Yukui Zhou, The Economic Fallout of COVID-19: Uncertainty, Sectoral Disruptions, and Policy Responses (2025)

Summary

Conference paper by Yukui Zhou of Boston University, The Economic Fallout of COVID-19: Uncertainty, Sectoral Disruptions, and Policy Responses, in the Proceedings of the 3rd International Conference on Management Research and Economic Development, DOI 10.54254/2754-1169/177/2025.22435, published in 2025. The paper reviews key studies of the pandemic's economic impact: work on COVID-induced economic uncertainty measured by stock market volatility and business surveys, projecting an estimated 11% year-on-year contraction in U.S. GDP by the fourth quarter of 2020 with about half attributed to uncertainty; private sector data showing reduced spending by high-income individuals and losses for small businesses and low-wage workers; and forecasting of prolonged declines in employment, consumption and production. It closes with proposed recovery policies.

Summary drafted by a model from the document's text below and checked by script against that text before publication. It is a navigation aid, not a reading of what the document proves. Where AI is used

Cite this document

“Yukui Zhou, The Economic Fallout of COVID-19: Uncertainty, Sectoral Disruptions, and Policy Responses (2025).” Original at aemps.ewapub.com. Archived at Pandemic Darlings, https://pandemicdarlings.com/source-documents/yukui-zhou-the-economic-fallout-of-covid-19-uncertainty-sectoral-disru-4180a1d524/, SHA-256 6e113db266f727bef930be69ed51f9c6c5be8518baf984efd944e6d875eb3dfc.

Full text

Proceedings of the 3rd International Conference on Management Research and Economic Development
DOI: 10.54254/2754-1169/177/2025.22435

The Economic Fallout of COVID-19: Uncertainty, Sectoral
Disruptions, and Policy Responses
Yukui Zhou
Boston University, Boston, USA
zhouyk08@bu.edu

Abstract: The COVID-19 pandemic triggered one of the most severe global economic
downturns since the Great Depression, marked by unprecedented uncertainty and widespread
disruption. This paper examines the pandemic's economic impacts through key studies.
COVID-Induced Economic Uncertainty highlights how surging uncertainty, measured by
stock market volatility and business surveys, contributed to an estimated year-on-year U.S.
GDP declines by late 2020, with half attributed to uncertainty alone. The Economic Impacts
of COVID-19: Evidence from a New Public Database uses private sector data to show
reduced spending by high-income individuals, leading to revenue losses for small businesses
and job losses for low-wage workers, while finding traditional stimulus measures less
effective than direct financial support. Macroeconomic Forecasting in the Time of COVID-
19 projects prolonged declines in employment, consumption, and production, with a slow
recovery. Based on these findings, this paper proposes strategic policies to mitigate the crisis
and support economic recovery.

Keywords: COVID-19 Impact, Economic Uncertainty, Macroeconomic Forecasting

1. Introduction
An economic recession is generally characterized by a significant and widespread decline in
economic activity that persists for several months. Key indicators of a recession include reductions
in GDP, income, employment, industrial production, and retail sales. During such periods, consumer
spending, business investment, and overall economic confidence typically diminish, further
exacerbating the downturn.
The COVID-19 pandemic triggered one of the most severe global economic crises since the Great
Depression, causing unprecedented levels of uncertainty and disruption across various sectors. The
pandemic's far-reaching effects have been extensively studied, revealing profound impacts on
economic activity, labor markets, and consumer behavior. Several key studies have shed light on
these dynamics, we will talk about it in the Related Paperwork section below.
This paper will build on these insights to further explore the economic repercussions of the
COVID-19 pandemic, offering a comprehensive analysis of its impact and proposing targeted policy
measures to address the ongoing challenges. The following sections will explore the key findings,
methodologies, and implications of these studies, offering an in-depth analysis of the pandemic's
economic impact and potential recovery strategies.

© 2025 The Authors. This is an open access article distributed under the terms of the Creative Commons Attribution License 4.0
(https://creativecommons.org/licenses/by/4.0/).

