Project
As a child growing up in Manhattan, New York City, I was able to first handedly witness the economic and health fields change with demographics simultaneously. It is often status-quo for counties with higher minority populations to have higher unemployment rates and lower income and health insurance coverage rates. The economic disparities seen in some of these counties are not unintentional, but instead a demonstration of historical redlining, public policies as well as systemic discrimination. Specifically in New York and New Jersey, the high cost of living, particularly in housing tends to be a major barrier to economic instability. As income inequality remains persistent, as does health insurance access as many New Yorkers and Jerseyans face barriers when qualifying for health insurance coverage. The cumulative effect of these factors is a significant racial wealth gap that inhibits the future generations of the North from the intergeneration wealth transfer that other counties have access to. These disparities warrant not only action for policymakers but also further discussion on the ways in which these counties regenerate the wealth loss to discrimination. The variables of interest I’ve chosen to explore demographics are racial makeup, voting population, total population and gender. To better understand economic disparities the variables I’ve chosen to explore are median and mean household income, population of employed and unemployed, population in and out of labor force, mean Social Security income, and the population with and without health insurance coverage. Using these variables my goal for this project will be to determine how economic and demographic factors shape healthcare outcomes in 59 New York and New Jersey counties.
To begin my analysis on the economic and demographic disparities that shape healthcare inequality I chose to create a correlation heatmap. This graphic provides a great overview of the relationships between the variables of interest in my dataset. As my project aims to weed out the most impactful variables, this heatmap is a great tool in helping to identify the strong relationships that should be further explored. The highest correlations are seen between employment status, coverage and total population suggesting further exploration using county-level comparisons. Suprisingly, mean income has the lowest correlation with every variable, alluding that it may not be as powerful as a contribution as the rest of the variables. This visualization acts as a “map” for the rest of my discussion and aided in providing me the best route forward when it comes to exploring variations across geography. In the New York correlation map we can see that employment status has a higher correlation with health insurance coverage than it does in New Jersey which is most likely due to population size. Lastly, this visulization has allowed me to conclude that the population variable will hold to be one of the most influential factors in ensuring healthcare access.
First I want to explore the relationship between median household income and health insurance coverage as it yeiled a low correlation in the beginning analysis. In order to interpret this in accordance with other earnings amongst the population I created a dashboard to showcase insurance coverage rates by specific income thresholds. According to the visulazations, low income communities have more fluctuation in coverage rates in comparison to average and high income counties. Suprisingly, the counties with the high income had the highest rate of people not covered by insurance. Consequently, counties with low household incomes had the lowest rate or people without insurance, illuminating that median income may not be such an impactful factor as confirmed by the earlier correlation. Out of the low income counties, CUmberland County, saw the highest rate of uncovered people whereas Chemung County, New York witnessed the highest rate of insured people. In the average communities Passaic County, New Jersey saw the highest rate of uncovered people whereas Erie County, New York witnessed the highest rate of insured people. Lastly in the high income counties Union County, New Jersey saw the highest rate of uncovered people whereas Hunterdon County, New Jersey witnessed the highest rate of insured people. This lack of impact on insurance coverage could be due to government assistance with healthcare as well as job benefits that often allocate a portion of earnings towards household insurance. However so, the extremes in the visualization allow for further discussion of other variables that do hold impact on insurance coverage and how they might be key in remdying its disparities.
Tableau Dashboard
Next, I explored the correlation between employment status and insurance coverage through a scatterplot. Employment rate and insurance coverage are seen to have a weak negative relationship, which intrigued me further. I chose to also create a scatter plot showcasing the unemployment rate so that i may better visualize the differences. The distribution between unemployment rate and coverage inevitably turned out to be a weak positive one. In a states as big as New York and New Jersey one would expect to see an increase in insurance rates as employment rates increase because of typical job benefits, however that is not the case here. Without establishing causation, this visual heeds greater policy-relevant discussions such as immigration and wage thresholds. New York being one of the biggest cities also causes it to have a large immigrant population. Thus proposing issues as undocumented and minority populations are inhibited from being covered by their employer. Additionally, high-costs of living in states such as New York may inhibit workers that own a low wage from seeking private insurance coverage. This highlights the need for New York and New Jersey policymakers to better examine industry composition and employment benefits.
