Democracy and Health II

This post is going to be another heavy slog through statistics. But I hope you’ll agree it is in pursuit of something important—imagining a healthy American democracy. So, I’ll start us off with a story that highlights the threat to it. In 1968 when I was only 12, I met one of my uncles from Yazoo City, Mississippi. Many in my family had ties to the White Citizens Council and other enforcers of racial segregation and I suspect he might have been one, but I didn’t know that then. I stopped at his house to introduce myself and we talked on the porch. He didn’t invite me in, which then was unusual in the South. He knew, of course, that I was a Yankee and the son of a civil rights supporter so even though I was only a kid, I could tell my presence made him nervous. So much so that he quickly began—strangely since this was a family visit—to explain to me his stance on the Voting Rights Act, which had only just been passed. He began by professing his love for the Black people of Yazoo County, but then descended to protesting that they must never be allowed to vote. He was soon yelling and shaking his fist, “I’ll be a dead man before a Black man votes in Yazoo County!” he shouted. Of course he did not die, and soon a Black man would become the mayor of Yazoo City. But the point is, preventing other people from voting has long been a way of life in this country. Now back to the science…

My last post, more than a year ago, was about the Health and Democracy Index, now produced by the Institute for Responsive Government. The Index (I’ll call it that for short) has a simple purpose, which is to demonstrate that when democracy is stronger people are healthier. It does this very well. States in which more people can vote tend to have residents who are much healthier than states where it is harder to vote. In the post I argued that this is at least in part because when people can express their preference for health through the ballot, politicians are forced to listen. 

I serve on the Board of Health of the Ingham County Health Department in Michigan. That means I am privileged to work with their talented staff, and we noted that the Index used data only at the national level. Since we work in local public health, we wondered together whether what the Index shows also holds true at the state and local level as well? In other words, within a single state or county, do areas where more people are civically engaged—voting for example—have healthier people than areas where folks are less engaged? 

We found that the answer is yes for the local level; neighborhoods differ in the same way as states. However, there are a lot of interesting details, so I encourage you to read on. 

I’m not going to describe the original Health and Democracy Index completely in this post. You can read all about it here, and anyway, I want to use this post to fully describe what we did to create the local version. But let me begin by showing you the main chart from the original national index. 

The cloud of data points represents the fifty states. The horizontal axis is an index composed of measures of state electoral laws and policies. Toward the left are states with laws and policies that tend to restrict access to voting and toward the right are states with easier access. The vertical axis is an index composed of common measures of health, with healthier states being toward the top. So you can see that states with more access to the ballot on the right side of the chart tend to cluster at the top, meaning the stronger a state’s democracy the healthier its residents are. For the stats nerds (who always complain if I don’t do it) I have included a regression line and the relevant statistics for this chart. If you just want to contemplate one number, look at the one called R2 (r-squared). It is .46 which means the relationship between voting and health at the national level is very strong. 

Before I go on, I want to clarify what I am claiming here. We are not arguing that access to the ballot by itself is responsible for all differences in health between states. What we are looking at are consistent and powerful historical differences between regions of the country. Different groups of people occupied different parts of the country and subsequent social and economic development transformed regions of the country differently. In brief, Appalachia, the Mississippi Delta and the southern border wound up with political systems more concerned with maintaining social control than other areas—control of unruly rural Whites, African Americans, Hispanics and Native people—and less concerned with winning support by demonstrating the ability to improve people’s lives. In other words, historical conflict created both the voting laws and health policies of the states. Once in place, the voting laws probably tended to reinforce health inequities. After we look at the data on the relationship between voting and health at the state and local level we will discuss how this may manifest itself even at the most local level. 

We decided to try to replicate the chart from the original Index twice: once at the state level and once at the county level, giving us three similar charts to compare. However, we had to do it differently in each case because of the difficulty of obtaining appropriate data at each level. The table below summarizes the differences between the measurement and results of the three charts and these will be discussed more below.  

