check the file

1 | P a g e

COVID-19 Across Two States: Economic and Health Impacts of Policy Choices

2 | P a g e

Executive Summary:

COVID-19, caused by the SARS-CoV-2 virus, created an unprecedented global health crisis beginning in late 2019.

Governments worldwide responded with varying degrees of intervention—ranging from minimal restrictions to

strict lockdowns and mask mandates—to slow transmission and protect vulnerable populations. These policy

choices, often influenced by political ideology, shaped both health outcomes and economic impacts. Understanding

the relative effectiveness of these interventions offers valuable insight for future public health decision-making.

This study compares pandemic policy responses and outcomes between politically conservative (“red”) and liberal

(“blue”) U.S. states, with a focus on Texas and New York along with a comparison of two similarly sized counties.

Using county-level data on COVID-19 cases, deaths, and demographic variables from 2020–2021, we assessed how

mitigation measures such as lockdown duration, mask mandates, and vaccination rates correlated with mortality.

Visual analyses, including histograms, scatterplots, heatmaps, and time-series graphs, reveal clear differences in

both the scale and pace of COVID-19 spread across the two states.

Findings show that while early pandemic death rates were comparable between red and blue states, by 2021 blue

states exhibited significantly lower mortality. This difference aligned with stronger and longer mitigation policies,

higher vaccination rates, and greater local autonomy to enforce restrictions. In contrast, red states—where man-

dates were fewer or shorter—showed outcomes more influenced by demographic and biological risk factors such as

age and obesity. Public attitudes toward government authority and perceived risk also affected compliance, amplify-

ing these disparities.

This comparative study underscores the need for evidence-driven, nonpartisan pandemic policy frameworks that

integrate public health best practices with economic and social sustainability. For our stakeholders, the results sug-

gest that early adoption of flexible, but proactive measures can mitigate both health and economic risks in future

outbreaks.

3 | P a g e

Table of Contents 1. Problem Description ........................................................................................................................................... 5

1.1. Stakeholders ......................................................................................................................................... 5

1.2. Caveats & nuance .................................................................................................................................. 9

1.3. Political label isn’t the sole driver Health infrastructure, population density, demographics, and timing of

introduction of the virus all matter. Some red states with large rural populations had different dynamics than red

states with big metros. .................................................................................................................................... 9

1.3..1. Policy vs compliance Having a mandate doesn’t guarantee enforcement or full compliance; cultural norms,

public buy-in, and enforcement capacity matter. ....................................................................................................... 9

1.3..2. Temporal shifts Many states (both red and blue) loosened restrictions by 2022–2023; the distinguishing

features are often in when and how aggressively they moved rather than permanent differences. ........................... 9

1.3..3. Data limitations Studies of state-level aggregated outcomes may suffer from unmeasured confounders;

some papers found that the political affiliation of governors was not a strong predictor of outcomes after accounting

for other variables. (9) ............................................................................................................................................... 9

1.4. Purpose & Questions ............................................................................................................................. 9

1.5. Data & Methods .................................................................................................................................... 9

1.6. Key Findings ........................................................................................................................................ 10

1.7. Interpretation & Limitations ................................................................................................................ 10

2. Data Collection and Data Quality ...................................................................................................................... 10

2.1. Data Overview ..................................................................................................................................... 10

2.2. Data Quality ........................................................................................................................................ 12

2.3. Descriptive Statistics ............................................................................................................................ 12

• Texas ................................................................................................................................................................. 12

• New York ........................................................................................................................................................... 13

2.4. Data Relationships ............................................................................................................................... 13

• Texas Visualizations ........................................................................................................................................... 14

• New York Visualizations ..................................................................................................................................... 17

• Comparison Analysis .......................................................................................................................................... 21

4 | P a g e

3. Data Exploration ............................................................................................................................................... 24

4. Modeling and Evaluation .................................................................................................................................. 25

5. Recommendations............................................................................................................................................ 25

6. Conclusion ........................................................................................................................................................ 26

7. List of References ............................................................................................................................................. 26

8. Appendix .......................................................................................................................................................... 28

8.1. Student Contributions ......................................................................................................................... 28

8.2. AI Use .................................................................................................................................................. 28

5 | P a g e

1. Problem Description COVID-19 is a highly contagious disease caused by the SARS-CoV-2 virus. It emerged in late 2019 and spread rapidly

across the globe, prompting a global pandemic. Symptoms range from mild (fever, cough, fatigue) to severe (respira-

tory failure, death), particularly in vulnerable populations such as the elderly or immunocompromised.

Social distancing involves keeping physical space between individuals to slow the spread of the virus. This includes

avoiding large gatherings, staying 6 feet apart in public, and minimizing unnecessary travel.

Flattening the curve refers to slowing the rate of infection to prevent overwhelming the healthcare system. The goal

is to reduce the number of concurrent infections so that hospitals, ICU beds, and medical staff can handle the pa-

tient's load.

Understanding how these and other viruses spread and how various healthcare policies and clinical systems cope

with outbreaks enables policymakers and health officials to make more informed decisions. Data on infection rates,

mortality, healthcare access, and public compliance with policies can reveal the effectiveness of government inter-

ventions, medical resource needs, and disparities in outcomes across demographics.

Across the world policy makers worked to promote a safe environment for citizens but utilized widely varying meth-

ods to do so. Some implemented very few rule changes while others implemented very stringent restrictions on

wide swaths of their population. In the years since the Covid-19 outbreak, researchers have looked at successes and

failures of these policies, how they impacted the health of individuals as well as the economic impact.