164
Proceedings of the 3rd International Conference on Management Research and Economic Development
DOI: 10.54254/2754-1169/177/2025.22435

2. Related Paperwork
2.1. COVID-Induced Economic Uncertainty
This paper explores the economic impact of the COVID-19 pandemic, focusing on the surge in
uncertainty it brought about. The authors—Scott Baker, Nicholas Bloom, Steven Davis, and Stephen
Terry—highlight the unprecedented speed and scale of the crisis and outline three primary indicators
for measuring real-time uncertainty: stock market volatility (such as the VIX), newspaper-based
economic uncertainty indices, and business expectation surveys.
They demonstrate how these measures can quantify the uncertainty stemming from COVID-19
and project potential macroeconomic impacts, such as an estimated 11% year-on-year contraction in
the U.S. GDP by the fourth quarter of 2020.[1] About half of this projected decline is attributed to
uncertainty alone, illustrating the significant indirect economic consequences of the pandemic beyond
immediate demand or supply shocks.
2.2. The Economic Impacts of COVID-19: Evidence from a New Public Database Built Using
Private Sector Data
Written by Raj Chetty, John N. Friedman, Nathaniel Hendren, Michael Stepner, and the Opportunity
Insights Team, this paper introduces a real-time database using private data sources—such as credit
card spending, payroll records, and job postings—to track economic activity by geography, industry,
and income. The authors compiled anonymized data on key indicators like consumer spending,
business revenues, and employment, disaggregated by ZIP code, industry, and demographic segments,
applying privacy protocols, data cleaning, and validation against public benchmarks [2].
Findings reveal that high-income individuals sharply reduced spending in sectors requiring in-
person interaction, leading to significant revenue drops for small businesses in affluent areas and
persistent job losses for low-wage workers there. Instead of relying on expert advice, consumers
increasingly turn to fellow customers when choosing among products, a trend facilitated by the
emergence of social media and the associated creation and consumption of user-generated content ,
for instance, finds that 82% of US Internet consumers report to be directly influenced by peer reviews
in their purchasing decisions.[3] The study suggests that direct financial support, such as extended
unemployment benefits, may be more effective than traditional stimulus during health crises and
demonstrates the utility of private-sector data for real-time economic monitoring and policy
evaluation.
2.3. COVID-19 and The Macroeconomic Effects of Costly Disasters
Written by Sydney Ludvigson, Sai Ma, and Serena Ng, this study investigates the broader economic
impacts of large-scale disasters, focusing on the COVID-19 pandemic as a prolonged, global, multi-
sector disruption unlike typical, localized natural disasters like hurricanes or earthquakes. While such
events typically produce short-term local economic impacts, COVID-19 has simultaneously disrupted
global markets, labor, and production for an extended period. Leveraging 40 years of historical U.S.
disaster data, the authors develop a "costly disaster" (CD) series to quantify the pandemic's economic
damage. Using a vector autoregression (VAR) model, they predict COVID-19's impact across sectors,
assessing industrial production, service employment, unemployment claims, and flight departures
based on prolonged relief costs and lockdown durations.
Key findings include an estimated 20% decline in industrial production over 12 months, potential
39% job losses in service employment impacting up to 55 million jobs, an unprecedented rise in
unemployment claims, and a significant, sustained drop in flight departures, signaling broader
declines in travel and tourism.[4] The study highlights those longer shocks delay recovery and

165
Proceedings of the 3rd International Conference on Management Research and Economic Development
DOI: 10.54254/2754-1169/177/2025.22435