Attributes of the Three Charts	National	State	Local
Measurement of Health	Index of Health Data	Stroke Mortality	Life Expectancy
Measurement of Civic Engagement	Index of Electoral Policies	Percent Voting	Percent Voting
Units of Analysis	States	Counties	Census Tracts and Precincts
Strength of Relationship	Strong	None	Moderate

Here is the chart replicating the Index at the state level. In this case the state is Michigan. The dots represent Michigan’s 83 counties. 

The horizontal axis is the percent of eligible voters in each county who voted in the 2024 general election. Some counties had very high voter turnout, around 90 percent, while others had very low turnout, around 60 percent. The vertical axis is a measure of health, the stroke mortality rate per 100,000 people. Stroke mortality is a robust indicator of population health that was included in the original health index. The counties with the highest stroke mortality rate (near 50 per 100,000) had rates almost 5 times higher than those with the lowest (near 10). 

There is one important difference between the display of the national and state level data. The index of health in the national data is positive—that is when it is higher it is good. But stroke mortality is negative. When it is higher that is bad. So we flipped the direction of the vertical axis in the state data so the two charts would look the same and be easily comparable. 

What you can see is that the relationship between civic engagement and health that we saw in the national data is almost gone in the state data. The data do not cluster and the regression line is almost flat. If you want a number to confirm what your eyes tell you, the R2 is only .05, almost zero. 

The reason the relationship between civic engagement and health has disappeared is because many of the low-income counties in Michigan that have high stroke mortality rates also have high voter turnout. For example, Montmorency county had the highest stroke mortality rate at 48.7 per 100,000, but it also had sky high voter turnout at 79.7 percent (Montmorency is the dot in the lower right hand corner). Rural Michigan counties are not like Appalachia or the deep South. They may have health problems, but they can also have high levels of political participation. This is at least partly because what counties must do under Michigan electoral law is set by the state, so it is no wonder that what we saw occurring between states has disappeared when we look within one. Of course, we only looked at one state; there may be other states, depending on their history and laws, where voting is related to health. But here we have showed, at least, that it isn’t necessarily so. 

Surprisingly then, the relationship between civic engagement and health reemerges when we look at the most local level, within one county, in this case Ingham County, Michigan. Before we look at the chart let’s look at two maps to get an idea of what our County level health and voting data are like. The map below shows life expectancy by Census tract in Ingham County. 

The data in the map are grouped into equal sized quartiles and color coded to make it easy to see where life expectancy is longest and shortest. Darker colors indicate longer life expectancy. The white spaces had too few deaths to calculate life expectancy. The tract with the longest life expectancy is an affluent suburb of East Lansing with an average life span of 86.6 years. The tract with the lowest life expectancy at only 69.7 years is an urban neighborhood located in central Lansing. 

The next map shows voter turnout in the 2024 general election by electoral precinct, also color coded in equal sized quartiles. Voter turnout varied widely with the highest turnout being in Meridian Township, an affluent suburb where 86.4 percent of those eligible voted. The lowest turnout was in Central Lansing at only 42.9 percent. 
 

When you consider the two maps together, they appear to be similar in some ways. The longest life expectancy and highest voter turnout (dark colors) are both found in the northeastern part of the county where home values and incomes are high. The lowest life expectancy and lowest voter turnout are in the urban core. There isn’t a clear pattern in the southern part of the county. So, when we chart the data what will we get? 
 

There is a moderately strong relationship between civic engagement and health at the local level. Neighborhoods with the highest voter turnout tend to be the ones with the longest life expectancy and they are clustered in the upper right-hand corner. Neighborhoods with low turnout and shorter life spans tend to be in the lower left. The regression line is steep again and R2 is .30, not as strong as at the national level, but still large for social science data. 