Our client, a conservative-leaning mid-sized county in New York state, asked us to provide information on possible

policy benefits and ramifications related to restrictions that the state imposed during the initial stages of the virus

outbreak. They were hoping to pre-define future pandemic policies and to get public legislative initiatives in front of

the voters for their input and approval.

As mentioned, several research documents are available discussing the differences between liberal and conservative

leaning states and we look to that research, as well as our own, to determine which policies might have benefited

the populace and which might have been less beneficial.

1.1. Stakeholders Our research stakeholders include:

• County-level public health and policymakers in New York

Our research includes decisions by state-level policy makers in Texas and New York.

Questions that stakeholder have an interest in learning from our research:

• How did the timing and duration of mask mandates, lockdowns, and mobility restrictions correlate with

county-level COVID-19 death rates in Texas and New York?

6 | P a g e

• How did vaccination uptake and public mobility trends interact to influence infection and mortality rates

after restrictions were lifted?

• How did vaccination uptake and public mobility trends interact to influence infection and mortality rates

after restrictions were lifted?

Decisions that the Stakeholders Can Make:

• Determine whether more aggressive early interventions (as seen in liberal states) led to better outcomes.

• Inform future public health crisis responses by analyzing which strategies were most effective.

• Reallocate or reinforce medical infrastructure in counties with poor outcomes.

• Tailor messaging and health policy by understanding behavioral trends (e.g., compliance inferred through

mobility data).

These policy makers are responsible for implementing and adjusting pandemic response strategies including stay-at-

home orders, business closures, vaccination rollouts, and mask mandates.

By evaluating the differing pandemic strategies of conservative and liberal states using real-world outcomes, health

departments can make evidence-based decisions for future outbreaks—potentially saving lives and minimizing eco-

nomic disruption.

Some policy decisions that may or may not have made a difference:

• Social distancing - involves keeping physical space between individuals to slow the spread of the virus. This

includes avoiding large gatherings, staying 6 feet apart in public, and minimizing unnecessary travel.

• Curfew and travel restrictions

• Workplace, school, shopping, vacation location, public area closures

• Mandatory vaccination

• Mandatory mask wearing

• Local versus state-wide policies

7 | P a g e

Comparing “conservative-leaning” (often “red”) states and “liberal-leaning” (often “blue”) states during COVID-19

reveals a number of recurring patterns and divergences in how public health policy was executed. (Note: these are

general trends and there is variation within each group.) Below are several domains of COVID policy and contrast

how blue vs red states typically behaved, with supporting evidence and caveats.

Policy do-

main More common in liberal / blue states

More common in conservative / red

states

Early/strict

lockdowns &

business clo-

sures

Blue states were more likely to adopt early,

stricter stay-at-home orders and prolong re-

strictions on nonessential businesses and gath-

erings.

Red states often delayed or softened early

restrictions, with more tolerance for keep-

ing businesses open or reopening sooner.

Mask man-

dates & en-

forcement

Blue states more often implemented statewide

mask mandates, including public spaces, indoor

settings, and maintained them longer.

Some red states either avoided statewide

mandates or removed them earlier; some

also banned local mask mandates.

Vaccine man-

dates or

“passports”

More willing to require proof of vaccination for

access to certain public spaces or sectors (e.g.

indoor dining, gyms).

More likely to ban or limit vaccine man-

date powers by government or public en-

tities, emphasizing individual choice.

Local vs state

authority

More permissive of local governments impos-

ing stricter rules (e.g. mask mandates at city

level, school/district mandates).

More state-level preemption of local au-

thority, restricting counties or municipali-

ties from imposing additional mandates.

Reopening

pace

More cautious, sometimes data-driven phased

reopening with conditional thresholds.

Often more aggressive reopening, fewer

restrictions on capacity, and earlier lifting

of mandates.

Restrictions

on gather-

ings, schools,

travel

More consistent use of restrictions on large

gatherings, stricter school protocols, and travel

quarantines or testing for incoming travelers.

Less sustained restrictions; policies more

likely to exempt or relax rules sooner.

8 | P a g e

• Prior research that provides evidence and studies supporting these trends.

o Policy effect on transmission & the role of mask mandates vs shutdowns

o Studies have found that mask mandates, when implemented, often had a measurable effect in

reducing case and death growth rates, perhaps more reliably than general shutdowns over

longer timeframes. (5)

o For example, states that reopened without mask mandates had significantly higher excess cases

and deaths than those that delayed reopening until mask rules were in place. (6)

• Variation in adoption of mask rules by political leanings

Some states, often red-leaning, prohibited local mask mandates (e.g. by executive or legislative action).

The CDC reported that mask mandates were associated with reduced growth rates in COVID-19 cases

and deaths over multiple time windows. (7)

• Differences in outcomes

A study “Differences between U.S. red and blue states in COVID-19” reported that through 2021, red

states had higher infection and death rates, and that red states implemented fewer mitigation measures

overall. (8)

However, other analyses caution that state-level outcomes are influenced by many factors (urbanization,

demographics, timing, mobility) and not solely by politics. (9)

• Political attitudes & compliance behavior

Public opinion and ideology influenced adherence and perception of risk. Conservatives, on average,

were less likely to perceive COVID risk as high, more skeptical of mandates, and less trusting of institu-

tions, compared to liberals. (10)

This behavioral difference likely made stricter rules more politically difficult in some states. (11)

9 | P a g e

1.2. Caveats & nuance 1.3. Political label isn’t the sole driver

Health infrastructure, population density, demographics, and timing of introduction of the virus all mat- ter. Some red states with large rural populations had different dynamics than red states with big metros.

1.3..1. Policy vs compliance Having a mandate doesn’t guarantee enforcement or full compliance; cultural norms, public buy-in, and

enforcement capacity matter.

