Skip to main content
. 2021 Aug 19;27(9):1622–1628. doi: 10.1038/s41591-021-01487-3

Table 2.

Effect of treatment on movement outcomes

Mean (95% CI) OLS model Number of days × counties
Campaign Outcome Period High-intensity county Low-intensity county High-intensity county (95% CI) P value RI P value
Both campaigns Distance traveled From day − 3 to day − 1 −4.384 (−4.973,−3.796) −3.603 (−4.254, −2.952) −0.993 (−1.616, −0.371) 0.002 0.002 4,059
Both campaigns Share ever left home Thanksgiving (26 November) or Christmas (24–25 December) 72.326 (72.012, 72.639) 72.381 (72.092, 72.670) 0.030 (−0.361, 0.420) 0.881 0.911 2,017
Thanksgiving Distance traveled From day − 3 to day − 1 −6.082 (−6.822, −5.341) −5.320 (−6.113, −4.527) −0.924 (−1.785, −0.063) 0.035 0.030 2,072
Thanksgiving Share ever left home Thanksgiving (26 November) 71.308 (70.885, 71.731) 71.468 (71.071, 71.866) 0.012 (−0.438, 0.461) 0.959 0.966 689
Christmas Distance traveled From day − 3 to day − 1 −2.603 (−3.279, −1.927) −1.823 (−2.588, −1.057) −1.041 (−1.847, −0.235) 0.011 0.008 1,987
Christmas Share ever left home Christmas (24–25 December) 72.859 (72.507, 73.210) 72.852 (72.520, 73.185) 0.095 (−0.289, 0.479) 0.629 0.580 1,328

The control and treatment means at the county level and different periods, in addition to the estimate of the treatment coefficient in Eq. (1). Standard errors are clustered at the county level. 95% CIs are reported in parentheses. P values are based on a two-sided test. RI P values are computed using randomization inference, accounting for the two-stage design.