Mobility is one of the potential factors in contributing to the spread of COVID-19. We assume that every county will follow the state-wide policy accordingly. Our goal is to estimate the effect of state-wide policies on mobility signals at a county-level in this work to answer the following questions:
-
Do mobility signals decrease more because of the governemnt interventions or public reaction to case counts?
-
Once we can measure the effects of governemnt interventions on mobility, can we rank the effectiveness of the interventions at a county level?
-
Can we characterize different counties based on the known effects of the governemnt interventions on mobility?
-
Do mandatory policies seem to be more effective than recommended policies in reducing mobility?
This work has shown a way to estimate the effect of the emergency declaration on mobility during the pandemic. The emegency declaration tend to be more effective in reducing mobility in the areas that have large population, small percent of people in poverty, high percent of people with education backgrounds, low unemployment rate. One can apply the same strategy to estimate the causal effect of a single intervention so long there is no any other interventions happening concurrently.
Also, we provide a way to estimate the effects of interventions when the potential confounding variables are observed. Having accounted for case count signals and number of outpatient visits, we see that governemnt interventions can be more significiant in reducing the mobility in terms of restaurant visit. For example, in Allegaheny county in Pennsylvania, among all governemnt interventions, only mandatory stay at home order reduces restaurant visit significantly at 0.05 significant level. On the other hand, bar restriction and gathering restriction significantly reduce restaurant visit in Yolo county in California in comparison with other interventions.
We leave characterization for the ranks of the effect of interventions on county-level as a future work. Other regression methods such as non-parametric regression such as generalized additive models can also be used for further study. One should note that the effects of the interventions vary across counties in general. This study assumes that every county strictly follows all state-wide policies. It is encouraged to study a specific county in order to make a more precise conclusion on the effects of the intervention.
-
Lag analysis for case count signals, doctor visits on mobility across states. See 00_initial_exploration.md
-
Mobility signals across states under different interventions. See 00_initial_exploration.md
-
Regression discontinuity design to measure the effects of interventions on mobility. See 02_estimating_effect_of_intervention_via_RDD.md
-
Correlation between demographics and the effect of the emergency declaration. See 02_estimating_effect_of_intervention_via_RDD.md
-
Rank of the effectiveness of the intervention at a county-level. See 03_ranking_effect_of_interventions.md
-
Comparison between recommended policies and mandatory policies on a county-level [See 03a_allegaheny_county_recommend_mandate_policy.md]
-
Main report: a summary of the entire project See 04_main_report.md
-
/data/: this folder stores all data files. -
/code/: this folder contains all the codes to generate the files in/reports/and/html/folders. -
/historic/: this folder contains historic files, codes for implementing different ideas, but not officially used. -
/html/: this folder contains the main reports in HTML produced by R Markdown. -
/reports/: this folder contains the main reports in Markdown format.
-
Country-level:
-
State-level:
-
County-level: