Table 3. Alternative scenario with different testing efficiency.

	30%	50%	70%	90%
Panel A: Until 3.15
Scenario 1: No lockdown ever
No. of cases	1.55×107	4.28×106	2.58×106	2.05×106
simulated cases/actual cases	41.5	11.5	6.9	5.5
Scenario 2: Lockdown since Jan 23
No. of cases	6.28×105	4.70×105	4.40×105	4.27×105
simulated cases/actual cases	1.74	1.30	1.22	1.18
Panel B: Until 7.15
Scenario 1: No lockdown ever
No. of cases	6.95×108	7.72×106	2.68×106	2.07×106
simulated cases/actual cases	1866.9	19.4	7.2	5.6
Scenario 2: Lockdown since Jan 23
Cases	2.22×108	5.01×105	4.51×105	4.41×105
simulated cases/actual cases	595.14	1.38	1.22	1.18
Scenario 3: Lockdown from Jan 23 to Mar 15, no lockdown after Mar 15
No. of cases	2.89×108	4.97×105	4.40×105	4.27×105
simulated cases/actual cases	799.59	1.38	1.22	1.18

This table presents alternative counterfactual scenarios with different detection rate after March 15. We let the detection rate to increase from the initial level before January 23, gradually to the designated level from January 23 to March 15. We simulate different lockdown scenarios and set the inner city mobility to follow the level on the same lunar calendar date in 2019 in the counterfactual case of no lockdown. The three scenarios are (1) No lockdown ever from January 10 to July 15; (2) Lockdown from January 10 to July 15; (3) Lockdown from January 10 to March 15, no lockdown from March 15 to July 15. Results in this table are the mean value derived from 200 simulations. After March 15 when the daily mobility data were no longer available, we use the average migration and inner-city flow intensity from March 9 to March 15 (one week before March 15) as a proxy.