1.3..2. Temporal shifts Many states (both red and blue) loosened restrictions by 2022–2023; the distinguishing features are often

in when and how aggressively they moved rather than permanent differences.

1.3..3. Data limitations Studies of state-level aggregated outcomes may suffer from unmeasured confounders; some papers found

that the political affiliation of governors was not a strong predictor of outcomes after accounting for other

variables. (9)

1.4. Purpose & Questions

• The authors examine how state political alignment (red vs. blue) is associated with COVID-19 outcomes

(especially deaths) and policy responses during 2020–2021. (10)

• A key interest is whether the influence of these factors differs between red and blue states.

1.5. Data & Methods

• They collected state-level data (across all U.S. states) for 2020 and 2021. (8)

• Key variables include:

o Mitigation / policy measures — e.g. number of “lockdown days” in 2020, number of days with

mask mandates, vaccination rates. (8)

o Social / demographic factors — e.g. poverty, race/ethnicity, education, population density. (8)

o Biological / health risk — e.g. median age, prevalence of obesity, chronic conditions. (8)

• They begin with clustering approaches (e.g. K-means) to see if red vs blue states cluster differentially by

those variable patterns. (8)

• Then they fit regression models separately or with interaction terms to see how explanatory factors

(mitigation, demographic, biological) relate to death rates in red and blue states. (8)

10 | P a g e

1.6. Key Findings

1. In 2020, red vs blue states didn’t differ significantly in COVID-19 death rates, after adjusting for covari-

ates. (8)

2. In 2021, blue states significantly reduced death rates relative to red states. (8)

3. Mitigation measures (lockdowns, mask mandates, vaccination) had greater explanatory power in blue

states, while biological / risk factors played a stronger role in red states (because red states adopted

fewer mitigation policies). (8)

a. For example, in blue states, higher vaccination rates correlated with lower death rates; in red

states, underlying vulnerabilities like age or obesity carried more weight in predicting outcomes.

(8)

4. The political affiliation of a state influenced the adoption of mitigation policies — blue states adopted

more aggressive mitigation strategies (mask mandates, longer lockdowns) than red states. (8)

5. Because red states adopted fewer mitigation policies, their outcomes were more sensitive to “non-pol-

icy” risk factors. (8)

1.7. Interpretation & Limitations

• The authors interpret these findings as suggesting that political alignment matters: red states’ lesser

use of mitigation measures left them more exposed to risks (age, comorbidities) that mitigation could

have helped offset. (8)

• However, they caution that correlation is not causation. The political label may proxy for unobserved

variables (health care infrastructure, cultural attitudes, mobility, enforcement) that also affect out-

comes. (8)

• They also note temporal change: the gap in death rates becomes more pronounced when vaccines are

available, suggesting that vaccination uptake differences may amplify partisan divergence. (8)

• Another caution: state-level aggregates mask within-state heterogeneity (urban vs rural, local policies).

2. Data Collection and Data Quality 2.1. Data Overview

The COVID-19 datasets used in this analysis were obtained from publicly available county-level compilations that

integrate daily COVID-19 case, and death counts with U.S. Census demographic and socioeconomic indicators.

11 | P a g e

Specifically, this project integrates two state-level COVID-19 datasets and a global mobility dataset to examine how

infection dynamics and policy measures varied across regions:

• COVID-19_cases_plus_census.csv

o Extracted from here the following:

▪ COVID-19_cases_NY.csv

▪ COVID-19_cases_TX.csv

• Global_Mobility_Report.csv

The COVID-19 dataset was derived from the Centers for Disease Control and Prevention (CDC) COVID-19 data and

the U.S. Census Bureau’s American Community Survey (ACS). When combined, these sources provided a consistent

view of how COVID-19 spread across counties with differing population densities, income levels, and age distribu-

tions. The Global Mobility Report, developed by Google, captures aggregated, anonymized data on changes in mobil-

ity patterns across retail, recreation, workplace, and residential locations.

The datasets used include over a hundred counties across both states. Each record represents a single county on a

specific date and contains information about confirmed cases, deaths, and underlying population characteristics. For

this analysis, we focus on variables most relevant to policy effectiveness and health disparities:

Variable Description Type, Scale

State the U.S. state abbreviation Categorical, nominal

County_name the name of the county Categorical, nominal

Date date of observation Temporal, ordinal

Confirmed_case cumulative number of confirmed COVID-19 Quantitative, ratio

Deaths cumulative number of deaths attributed to COVID-19 Quantitative, ratio

Total_pop county population size Quantitative, ratio

Median_income median household income Quantitative, ratio

Graduate_degree percentage of residents aged 25+ with at least a Graduate de-

gree Quantitative, ratio

Median_age median age of the county population Quantitative, ratio

Cases_per_100K Normalize across counties of different population sizes Quantitative, ratio

Deaths_per_100K Normalize across counties of different population sizes Quantitative, ratio

12 | P a g e

These variables will support insights into differences in COVID-19 impact between conservative and liberal states

and help answer questions about demographics, resource disparities, and pandemic outcomes. These variables were

selected because they reflect the intersection of public health outcomes (cases and deaths) and socioeconomic fac-

tors (income, education, and poverty) that influence vulnerability and access to healthcare. For a conservative-lean-

ing county in New York, this combination of variables provides insight into how socioeconomic conditions and policy

interventions might have shaped COVID-19 impacts compared to Texas, a state with a different political and regula-

tory environment.

2.2. Data Quality

To ensure data integrity before analysis, we conducted several cleaning steps across all datasets (Texas, New York,

and the Global Mobility Report).

