Home
Predicting Flight Delays Exploring Catastrophic Forgetting for Question Answering Domain Adaptation in Text-to-Text Transformers Effect of Political Ideology on Consumer Behavior Pulmonary Nodule Segmentation Poverty and COVID-19 Barriers to entry in Open Source Software Certifications and Discounts Interactive Wheel Based Image Recoloring GUI Coreless Promoters in Arabadopsis DNA Health Portal - Built for All

Poverty and COVID-19

The Purpose

Our questions were:

  • How were those living in poverty impacted by COVID-19?
  • How do other factors related to poverty such as number of homeless and percent unemployed affect the COVID-19 case rate?
  • How do state characteristics affect the COVID-19 case rate?

The Quest

Our hypothesis was that those living under poverty will have a significant impact on the COVID-19 case rate because these are the people that are probably living in close quarters where social distancing is hard to follow. We are also interested in seeing if certain characteristics and pdolicies such as mask mandate and medicaid expdentiure had an impact on the COVID-19 case rate.

These are the variables we are interested in:

  • Population density per square miles: Population density of each state per square mile.
  • Percent Unemployed (2018): The percentage of people unemployed in each state. This is 2018 data.
  • Number Homeless (2019): The number of homeless people in each state. This is 2019 data.
  • Median Annual Household Income: The median annual income for a household in each state.
  • Percent at risk for serious illness due to COVID: Percentage of state population that may develop a serious illness due if they are infected with COVID-19.
  • Nonelderly Adults Who Have A Pre-Existing Condition: Total number of non-elderly adults that have a pre-existing condition.
  • Life Expectancy at Birth (years): The average life expectancy for an individual in that state at birth.
  • Medicaid Expenditures as a Percent of Total State Expenditures by Fund: The percent of money that a state spent on medcaid expenses in relation with the total money a state spent.
  • No legal enforcement of face mask mandate: Whether the state enforced a legal requirment to wear a mask.
  • Mandate face mask use by all individuals in public spaces: Start date of face mask mandate for all individuals in state

We sanitized our data and looked at correlation between the variables:

From this correlation plot, we could see that povertyLine and lifeExpectancy have a strong negative correltaion. This indicated that living under the poverty line correlates to a shorter life expectancy. Life expectancy is also strongly negatively correlated to percent at risk for a serious illness due to COVID-19. The higher the percentage at risk, the lower the average life expectancy. We dropped lifeExpectancy from our analysis to ensure that we don’t violate the multicollinearity assumption. The number of homeless people is strongly positively correlated to non-elderly adults that have pre-existing conditions.

Next we built visualized the distributions of these variables:

From these distributions, we noticed that population density, non-elderly adults with pre-existing conditions, and number of homeless are skewed to the left so we applied a log tranformation on these variables. Here are the distributions after the transformation:

Redesigned UI

The Solution

We built three models:

  • Model 1: lm( formula = caseRate ~ povertyLine)
  • Model 2: lm(formula = caseRate ~ povertyLine + logPopDensity + percentUnemployed + logNumHomeless)
  • Model 3: lm(formula = caseRate ~ povertyLine + maskMandate + logPopDensity + logPreExistingCondtions + percentAtRisk + logNumHomeless + medicaidSpending + percentUnemployed + medianIncome, data = covid_19TData)

The main takeaways for each model are:

  • Model 1: Adjusted R squared is 0.07. The p-value (0.037) for the percentage of people living under the federal poverty line (2018) is just about within the confidence level. From this interpreted, that the percentage of people living under the federal poverty is related to the COVID-19 case rate. However, from the low R squared, we know that there are other variables that also affect the COVID-19 case rate.
  • Model 2: Adjusted R squared is 0.09. The p-value (0.007) for the percentage of people living under the federal poverty line (2018) has now increased in signficance. Another interesting finding is that the percent unemployed is also negatively related to COVID-19 case rate. However, the R squared value has only increased a bit from our last model. From the low R squared, we know that there are other variables that also affect the COVID-19 case rate.
  • Model 3: Adjusted R squared is 0.4. The p-value (0.001) for the percentage of people living under the federal poverty line (2018) has is at the same significance level from Model 2. However, other factors such as whether masks were mandated p-value: 0.003) and the percentage of people at risk of developing a serous illness were both significant factors p-value: 0.0004). They both had a negative relation with the COVID-19 case rate. THe r-squared value is also much better and speaks to a better fit.

The Impact

The percent of peole living under the poverty line in a state has a significant positive relationship on the COVID-19 case rate per 100k people in that state. This was a consistent relationship across all of our models and only became more significant as more factors affecting poverty were considered. The face mask mandate was effective in lowering the COVID-19 case rate. We were surprised that percentage of people at risk for developing a serious illness due to COVID-19 had a negative impact on the COVID-19 case rate. We did not expect that public health education would be successful in warning people at risk. We think this maybe because people at risk are being more careful and are less likely to contract COVID-19

Team Members

Ram Ben-David, Dwight Liu, Susmita Padala, Kineret Stanely