The Effect of the Franchises on Local Bakeries with High-Dimensional Controls

The Effect of the Franchises on Local Bakeries with High-Dimensional Controls

Since the 2013 designation of the bakery industry as a suitable business for small- and medium-sized enterprises in South Korea, there have been persistent questions about the validity of the designation and the suitability of the 500-meter store distance limits. This paper analyzes the effect of the entrance of a large franchise bakery on the closing rates of neighboring rival bakery using the Cox proportional hazard model. Our empirical results reveal that standard onestep variable selection like Lasso selection can lead to omitted variable bias in the Cox proportional hazard model, since it may exclude possible confounding variables from the estimation. We first show that the double-Lasso (Least Absolute Shrinkage and Selection Operator) selection method proposed by Belloni et al. (2014) can resolve the omitted variable biases through the Monte-Carlo simulations. The empirical results using the double-Lasso selection procedure suggest that the number of franchise bakeries within 200 meters of a bakery significantly increases the closure rate of that bakery, but that this effect becomes insignificant when we restrict our analysis to small- and medium-sized bakeries.