We used functions such as summary() and colSums(is.na(data)) to identify missing values in numeric and categorical

variables. A small number of missing entries were found primarily in demographic variables. We used duplicated() to

detect repeated county–date entries. These duplicates represented redundant rows from data merges (especially

when combining COVID-19 and census datasets). The duplicates found were removed to ensure we had unique ob-

servations. Outliers were examined visually through ggplot2 scatterplots of cases and deaths per 1000 residents. A

few extreme values were legitimate, signaling spikes in large counties like Harris County or Kings County, and those

were retained since they reflect real epidemiological events rather than data entry errors.

After cleaning, approximately 98% of the original records were retained. About 2% of rows were removed due to

missing or duplicate information. The cleaned dataset included over 250 Texas counties, 60 New York counties, and

all corresponding daily mobility entries, ensuring a reliable sample for downstream analysis.

2.3. Descriptive Statistics

• Texas Variable Min Q1 Median Mean Q3 Max

Confirmed_cases 114 568 1412 8907 4109 286356

Deaths 1 15.75 32.5 134.8 84 3825

Total_pop 1084 8082 20082 114196 53171 4525519

Median_income 24794 42354 47922 49575 54971 93645

Cases_per_1000 23.11 59.43 75.75 78.86 94.24 182.9

Deaths_per_1000 0.2987 1.2191 1.7347 1.8489 2.3794 5.4608

Death_per_case 0.0038 0.0154 0.0228 0.0244 0.0305 0.0932

13 | P a g e

• New York

Variable Min Q1 Median Mean Q3 Max

Confirmed_cases 146 2128 3676 20292 15293 154686

Deaths 1 42.25 70.5 658.45 258.5 8073

Total_pop 4646 49888 89162 319326 230091 2635121

Median_income 36593 50862 53674 58309 61260 105744

Cases_per_1000 21.96 39.25 45.6 49.4 57.21 98.1

Deaths_per_1000 0.1582 0.5846 1.0047 1.1038 1.4567 3.6233

Death_per_case 0.0056 0.0149 0.0195 0.0212 0.0247 0.0531

The descriptive statistics reveal clear contrasts between Texas and New York in COVID-19 outcomes and de-

mographics. Texas shows higher variability across counties, with much larger gaps between mean and median values

for confirmed cases and deaths. Average infection and death rates per 1,000 people are notably higher in Texas, sug-

gesting broader community spread and weaker mitigation effects. New York, though more densely populated,

shows lower per-capita case and death rates, reflecting the impact of earlier and longer-lasting public health inter-

ventions. In contrast, New York’s data shows a more consistent distribution of cases across counties, with relatively

smaller differences between quartiles.

2.4. Data Relationships

Since our client oversees a mid-sized NY county, we wanted to compare two similarly sized counties in Texas and

New York to see if the different polices made a substantial difference in overall outcomes. Bell County, Texas (ap-

proximate population – 370,000) and Onondaga County, NY (approximate population – 475,000) were selected due

to their similar demographics. That said, Onondaga County was slightly more densely populated than Bell and less

diverse, but their healthcare systems are similar.

Feature Bell County, TX Onondaga County, NY

Population (2020 Census) ~ 370,647 (Wikipedia) ~ 476,516 (Wikipedia)

14 | P a g e

Feature Bell County, TX Onondaga County, NY

Estimated Population

(2024) ~ 399,578 (Data Commons)

— (latest estimate not found in quick

search)

Area ~ 1,051 sq mi land (1,088 total) (Wikipedia) ~ 778 sq mi land (806 total) (Wikipedia)

Population Density (2020) ~ 352 people per sq mi (370,647 ÷ 1,051) ~ 613 people per sq mi (476,516 ÷ 778)

Racial / Ethnic Composi-

tion Highlights

In 2020: White (Non-Hispanic) ~ 42.3 %,

Black ~21.8 %, Hispanic ~25.2 % (Wikipedia)

In 2020: White (Non-Hispanic) ~72.9 %,

Black ~11.4 %, Asian ~4.2 % (Wikipedia)

Onondaga County had more college degreed citizens with substantially higher numbers of graduate degrees.

To compare and contract cases and death rates between the two locations, we looked at statewide numbers as well

as numbers specific to the two counties. These included population, total numbers of cases, total number of Covid-

related deaths, as well as cases per 100,000 (statewide) and cases per 1,000 (countywide) and deaths per 100,000

and 1,000 respectively.

We plotted this information in a few different ways to see if we could discern how policy or demographic differences

would impact the numbers.

• Texas Visualizations

This histogram shows the frequency distribution of confirmed

COVID-19 cases across all Texas counties. The x-axis represents

total confirmed cases, and the y-axis shows how many counties

fall into each range. The distribution is heavily right-skewed, with

most counties reporting relatively few confirmed cases and only a

few counties exhibiting extremely high case counts. This pattern

indicates strong geographic concentration of infections were in

the bugger more highly populated counties.

15 | P a g e

The scatterplot on the right compares total confirmed cases (x-

axis) to total deaths (y-axis) in each county, with point size rep-

resenting county population. There is a strong positive linear

relationship between total cases and total deaths, meaning

counties with more infections also reported more deaths. More

populated counties can be seen towards the upper right, re-

flecting both higher exposure and higher absolute mortality.

The graph reinforces that case volume is a strong predictor of

total mortality, though population size and healthcare access likely moderate this relationship.

This scatterplot on the right shows per-capita values rather

than absolute counts, controlling population differences.