emphasizes the necessity of sustained policy support and targeted aid to stabilize affected sectors,
suggesting that pandemic-driven disruptions could influence labor markets and business operations
well beyond the immediate crisis.
2.4. Macroeconomic Forecasting in the Time of COVID-19
Written by Giorgio E. Primiceri and Andrea Tambalotti, this paper presents a forecasting model for
the economic impact of COVID-19 in the United States, highlighting the unique assumptions needed
to predict such an unprecedented shock. The authors note that COVID-19 challenges standard
economic models, requiring scenario-based assumptions beyond typical disruptions. By synthesizing
COVID-19 as an exogenous shock within macroeconomic frameworks, the authors use historical data
combined with projections of the virus's spread and economic impacts. They develop three
forecasting scenarios: a baseline where the pandemic fades by late 2020, an optimistic scenario ending
by September 2020, and a pessimistic scenario with a second wave peaking in early 2021.
Findings indicate significant economic decline: in the baseline, employment falls by over 20%
with a slow, multi-year recovery; consumption and industrial production sharply drop and remain
below pre-COVID levels across scenarios; unemployment peaks at over 20% in the baseline, with
prolonged high rates in the pessimistic scenario; and inflation remains muted, with core PCE inflation
temporarily turning negative due to demand-side constraints.[5] This analysis underscores that
economic recovery depends heavily on the pandemic's trajectory, suggesting policy actions to
stabilize demand and support labor markets are essential for a robust recovery.
2.5. Discussion
The COVID-19 pandemic triggered an unprecedented surge in economic uncertainty, measured
through indicators such as stock market volatility, media coverage, and business surveys, which
significantly contributed to economic contraction. Projections indicated substantial declines in GDP,
employment, and industrial production, with one study estimating an 11% year-on-year contraction
in U.S. GDP by late 2020, half of which was attributed to uncertainty alone.
The studies emphasize the need for sustained and targeted policy measures to stabilize affected
sectors and labor markets, particularly given the prolonged nature of the crisis. Additionally, the
pandemic underscored the importance of real-time data, such as credit card spending and payroll
records, for monitoring economic activity and evaluating policy effectiveness. However, forecasting
the economic trajectory of COVID-19 required scenario-based assumptions due to the unprecedented
nature of the shock, with recovery heavily dependent on the pandemic's course.
3. Data and Correlation Analysis
The correlation analysis presented in the document examines the relationship between economic
mobility data and COVID-19 case data in the United States. The economic mobility data, sourced
from indicators such as retail and recreation, grocery and pharmacy, parks, transit stations,
workplaces, and residential activity, reflects changes in mobility patterns compared to a pre-pandemic
baseline. These mobility metrics are derived from anonymized location data, and provide insights
into how people's movement and economic activities have shifted in response to the pandemic. On
the other hand, the COVID-19 case data represents daily confirmed cases, which serve as a proxy for
the spread of the virus.
Over time, the economic mobility data shows fluctuations in activity across different sectors. For
example, retail and recreation, transit stations, and workplaces experienced declines in activity,
reflecting reduced public interaction and shifts to remote work during the pandemic. Conversely,
residential activity increased as people spent more time at home. These trends reflect public health

166
Proceedings of the 3rd International Conference on Management Research and Economic Development
DOI: 10.54254/2754-1169/177/2025.22435

interventions, including lockdowns, social distancing, and remote work policies, aimed at controlling
the spread of COVID-19.
The correlation analysis was conducted by plotting daily confirmed COVID-19 cases against
percent changes in mobility from the baseline for each sector. The results reveal mostly weak
relationships between mobility and case counts. For instance:
1. Retail and recreation showed a weak negative correlation, suggesting that as mobility in these
areas decreased, there was a slight reduction in cases, though the relationship was not strong.
2. Grocery and pharmacy exhibited a weak positive correlation, indicating a slight increase in cases
with higher mobility, but the trend was not pronounced.
3. Parks also showed a weak positive correlation, with a small increase in cases as park activity rose,
though the data points were widely dispersed.
4. Transit stations and workplaces both displayed weak negative correlations, implying that reduced
activity in these areas slightly coincided with fewer cases, but the relationships were not robust.

Figure 1: Transit Stations Percent Change from Baseline VS Daily Confirmed Cases

In conclusion, the analysis indicates that while certain trends between mobility and COVID-19
cases can be observed, the relationships are generally weak and not clearly defined. This may show
up the complex interaction of various factors affecting case numbers, such as public health measures,
vaccination rates, and changes in behavior, which mobility data cannot fully account for. The findings
underestimate the difficulty in making definitive conclusions about the direct influence of mobility
on the spread of COVID-19, highlighting the importance of detailed analyses that incorporate
additional variables.
4. Predictive models
4.1. Variables
The dataset used in this analysis contains annual economic and financial data from 1984 to 2023
including key variables as:
1. GDP (Billions of USD), also treated as target variable.

