Notes: Nearest neighbor propensity score matching was done using Stata 12 and the psmatch2 package developed by Leuven and Sianesi (2003). Matching is done using a caliper of 1/4 of the standard deviation of the estimated propensity scores and by allowing matches only within narrowly defined year-sector-export value bracket combinations. Export values are expressed in constant euro’s and are deflated using sector level price indices for gross output. The balancing property condition, requiring absence of statistically significant differences between mean values of the matching characteristics of the treatment and the control group is fully satisfied in all instances. The bias-corrected confidence intervals are constructed by bootstrapping the ATT with 1,000 replications. *p < 0.05