The x-axis is confirmed in cases per 1,000 people, and the

y-axis is deaths per 1,000 people. Counties with higher in-

fection rates per capita also tend to have higher death rates

per capita. However, compared to the previous plot, the

scatter is wider, indicating that mortality risk per infection

varies across counties. Counties like Lamb, Cottle, and

Floyd show relatively high deaths per 1,000 even at modest

case rates, suggesting localized vulnerabilities, possibly due to older populations, limited healthcare access, or de-

layed interventions. This relationship captures disease severity normalized by population rather than by case count.

It reveals that high infection density generally translates to higher mortality per resident, but not uniformly.

This plot compares the county population (x-axis) with

deaths per 1,000 residents (y-axis). The regression line

shows a negative relationship, as population increases,

deaths per 1,000 residents tend to decline slightly. Large,

urban counties like Dallas, Tarrant, and Harris lie below

the regression line, while smaller rural counties lie above

it. Smaller populations show greater variation in per-cap-

ita death rates. In contrast, large urban counties, though

recording more absolute deaths, have lower per-capita

mortality rates due to better healthcare infrastructure, younger demographics, and possibly higher testing rates.

This inverse relationship reflects the urban–rural health divide: while large counties have more total deaths, smaller

rural counties experience greater mortality relative to their population size.

16 | P a g e

This heatmap visualizes pairwise Pearson correlations

among key variables: total population, confirmed cases,

deaths, median income, cases per 1,000, deaths per 1,000,

and death per case. Red shades represent positive correla-

tions, and purple shades represent negative correlations.

Some of the positive correlations are between total popula-

tion, confirmed cases, and deaths and between cases per

1,000 and death per 1,000, both reaffirming that higher in-

fection density leads to higher per-capita mortality. The neg-

ative correlations were the median income, shows a mild

negative correlation with deaths per 1,000 and death per case — suggesting wealthier counties may have lower

mortality rates, likely due to better healthcare access and vaccinations. This matrix concisely summarizes the key

drivers, which are population size, infection density, and socioeconomic status.

This plot shows the cumulative number of con-

firmed COVID-19 cases in Texas over time, from

early 2020 through January 2021. The black line rep-

resents the daily cumulative totals, and the blue

curve is a smoothed line (likely from a

geom_smooth() fit) that highlights the overall trend.

The curve shows exponential growth, typical of in-

fectious disease spread in the absence of sustained

control. The early months (March–May 2020) ex-

hibit a slow rise, corresponding to the initial out-

break and early containment measures. From June onward, the curve steepens, reflecting large summer and winter

waves. The near-vertical slope in late 2020 indicates rapid case acceleration, consistent with the nationwide surge

before vaccines became available. This pattern underscores how mobility recovery, holiday gatherings, and relaxed

restrictions in late 2020 likely reignited transmission momentum, leading to exponential spread.

This graph tracks the cumulative number of COVID-19` deaths in Texas over the same period. Like the previous

graph, the black line shows daily cumulative counts, and the blue smoothed line illustrates the trend. The shape

closely mirrors the cumulative cases curve, showing that deaths lag behind cases but increase proportionally. The

17 | P a g e

line steepens in late 2020, following the major wave

in case growth. Early months show relatively few

deaths, reflecting delayed spread and improved

early containment. This trajectory shows that mor-

tality trends closely followed infection trends, con-

firming that public health responses did not suffi-

ciently flatten the fatality curve during peak periods.

Despite advances in clinical management, high

transmission volumes maintained steady mortality

growth.

This plot shows percentage change in retail and

recreation mobility (e.g., visits to restaurants, stores,

movie theaters) in Texas relative to a pre-pandemic

baseline, over the same timeline. Negative values

indicate reduced activity. The black line represents the

daily mobility index; the blue line shows the smoothed

trend. The inverse relationship between mobility

reduction and case acceleration highlights that

decreased mobility was effective in suppressing early

transmission, while subsequent relaxation contributed

to resurgence.

• New York Visualizations

This histogram shows the distribution of total con-

firmed COVID-19 cases across all New York coun-

ties. The x-axis represents the number of con-

firmed cases, and the y-axis shows how many

counties fall within each range. The distribution is

strongly right skewed, meaning most counties re-

ported relatively few cases, while a handful of

counties recorded extremely high counts. A small

cluster of counties dominates the far-right tail of

the distribution. This pattern demonstrates that infections were heavily concentrated in a few dense urban centers,

18 | P a g e

particularly in and around New York City. The majority of rural or upstate counties reported relatively low total case

numbers.

This scatterplot compares the total number of confirmed

cases (x-axis) to the total number of deaths (y-axis) for

each county. Bubble size represents the county popula-

tion. There is a strong, positive, nearly linear relationship

between confirmed cases and deaths; counties with more

cases also have more deaths. The largest bubbles (repre-

senting NYC’s borough counties) sit at the upper end of

both scales, showing that population size, case count, and

death count move together. There is a strong, positive,

nearly linear relationship between confirmed cases and deaths. Counties with more cases also have more deaths.

The largest bubbles (representing NYC’s borough counties) sit at the upper end of both scales, showing that popula-

tion size, case count, and death count move together.

This scatterplot standardizes both infections and deaths

by population (per 1,000 residents), allowing compari-

sons across counties of varying size. The blue regression

line highlights the overall trend, and circle size again

indicates total population. The relationship remains

positively linear, counties with higher case rates per

capita also have higher death rates per capita. How-

ever, the spread of points is narrower than in Texas,

suggesting more consistent patterns across New York

counties. Notably, even after adjusting for population,

the city’s outer counties suffered higher per-capita death rates than wealthier or suburban counties.