167
Proceedings of the 3rd International Conference on Management Research and Economic Development
DOI: 10.54254/2754-1169/177/2025.22435

2. Unemployment Rate (%), which affects GDP by reducing consumer spending, labor productivity,
and business investment, as fewer people earning income leads to lower demand and underutilized
resources.
3. Housing Price Index & Producer Price Index (PPI), which are indicators of the housing market's
health and production costs.
4. Population (Millions), which affects GDP in shaping economic activity, as a larger population
generally leads to increased consumption and production levels.
5. Stock Market Capitalization, which represents the total valuation of companies, serving as a gauge
of investor confidence and economic stability.
6. 1-Year Treasury Rate (%) & Corporate Bonds Spread (%), which influence investment decisions
and economic growth.
7. Net Exports (USD Billions), which contribute to GDP by reflecting the trade balance.
8. Savings Rate (%), which measures the portion of income saved rather than spent, influences
consumption behavior.

Figure 2: GDP between 1984 to present

Based on their information regression scores, which measure the dependency between each feature
and the target variable, GDP, the top 8 features, including Population (Millions), Housing Price Index,
PPI, Net Exports, 1-Year Treasury Rate, Savings Rate, Corporate Bonds Spread, and Stock Market
Cap, was selected. These variables will be used in the following analysis with chosen models.
4.2. Models
Several models were employed to forecast GDP, the target variable. First, the data was split into a
70:30 ratio for the training and testing set, and the model was trained on the training data to forecast
GDP for the testing period. The first model used was the Vector Autoregression Model (VAR Model).
This is a statistical model to capture linear interdependencies among multiple time series variables. It
works by predicting each variable based on its own historical values as well as the past values of other
related variables in the system. This model is an effective one for analyzing linear relationships and
dependencies among variables over time, which is widely used for studying time series data. By using

168
Proceedings of the 3rd International Conference on Management Research and Economic Development
DOI: 10.54254/2754-1169/177/2025.22435

Granger causality tests, GDP, Unemployment Rate, Housing Price Index, and Savings Rate were
selected as the key variables for the VAR model. RMSE (Root Mean Squared Error) and MAE (Mean
Absolute Error) were used for the model’s performance.
Besides the VAR model, there are also several machine learning models applied in this task. These
included:
1. Support Vector Regression (SVR), which is an adaptation of the support vector machine algorithm
for regression. It aims to find the hyperplane that best fits the data while providing a certain margin
of error.
2. Decision Tree Regressor, which is a tree-based model that divides data into branches based on the
values of input features,with each division creating a structure of decisions.
3. XGBoost, which is a boosting algorithm that sequentially builds decision trees. It focuses on
correcting errors in each new tree from the previous ones, and incorporating regularization
techniques to prevent overfitting.
4. LightGBM, which is another boosting algorithm similar to XGBoost but optimized for speed and
efficiency, leading to fast training times by using a growth strategy.
5. Long Short-Term Memory (LSTM), which is a specialized type of recurrent neural network (RNN)
designed to work with sequential data, allowing information to be stored or deleted for long periods
of time.
For all models mentioned above, all variables except year and GDP were used as features and the
top 8 features were selected using SelectKBest and mutual information regression. Each model was
trained on the scaled training data and hyperparameter tuning was performed using GridSearchCV to
optimize the performance. RMSE (root mean square error) and MAE (mean absolute error) were used
to evaluate the performance of these models.
5. Prediction results of US
After evaluating all the models, XGBoost was selected to be the best prediction model due to its
performance among all evaluation metrics. XGBoost has the lowest RMSE (6093.45) and MAE
(5175.55), indicating that the predictions of XGBoost were the closest to the actual GDP values, with
the smallest average errors.
The hyperparameters of the XGBoost model were optimized using GridSearchCV, and
learning_rate: 0.1, max_depth: 7, n_estimators: 100 were identified as the optimal parameters, which
improved the predictive accuracy of the model. When the predictions of the XGBoost model were
plotted against the actual GDP values for the testing period (2012-2023), the predictions were nearly
consistent with the actual trend, providing reliability and accuracy of the model.
6. Limitations and Further Suggestions
6.1. Limitations
Although this analysis is visually strong and insightful, there are still some limitations to be noticed.
First, the dataset covers the period only from 1984 to 2023, and even if it provides a solid historical
perspective, it may miss recent economic changes and future trends. Economic conditions change
rapidly, and models may perform relatively differently when applied to certain environments. Second,
this dataset only uses annual data, which may not capture short-term fluctuations or seasonal trends
for more detailed forecasts. Another limitation is the model itself. For example, VAR models only
apply linear relationships between variables, but may not fully capture the complexity that exists in
real economic systems. Machine learning models such as XGBoost are known as powerful, but their

