Generalized Synthetic Control Method Results


We have multiple treated (179 regulated SKUs) units from India. We created a clean donor pool of 747 SKUs for creating synthetic controls from the Philippines data. The synthetic control for each treated SKU (179 regulated SKUs from India) was created using the pre-regulation sales volumes of the control units (747 SKUs from Philippines). As we have multiple treated units (but treatment period is the same), we employ the generalized synthetic control method (Xu 2017). We run the generalized synthetic control method for the 179 treated SKUs and 747 control SKUs. We then compared the treated and synthetic units in the four quarters post-regulation (post normalizing the data). The results of GSCM are presented in Table below. A summary plot of regulated SKUs is presented in Figure below.

Main Analysis Results: Generalized Synthetic Control Method Estimates

Quarter # Negative treated effects (a) Average ATEs (b)
Jul 2013 - Sep 2013 90 out of 94 SKUs -1,727,272**
Oct 2013 - Dec 2013 88 out of 94 SKUs -1,648,094**
Jan 2014 - Mar 2014 87 out of 94 SKUs -1,620,187**
Apr 2014 - Jun 2014 83 out of 94 SKUs -1,450,358**

Results are based on 94 Treated SKUs with significant treatment effects
(a) Number of SKUs with negative (treated-synthetic) values in each quarter
(b) Average of ATE (Average Treatment Effects) of all 94 SKUs

Results of Generalized Synthetic Control Method

Notes: Vertical line (-) marks July 2013 (implementation of price regulation)