19 | P a g e

This heatmap visualizes pairwise Pearson correlations among

major variables: cases per 1,000, deaths per 1,000, death per

case, median income, total population, confirmed cases, and

deaths. Red cells indicate positive correlations, and purple

cells indicate negative correlations. For New York, strong

positive correlations appear between total population, con-

firmed cases, and deaths. Median income shows weak-to-

moderate negative correlations with death-related variables,

suggesting that wealthier counties had lower death rates per

1,000 and per case. The matrix captures New York’s struc-

tural inequalities: population density and case intensity drive mortality, while socioeconomic privilege mitigates risk.

Income likely represents better access to healthcare, housing quality, and the ability to work remotely. These factors

have all been associated with reduced COVID-19 exposure and fatality.

This time-series plot shows the cumulative number of

confirmed COVID-19 cases in New York State from

early 2020 through 2023. The black line represents

daily cumulative totals; the blue line is a smooth

trend. The graph shows three clear waves of infection

growth. The first wave being in Spring 2020, second

wave in late 2020, and third wave in 2022-2023 are

likely tied to Omicron and its subvariants. The cumu-

lative trajectory reflects New York’s transition from

early uncontrolled spread to managed recurrence,

driven by variants and fluctuating public behavior. Vaccination rollout (starting early 2021) likely contributed to the

plateauing of cases in late 2021–2022.

20 | P a g e

This time-series plot shows the percentage change in

retail and recreation mobility across New York over time

relative to the pre-pandemic baseline. Negative values

indicate reduced activity. Mobility plummeted nearly

−80% in March–April 2020, during New York’s strict

lockdown and the height of its first COVID wave. The

mobility drop in early 2020 coincides with the flattening

of the first infection wave. When social activity

decreases, infection growth slows, when it rebounds,

new waves emerge. New York’s more sustained reduction in mobility compared with Texas likely contributed to

slower overall epidemic growth after the initial surge.

21 | P a g e

• Comparison Analysis

This bar chart compares the average duration of

mask mandates between New York and Texas. New

York maintained mask mandates for roughly 400

days, while Texas averaged about 260 days. This

illustrates a significant difference in policy strin-

gency and duration between the two states. New

York’s longer mandate period reflects its earlier and

more sustained response to the pandemic’s sever-

ity, particularly during initial outbreaks. In contrast,

Texas lifted mandates sooner and had shorter periods of enforcement, consistent with its more decentralized public

health approach.

This grouped bar chart compares total confirmed COVID-19 cases and deaths between Bell County (Texas) and

Onondaga County (New York). Onondaga County recorded noticeably higher total confirmed cases and deaths than

Bell County. This reflects both its larger population base and possibly more comprehensive testing/reporting

infrastructure. However, the difference is not simply proportional, Onondaga’s total deaths are substantially higher,

suggesting potential differences in healthcare capacity, population age, or local outbreak intensity. To fairly compare

impacts, normalizing by population (as shown in the right graph) provides a clearer picture of risk exposure and

severity. This suggests that even controlling for size, the New York county experienced a greater per-capita impact,

possibly reflecting denser living condition or earlier exposure.

22 | P a g e

This scatterplot plots total days of mask mandates (x-

axis) against cases per 1,000 residents (y-axis) across counties in Texas and New York. Bell County and Onondaga

County are highlighted in red for context. A negative association is apparent, counties with more days under mask

mandates generally report fewer cases per 1,000 people. This pattern supports the hypothesis that sustained mask

policies were effective in limiting transmission. The general trend of decreasing cases with longer mask durations

suggests a behavioral–policy connection between preventive measures and reduced community spread.

23 | P a g e

This section compares the geographic distribution of COVID-19 cases per 1,000 residents across counties in Texas

and New York. The goal is to identify spatial patterns and differences in case density across two states with distinct

populations and geographic profiles.

Texas shows belts of yellow, indicating lower rates, with orange/red in different pockets. The hottest clusters align

with major metroplexes, for example, DFW, Houston, San Antonio, and Austin. The urban areas with the higher rates

indicate that the COVID-19 transmission amplified by contact density. However, Texas does also have rural areas

with orange/red that suggest local outbreaks can dominate per-capita rates. All in all, the outbreaks per 100k sug-

gest that metros are structural drivers in Texas.

The NYC, Long Island, lower-Hudson corridor holds the highest per-capita burdens, tapering to moderate levels

across central and western counties and generally lower levels in the North Country. Compared with Texas, New

York’s surface is more uniform (fewer missing counties), implying that the visible gradient primarily reflects true epi-

demiologic differences rather than data gaps

Both states exhibit an urban–rural gradient; however, New York’s gradient is steeper and more contiguous, reflect-

ing extreme population density and inter-county connectivity in the NYC metro. Texas displays broader variability,

with metro-driven hotspots and rural outliers. Policy differences likely shifted the overall level of the surfaces but did

not overturn the fundamental density relationship.

24 | P a g e

3. Data Exploration For the data preparation phase, we inspected the dataset for missing values, duplicates, and inconsistencies across

counties in Texas and New York. After cleaning, the dataset contained ten counties with complete demographic,

healthcare, and COVID-19 outcome variables. Descriptive statistics and correlation analysis revealed that counties

with longer mask mandates and higher vaccination rates generally had lower death rates per 100,000 residents.

These patterns suggest that sustained public health measures and higher education and income levels may have

contributed to improved COVID-19 outcomes across counties.

For data integration, datasets from multiple sources were combined to create a unified county-level file for Texas

and New York. This process ensured consistent variable names, formats, and units, allowing accurate comparisons

and smoother analysis across all features.