To further investigate insurance coverage on behalf of race, I decided to make an animated line plot that shows the distribution of insurance coverage as the black population in counties increase. Compared to other racial groups, Black Americans have historically suffered economic and healthcare disparities due to low wages, marginalization, and a greater exposure to medical debt. Though there are other factors outside of societal control, black people make up a large proportion of these counties as seen in the earlier analysis, and thus deserve fair and equitable treatment. Due to these inequalities and large proportions, I decided to showcase this in contrast to the other races that makeup NY/NJ counties. In the visual, there is significant variation in coverage rates as the black population increases. Moreover, a lot of the spikes in the graph are concave downward, illuminating to high coverage rates in areas with a smaller black population. At the midpoint of the plot the graph shifts with more spikes concave upward indicating lower insurance coverage rates in areas with a higher black population. As the disparities in healthcare access persist in the black population, visualizing these disparities helped to idenitfy whether these barriers were present in New York and New Jersey counties. According to the animation they are.
The chloropleth map of uninsurance rates across New York and New Jersey give rise to discussion of the geographic implications for healthcare access. The darkest parts of the map, indicating high uninsurance rates, are found in counties such as Passaic and Union, New Jersey, both of which have large Hispanic populations. The lightest clustering is found in New York state, indicating that they have higher rates of insured people. These findings suggest that though geographic variation and insurance coverage have a good relationship, demographic composition remains a pivotal factor in shaping healthcare coverage. Although I did not originally plan to focus heavily on race, the choropleth map made the pattern hard to ignore. This analysis was particularly surprising because both states—especially the New York City and northern New Jersey regions—are extremely diverse, so I expected the uninsurance rate to be more evenly distributed across counties.
Chloropleth of Insurance Coverage
I wanted to explore next if there are any noticable differences between the female population in counties and coverage rates. Isolating the gender comparison allows for a clearer interpretation of key trends in insurance coverage that might often be obscured when solely looking at racial demographics. Here, I demonstrated a line plot, that has a size determined by the proportion of females in the population. The plot is faceted by the uninsured and insured rates for that particular county. Though also a weak positive correlation, noticably counties with higher female populations have greater rates of uninsurance.
Gender (Female) Line Plot
In the Coverage Modeling Dashboard I created an interactive regression model that plots health insurance coverage by various other economic factors. With the exception of median household income, all other factors included in the menu yieleded a moderate negative slope. The goal of the app is to reflect broader the strength of the relationship between these factors and healthcare access as well as evaluate them as predictors. In the future, I would move towards gathering more data on Northern states so that a more significant p-value would be shown during this analysis. This dashboard allowed me to visualize the ways in which coverage impacts with other economic factors in a way that I could not predict.
For a more accurate conclusion of my research I created a New York and New Jersey explorer, to not only allow the aduience a chance to interact with demographics I’ve ignored, but also to better showcase the effects of gender on health insurance coverage, as it plays a pivotal factor. Specifically, when all states are selected the graph shows significant variability, especially with gender as a factor. However, when only New York or New Jersey is selected we can determine a clear trend in coverage rates by a specific variable. When mapping coverage rates by housing units and setting gender ratios to the minimum value, New York and New Jersey, had robust positive and negative slopes respectively. This was helpful in confirming my earlier assumption about geographical factors playing a role in shaping healthcare access. To aid further interpretation, I also included summary statistics for each variable included in the correlation analysis which shows that coverage rates in New York City and New Jersey have a narrow rightly-skewed distribution as evidenced by the mean(1.9) and minimum(1.57).
Synthesizing the elements of my project, the results of statistical analysis’ and visualizations support a nuanced conclusion about health insurance coverage in New York and New Jersey. The data has shown while health insurance coverage isn’t primarily driven by median household income levels, it is driven by the intersection of population density, gender, local labor markets and polocies, as well as race. My study did also not do well in accurately showcasing the rate of uninsured Black and Hispanic populations due to the small sample size. Though unfortunately, the trends in my data are still consistent with historical patterns overtime as minorities have tried to navigate the wealth gap in the North. The results demand intervention into healthcare access and future research on how this can be inhibited from returning. Additionally, the results demonstrate that both New York and New jersey may be due for political reinforcements such as expanding federal insurance eligibility as well as enforcing more equitable practices, in order to make health insurance coverage accessible for all.