Results

For each host market we compare the observed search index to its synthetic-control twin. With no game being played the two should match. When games are played, any gap is the estimated hosting effect, and difference-in-differences nets out the baseline gap. Select a single term to view it alone; select several and we show their average (each term indexed to its own pre-tournament baseline = 100 so big and small terms weigh equally). Intervals are 95% moving-block bootstrap CIs — robust to the week-to-week autocorrelation in search data (naïve standard errors would be too narrow).