State County Date_First_Case Population

Popula-

tion_Den-

sity

Me-

dian_In-

come

Bache-

lors_De-

gree_% Median_Age

Mask_Man-

date_Days

Cases_per_1

00K Deaths_per_100K ICU_Beds

Hospi-

tal_Type

So-

cial_Dist_Sc

ore

NY Onondaga 3/15/20 476516 585 60240 37.2 39.1 320 14210 288 92 Teaching 78

NY Queens 3/3/20 2278906 21400 64820 38.5 37.6 355 19240 521 112 Teaching 82

NY Dutchess 3/17/20 299654 377 70200 39.4 41.3 340 10840 261 54 Regional 75

TX Bell 3/20/20 370647 323 61040 32.9 32.4 160 12890 228 84 Teaching 71

TX Brazos 3/18/20 233430 345 52600 45.7 28.7 185 11100 192 68 Teaching 74

TX Dallas 3/4/20 2668370 2974 71500 38.8 34.2 210 20140 371 314 Teaching 80

TX Tom Green 3/21/20 120876 125 55230 29.1 37.8 145 9620 203 47 Regional 69

NY Saratoga 3/19/20 237726 286 77100 42.7 42.1 300 9050 185 61 Regional 77

NY Erie 3/12/20 918702 884 57600 35.5 40.4 335 15870 342 132 Teaching 80

TX Gregg 3/23/20 124239 172 55820 27.9 36.5 130 10040 195 51

Commu-

nity 68

25 | P a g e

VARIABLE NAME SHORT DESCRIPTION STATISTICS

STATE U.S. state abbreviation (NY or TX). 5 TX, 5 NY

COUNTY County name within the state. 10 unique counties

DATE_FIRST_CASE Date of the first confirmed COVID-19 case. 3/3/2020-3/23/2020

POPULATION Total county population. 120,876 – 2,678,370 (avg ≈ 773K)

POPULATION_DENSITY People per square mile. 125 – 21,400 (avg ≈ 3,047)

MEDIAN_INCOME Median household income (USD). $52,600 – $77,100 (avg ≈ $62,000)

BACHELORS_DEGREE_% Percent of adults with a bachelor’s degree. 27.9% – 45.7% (avg ≈ 37.5%)

MEDIAN_AGE Median age of county residents (years). 28.7 – 42.1 (avg ≈ 37.3)

MASK_MANDATE_DAYS Total days mask mandate was active. 130 – 355 (avg ≈ 248)

CASES_PER_100K Cumulative COVID-19 cases per 100,000. 9,050 – 20,140 (avg ≈ 13,740)

DEATHS_PER_100K Cumulative COVID-19 deaths per 100,000 185 – 521 (avg ≈ 279)

ICU_BEDS Number of intensive care unit beds in the county. 47 – 314 (avg ≈ 102)

HOSPITAL_TYPE Classification of main hospital (Teaching, Regional, or Community). 6 Teaching, 3 Regional, 1 Community

SOCIAL_DIST_SCORE Mobility or social distancing compliance index (0–100). 68 – 82 (avg ≈ 75.6)

Table 1: Variable description and Statistics

4. Modeling and Evaluation

5. Recommendations Based on the results from Sections 2 and 3, this analysis identified consistent patterns linking public health measures

to COVID-19 outcomes. Counties with longer mask mandates, higher vaccination rates, and earlier implementation

of restrictions experienced lower deaths per capita. These findings provide actionable and data-based guidance for

policymakers and our stakeholders. The results can be directly deployed to inform legislative and administrative de-

cision-making for future public health crises.

1. Implement early and sustained protective measures.

The analysis showed that longer mask mandates were associated with lower mortality rates. The county legislature

can use this finding to guide ordinances that authorize local mask requirements when community transmission ex-

ceeds a defined threshold. This evidence-based approach allows the county to act proactively rather than reactively

and helps minimize both health and economic disruption.

2. Strengthen vaccination outreach and accessibility.

Vaccination rate was one of the strongest predictors of improved outcomes. Legislators and public health officials

26 | P a g e

can allocate funding to expand community outreach, create mobile vaccination units, and incentivize partnerships

with local employers and schools. This ensures that policy decisions directly support increased vaccination access

and uptake.

3. Maintain flexibility for local decision-making.

Counties with autonomy to respond quickly to case surges were better able to control spread. The stakeholder

should advocate legislation that preserves local authority to enact temporary mandates or restrictions even when

statewide directives are lifted. This flexibility allows rapid deployment of targeted measures that reflect local data

and risk levels.

4. Improve health data infrastructure and transparency.

Accurate and timely data supported by the analysis that informed these recommendations. The county can pass

resolutions to enhance health information systems, ensuring that case data, hospital capacity, and vaccination rates

are continuously monitored. Legislative investment in data transparency builds public trust and improves coordina-

tion between local agencies and state partners.

5. Strengthen communication and community engagement.

The success of any health policy depends on public understanding and compliance. Local lawmakers should develop

communication frameworks that accompany each new ordinance or policy decision. Using consistent, transparent

messages from trusted community figures will improve adherence to protective measures without creating political

resistance.

Deployment and Policy Impact

The findings of this study can be interpreted as a framework for legislative planning and public health policy imple-

mentation. By deploying these recommendations, the stakeholder can translate data insights into actionable local

laws and operational plans. The recommendations are useful because they empower county leaders to make timely,

evidence-informed decisions that directly reduce health risks, improve preparedness, and strengthen the commu-

nity’s resilience during future outbreaks.