Before I go on you may notice that some of the dots in the chart seem to be arranging themselves in straight lines. What’s going on? In the earlier charts we had only one unit of analysis: just states at the national level and just counties at the state level. But at the local level we have two units of analysis: Census Tracts for life expectancy and electoral precincts for voter turnout. They don’t line up perfectly. So the chart of local data is asking a slightly different question than the charts of national and state data. At the national level, for example, we are asking if our units of analysis, states, have both better health and easier access to the ballot, or vice versa. At the local level we are asking if one of our units of analysis, Census Tracts, have long life expectancies and are overlapping other units of analysis, electoral precincts, with high voter turnout, or vice versa. This approach is called overlay analysis. 

You can see this clearly by looking at the line of dots at the top of the chart. That is our previously mentioned affluent East Lansing neighborhood with a life expectancy of 86.6 years. It overlaps ten different electoral precincts. Most of the precincts it overlaps had high turnout with the highest being 81.6 percent. But a few had low turnout, including one as low as 58.4 percent where East Lansing meets Lansing. Taking all such pairs of overlapping Census Tracts and precincts, we ended up with 380 data points in the chart. The statistics on this chart tell us that in spite of some exceptions, high life expectancy Census Tracts tend to be overlapping high turnout precincts and vice versa. 

So what does it mean that we find the relationship between health and civic participation at the local level that we first saw at the national level? It makes sense that we should see that relationship at the national level because states developed different electoral and health policies for important historical reasons and they apply them throughout their jurisdictions. But at the local level Census Tracts and precincts are all subject to the same state and local laws and policies. Shouldn’t they all be the same? 

I think two things are going on. One is we know that lower income people tend to be less healthy. Furthermore, people are profoundly sorted along economic lines at the neighborhood level. As we say in local public health, “Your zip code determines your life span”. Lower income, less healthy people are concentrated in places with affordable, often poor quality, housing. This is the result of policy failures that entrench inequality. Furthermore, these people may have an ambivalent relationship with government including voting. While most low-income people do vote, many still might not feel safe or welcome at the polls and even if they want to vote they might not be able to get there. So lower turnout by low-income people means policies that entrench health inequities persist—civic participation or its absence is driving health even at the local level. 

Another possibility is that even though all local electoral precincts are subject to the same policies and laws, there may not in fact be equal access to the ballot. It could be that even though the County Clerk may intend for people in downtown Lansing to have the same access to the ballot as people in the suburbs, things may not have worked out that way. This is something we intend to test in the next half of this project by looking at the locations of polling places, drop boxes, staffing levels, accessibility and hours of operations. We will share our findings with community organizations with whom the Health Department collaborates to see if there is some basis for action to meaningfully increase access. 

There have been many partners in this project. The Ingham County Health Department has taken the lead, and will present its findings at Michigan’s Premier Public Health Conference in October. The Michigan Public Health Institute has advised on the project and is interested in determining whether it can encourage other local health departments to adopt this approach. The Institute for Responsive Government has offered a great deal of technical assistance. And Barb Byrum, the Ingham County Clerk has helped us understand the availability and use of electoral data. If things go as planned this Fall, we will have plenty of input from community organizations on whether and how to help increase access to the polls in Ingham County and Lansing neighborhoods.  

Sources: 

• Michigan 2024 Census Tracts and Voting Precincts, Center for Shared Solutions, Department of Technology, Management, and Budget, State of Michigan. https://gis-michigan.opendata.arcgis.com/. 
• Michigan 2023 Stroke Mortality Rates by County. Michigan Department of Health and Human Services. https://vitalstats.michigan.gov/osr/deaths/StrokeCrudeRatesTrends.asp.
• US Life Expectancy at Birth 2023. National Association of Public Health Statistical Information Systems. https://www.naphsis.org/usaleep.
• Ingham County 2024 General Election Turnout by Precinct. Ingham County Clerk. https://clerk.ingham.org/departments_and_officials/county_clerk/election_results.php. 
• Michigan 2024 General Election Turnout by County. Michigan Department of State. https://www.michigan.gov/sos/elections/election-results-and-data.