A country comparison of place-based activity response to COVID-19 policies

Grant McKenzie, Benjamin AdamsarXiv 2020 · arXiv

TL;DR

This paper quantifies the relationship between government COVID-19 policies and actual human behavior across 129 countries using Google mobility data and the Oxford Stringency Index. The researchers found that while mobility generally tracks policy closely—especially for transit and residential activities—there is a measurable global average lag of 2.4 days between policy enactment and public response, with significant variations tied to a country's level of economic development.

What Problem Does This Paper Address?

When the COVID-19 pandemic hit, governments worldwide scrambled to implement "non-pharmaceutical interventions" (lockdowns, school closures, travel bans). However, a policy on paper doesn't always translate to immediate action on the ground.

Prior to this study, most mobility research focused on single nations (primarily the U.S.) or looked at movement patterns in isolation. This paper addresses the global gap by asking: How does the magnitude and timing of government action correlate with specific place-based activities (like going to work vs. the grocery store) across different cultures and political systems? It seeks to understand if people respond to the law, the severity of the outbreak, or other socio-economic factors.

Methods

The authors developed three primary assessment measures to compare policy against behavior:

  1. Similarity in Magnitude: They used Cosine Similarity and Pearson Correlation to compare the "shape" of the government policy curve against the mobility curves. To make these comparable, they inverted mobility categories that were expected to decrease (like Retail) so they would align with the increasing Stringency Index.
  2. Temporal Lag: Using signal processing techniques (cross-correlation), they calculated the displacement in days between a policy change and the resulting shift in community activity.
  3. Subregional Variability: For countries with available data, they measured how much activity patterns varied between different states or provinces, providing a proxy for how "unified" a national response was.

Finally, they used Kendall’s $\tau$ to correlate these findings with global socio-economic indices.

Datasets

  • Google Community Mobility Reports: Aggregated, anonymized data showing daily percentage changes in activity across six categories: Grocery & pharmacy, Parks, Transit stations, Retail & recreation, Residential, and Workplaces.
  • Oxford Coronavirus Government Response Tracker (OxCGRT): Specifically the Stringency Index, a 0–100 score representing the severity of 13 policy indicators (e.g., school closures, travel bans).
  • Global Indices: The UN Human Development Index (HDI), World Bank Development Indicators (WDI), the Corruption Perception Index, and the World Press Freedom Index.

Key Insights and Ideas

  • Policy vs. Reality: The correlation between activity patterns and government policy ($\tau = 0.631$) was actually higher than the correlation between activity and the number of confirmed cases ($\tau = 0.610$) or deaths ($\tau = 0.569$). This suggests people were responding more to government mandates than to the perceived immediate threat of the virus itself.
  • Activity Sensitivity: Not all places are affected equally. Transit and residential patterns are highly sensitive to policy changes, while "Parks" and "Grocery & Pharmacy" show high volatility and lower similarity to policy, likely due to weather dependencies and the "essential" nature of food shopping.
  • The Development Gap: There is a significant link between a country’s development and its response time. Countries with lower HDI ranks, lower life expectancy, and lower GDP tended to have longer lags between policy enactment and behavioral shifts.
  • Decentralization Matters: The study highlights the difference between unified responses (like New Zealand) and decentralized ones (like the U.S.). The U.S. showed some of the highest subregional variability, reflecting the significant authority held by state governors compared to the federal government.

Key Results

The researchers found that Transit stations and Residential activities were the most consistent indicators of policy adherence across all nations.

Table 1: Average cosine similarity comparing each place-based activity category to the stringency index across all countries.
  • Average Lag: The global mean lag was -2.4 days, meaning behavior typically shifted a few days after a policy was announced. However, some countries like South Africa showed substantial lags (mean 8.5 days).
  • High Adherence: Countries like Venezuela, Oman, and New Zealand showed very high similarity between policy and response.
Figure 2: Highest average cosine similarity between stringency index and five place-based activity patterns.
  • Outliers: Sweden and South Korea appeared as outliers in similarity, but for different reasons. Sweden's low similarity reflected its unique "herd immunity" approach with low stringency, while South Korea's low similarity in this specific dataset was due to their successful early intervention occurring before the study's data window began.

Limitations and Future Work

  • Data "Black Box": The authors acknowledge that Google’s mobility data lacks transparency regarding sample sizes and exact data collection methodology.
  • Confounding Variables: Weather is a major factor for "Parks" activity that wasn't fully controlled for. Additionally, the study did not account for partisan politics, which heavily influenced compliance in several countries.
  • Future Directions: The researchers suggest that the next critical phase of study is analyzing the "relaxation" phase—determining if residents respond to the lifting of restrictions with the same patterns they showed during the initial lockdown.

Summary updated

Abstract

The emergence of the novel Coronavirus Disease in late 2019 (COVID-19) and subsequent pandemic led to an immense disruption in the daily lives of almost everyone on the planet. Faced with the consequences of inaction, most national governments responded with policies that restricted the activities conducted by their inhabitants. As schools and businesses shuttered, the mobility of these people decreased. This reduction in mobility, and related activities, was recorded through ubiquitous location-enabled personal mobile devices. Patterns emerged that varied by place-based activity. In this work the differences in these place-based activity patterns are investigated across nations, specifically focusing on the relationship between government enacted policies and changes in community activity patterns. We show that people's activity response to government action varies widely both across nations as well as regionally within them. Three assessment measures are devised and the results correlate with a number of global indices. We discuss these findings and the relationship between government action and residents' response.
BibTeX
@misc{mckenzie2020country,
  author = {Grant McKenzie and Benjamin Adams},
  title = {{A country comparison of place-based activity response to COVID-19 policies}},
  year = {2020},
  eprint = {2005.08738},
  archivePrefix = {arXiv},
  url = {https://arxiv.org/abs/2005.08738},
}

Loading related content...

Figure