169
Proceedings of the 3rd International Conference on Management Research and Economic Development
DOI: 10.54254/2754-1169/177/2025.22435

performance is still limited by the quality and quantity of data, which remains a risk of overfitting,
especially when applied to out-of-sample data.
6.2. Further Suggestions
Based on these limitations, there are several suggestions for further research to improve the analysis.
Initially, incorporating additional data can increase the predictive power of the model. This may
include variables such as the World Economic Indicator, the Technological Innovation Index, and the
Geopolitical Risk Index to grab a larger range of factors that affect GDP. Using higher frequency
data, such as quarterly or monthly data, can also improve the model's ability to forecast short-term
trends and economic fluctuations. Exploring more advanced models is also another way of
improvement. For example, trying Transformer-based models or hybrid models, such as integrating
VAR with XGBoost, may give better results. Additionally, investigating methods that combine
predictions from multiple models may improve overall accuracy and robustness. Feature engineering
can be enhanced by incorporating techniques as well, such as polynomial features, interaction terms,
and domain-specific transformations for interactions between variables.
By incorporating additional data, exploring advanced models, improving feature engineering, and
enhancing evaluation techniques, the accuracy and robustness of GDP forecasts can be improved
effectively. These advances will not only benefit economic forecasting, but also provide policymakers
and businesses with more reliable tools for decision-making in an ever-changing economic
environment.
7. Conclusion
Forecast models provide a valuable tool for understanding and predicting economic trends, but they
need to evolve to better capture the complexity of a rapidly changing world. The global economy is
deeply interconnected. Forecasting models should evolve by integrating AI-driven analytics, diverse
data streams, and cross-sector collaboration to build resilient systems capable of weathering shocks
and enabling sustainable recovery.
References
[1] Scott R. Baker, Nicholas Bloom, Steven J. Davis, Stephen J. Terry, COVID-INDUCED ECONOMIC UNCERTAIN
TY.
[2] Hailiang Chen, Prabuddha De, Yu (Jeffrey) Hu, and Byoung-Hyoun Hwang, WISDOM OF CROWDS: THE VALUE
OF STOCK OPINIONS TRANSMITTED THROUGH SOCIAL MEDIA.
[3] Private Sector Data*, Raj Chetty, John N. Friedman, Nathaniel Hendren, Michael Stepner, and the Opportunity
Insights Team, The Economic Impacts of COVID-19: Evidence from a New Public Database Built Using.
[4] Sydney C. Ludvigson, Sai Ma, and Serena Ng, COVID-19 and The Macroeconomic Effects of Costly Disasters.
[5] Giorgio E. Primiceri Andrea Tambalotti, Macroeconomic Forecasting in the Time of COVID-19*.

170

File and source

File: zhou-2025-the-economic-fallout-of-covid-19-uncertainty-sectoral-disruptions-and-policy-res.pdf
Size: 767,820 bytes
SHA-256: 6e113db266f727bef930be69ed51f9c6c5be8518baf984efd944e6d875eb3dfc
Our copy: zhou-2025-the-economic-fallout-of-covid-19-uncertainty-sectoral-disruptions-and-policy-res.pdf
Original: aemps.ewapub.com