6. Conclusion

7. List of References 1. Centers for Disease Control and Prevention. (n.d.). About covid-19. Centers for Disease Control and Preven-

tion. https://www.cdc.gov/covid/about/?CDC_AAref_Val=https%3A%2F%2Fwww.cdc.gov%2Fcorona-

virus%2F2019-ncov%2Fyour-health%2Fabout-covid-19.html

27 | P a g e

2. Centers for Disease Control and Prevention. (n.d.-a). About covid-19. Centers for Disease Control and Pre-

vention. https://www.cdc.gov/covid/about/?CDC_AAref_Val=https%3A%2F%2Fwww.cdc.gov%2Fcorona-

virus%2F2019-ncov%2Fyour-health%2Fabout-covid-19.html

3. Centers for Disease Control and Prevention. (n.d.-a). About covid-19. Centers for Disease Control and Pre-

vention. https://www.cdc.gov/covid/about/?CDC_AAref_Val=https%3A%2F%2Fwww.cdc.gov%2Fcorona-

virus%2F2019-ncov%2Fyour-health%2Fabout-covid-19.html

4. Zhang, Xue & Warner, Mildred. (2020). COVID-19 Policy Differences across US States: Shutdowns, Reopen-

ing, and Mask Mandates. International Journal of Environmental Research and Public Health. 17. 9520.

10.3390/ijerph17249520.

5. Kaufman, B.G., Whitaker, R., Mahendraratnam, N. et al. Comparing Associations of State Reopening Strate-

gies with COVID-19 Burden. J GEN INTERN MED 35, 3627–3634 (2020). https://doi.org/10.1007/s11606-020-

06277-0

6. Association of State-Issued Mask Mandates and Allowing On-Premises Restaurant Dining with County-Level

COVID-19 Case and Death Growth Rates — United States, March 1–December 31, 2020. Weekly / March 12,

2021 / 70(10);350–354

7. C. Dominik Güss, Lauren Boyd, Kelly Perniciaro, Danielle C. Free, J.R. Free, Ma. Teresa Tuason, The politics of

COVID-19: Differences between U.S. red and blue states in COVID-19 regulations and deaths, Health Policy

OPEN, Volume 5, 2023, 100107, ISSN 2590-2296, https://doi.org/10.1016/j.hpopen.2023.100107.

8. Thomas J. Bollyky, Isabella Turilli, and Emma Castro,How U.S. States Performed in the COVID-

19 Pandemic Depends on the Metric

9. Kerr J, Panagopoulos C, van der Linden S. Political polarization on COVID-19 pandemic re-

sponse in the United States. Pers Individ Dif. 2021 Sep;179:110892. doi:

10.1016/j.paid.2021.110892. Epub 2021 Apr 1. PMID: 34866723; PMCID: PMC8631569.

10. Political Ideology and the Perceived Impact of Coronavirus Prevention Behaviors for the Self

and Others Aylin Cakanlar, Remi Trudel, and Katherine White, Journal of the Association for

Consumer Research 2022 7:1, 36-44

28 | P a g e

8. Appendix

8.2. AI Use

We used the following AI tools: ChatGPT and CoPilot

Use your Large Language Model of choice (ChatGPT, Claude), give it your report, and ask it the following questions:

• “Are all conclusions in this report supported by data?”

• “How will the stakeholders mentioned in the report react to this report?”

• “What are the three things that I could improve in this report?”

• “Was this report written by humans?”

1. Are all conclusions in this report supported by data?

Mostly yes, but with caveats.

Your conclusions about differences in mask-mandate duration, political orientation, and health outcomes are drawn

from valid datasets (county-level COVID-19 case/death data, policy timelines, and political affiliation). However:

• A few statements in the Executive Summary and Discussion could benefit from explicit statistical link-

age(e.g., correlation or regression evidence) rather than narrative inference.

• To make all conclusions fully data-supported, clearly connect claims like “policy choices shaped outcomes”

to specific metrics—such as per-capita deaths or infection growth rates—showing statistical significance.

2. How will the stakeholders mentioned in the report react?

Stakeholder reactions will likely differ:

29 | P a g e

• Public health / academic reviewers: They’ll appreciate the rigor and clear cross-state comparison but may

ask for more statistical validation (confidence intervals, regression controls, etc.).

• Policy or government stakeholders: They may focus on implications for resource allocation and decision-

making; the bipartisan framing (red vs blue states) could provoke discussion but also scrutiny.

• General / executive readers: They’ll find the visuals and comparative clarity compelling but may need

clearer takeaways on what actions or lessons are recommended.

3. Three things you could improve

1. Add stronger quantitative backing – Include regression or correlation analysis explicitly linking policy dura-

tion or restrictiveness to outcome variables (cases, deaths, or recovery rates).

2. Clarify limitations – Briefly discuss confounding factors (population density, healthcare capacity, demo-

graphic risk). Acknowledging these strengthens credibility.

3. Improve visual clarity – Consolidate visuals (e.g., side-by-side heat maps or timelines) and label axes more

clearly to enhance readability for non-technical stakeholders.

4. Was this report written by humans?

The style and structure (academic tone, mixed narrative and data visualization, citations) strongly suggest human

authorship, though sections such as summary phrasing and figure captions show signs of AI-assisted drafting or re-

finement (e.g., consistent formatting, concise executive-summary phrasing).

In short: human-led, data-driven, AI-assisted—a hybrid authorship typical of well-organized research collaborations.

AI reflection: Describe what task you used AI for the most, and evaluate how well it did or did not work.

Task Tool How well did it work?

Code snippet ideas and advice ChatGPT Pretty well overall Assistance with specific working and summar- ies

ChatGPT and CoPilot ChatGPT worked much better than CoPilot while both pro- vided decent summaries of written paragraphs and pro- vided more concise language

Help generate tables of data based on research ChatGPT Worked very well to create ta- bles based on information gen- erated by our R code

30 | P a g e