Using Causal Inference to Explore Government Policy Impact on Computer Usage

Mingjia Zhu, Lechuan Wang, Julien Sebot, Bijan Arbab, Babak Salimi, Alexander CloningerarXiv 2025 · OpenReview

Abstract

We explore the causal relationship between COVID-19 lockdown policies and changes in personal computer usage. In particular, we examine how lockdown policies affected average daily computer usage, as well as how it affected usage patterns of different groups of users. This is done through a merging of the Oxford Policy public data set, which describes the timeline of implementation of COVID policies across the world, and a collection of Intel's Data Collection and Analytics (DCA) telemetry data, which includes millions of computer usage records and updates daily. Through difference-in-difference, synthetic control, and change-point detection algorithms, we identify causal links between the increase in intensity (watts) and time (hours) of computer usage and the implementation of work from home policy. We also show an interesting trend in the individual's computer usage affected by the policy. We also conclude that computer usage behaviors are much less predictable during reduction in COVID lockdown policies than during increases in COVID lockdown policies.

Introduction

While the shift toward remote work and digital entertainment during the COVID-19 pandemic is well-documented anecdotally, quantifying the exact impact of government policies on hardware usage has remained a challenge. This research, a collaboration between UC San Diego and Intel Corporation, moves beyond simple correlation to establish a causal link between workplace closure policies and personal computer (PC) usage. By analyzing massive hardware telemetry datasets, the study identifies how specific policy levers changed not just the duration of PC use, but also the intensity and nature of user behavior.

Data and Methodology

Datasets

The researchers integrated two primary data sources to conduct their analysis:

  1. Oxford Covid-19 Government Response Tracker (OxCGRT): This public dataset provides a standardized timeline of policy strictness across more than 50 countries. The study specifically focused on the workplace closure metric, which is categorized on a scale from 0 to 3.
The Oxford Policy scale used to categorize the strictness of workplace closure mandates.
  1. Intel DCA (Data Collection & Analytics) Telemetry: This anonymized dataset encompasses millions of worldwide devices, providing hardware-level metrics such as C0 state time (active usage hours), CPU power (watts), and device specifications (e.g., Notebook, Desktop, 2-in-1).

Statistical Framework

To isolate the effect of policy from general pandemic-era trends, the authors employed three rigorous statistical methods:

  • Difference-in-Differences (DiD): This method compared "treated" regions (those with strict lockdowns) against "control" regions (those that remained open). For example, by comparing California to Taiwan, researchers could filter out global noise to see the specific impact of the local lockdown.
  • Synthetic Control Method (SCM): SCM allowed for the creation of a "weighted average" of multiple control countries to better mirror the pre-lockdown behavior of a treatment country, providing a robust counterfactual.
  • Offline Change-point Detection: Algorithms were used to identify sudden "breakpoints" in time-series data, such as CPU power consumption, to see if these shifts aligned with policy enactment dates.
A Difference-in-Differences (DiD) visualization comparing average daily computer usage in California (treated) against Taiwan (control) around the March 2020 lockdown.

Key Findings

Quantifiable Usage Spikes

The study found that workplace closures caused a statistically significant increase of 1.2 to 2.3 hours in daily PC usage. This was accompanied by an increase in CPU power consumption (~1 watt for notebooks and ~2 watts for desktops), indicating that users were engaging in more computationally intensive tasks like video conferencing and gaming.

The Mobility Paradox

Interestingly, the most significant usage surge occurred in "lighter and more mobile" devices, specifically 2-in-1 convertible tablets (an increase of ~2.2–2.3 hours). Despite being confined to their homes, users gravitated toward flexible, portable form factors rather than traditional desktops, which saw the lowest relative jump (~1.4–1.5 hours).

Asymmetry and "Sticky" Habits

A critical insight from the research is the asymmetry of policy impact. While the implementation of lockdown policies led to a rapid and predictable spike in usage, the easing of these policies did not result in a corresponding return to baseline. This suggests a "sticky" shift in digital habits; once users integrated PCs more deeply into their daily lives for work and leisure, those behaviors persisted even after restrictions were lifted.

Conclusion and Future Work

This research demonstrates that government workplace policies were a direct causal driver of increased PC usage during the pandemic. However, the study is limited by the nature of the telemetry data, which relies on opt-in users (approximately 40% of the Intel base) and may lean toward tech-savvy demographics. Future work will aim to disentangle the effects of other specific measures, such as transit restrictions and stay-at-home orders, to further refine our understanding of how policy shapes digital life.

Summary updated

Abstract

We explore the causal relationship between COVID-19 lockdown policies and changes in personal computer usage. In particular, we examine how lockdown policies affected average daily computer usage, as well as how it affected usage patterns of different groups of users. This is done through a merging of the Oxford Policy public data set, which describes the timeline of implementation of COVID policies across the world, and a collection of Intel's Data Collection and Analytics (DCA) telemetry data, which includes millions of computer usage records and updates daily. Through difference-in-difference, synthetic control, and change-point detection algorithms, we identify causal links between the increase in intensity (watts) and time (hours) of computer usage and the implementation of work from home policy. We also show an interesting trend in the individual's computer usage affected by the policy. We also conclude that computer usage behaviors are much less predictable during reduction in COVID lockdown policies than during increases in COVID lockdown policies.
BibTeX
@misc{zhu2025using,
  author = {Mingjia Zhu and Lechuan Wang and Julien Sebot and Bijan Arbab and Babak Salimi and Alexander Cloninger},
  title = {{Using Causal Inference to Explore Government Policy Impact on Computer Usage}},
  year = {2025},
  eprint = {2503.09957},
  archivePrefix = {arXiv},
  url = {https://arxiv.org/abs/2503.09957},
}
Figure