; (4) risk of bankruptcy is related to the mismatch between how fast a firm attempts to grow versus what growth rate it realizes. We also test for a negative correlation between sales growth and return on assets, which is widely believed to be true but is not borne out by our data.