Regression?

(click on the graph to see it more clearly.)

I came across this article from the Economic Policy Institute the other day, and immediately I said, "Damn! I had no idea that health care disparities were so much more pronounced in Hispanic communities than Black communities." Then, immediately I thought, "Really? I wonder if they are lying with numbers." Or perhaps more appropriately, I realized that I didn't understand what these numbers meant.

It has been an enlivening experience to learn how to conduct quantitative analysis, manipulate data, and take weeks or months of research and condense it into a couple of graphs. Obviously, I've learned that some things necessarily have to be left out, but the more powerful lesson (the unspoken lesson) is that the researcher gets to decide what is left out. And, what I tell you or don't tell you about my numbers will lead you to different conclusions.

Since we are so programmed in our society to see and relate to numbers as objective, absolute, and irrefutable, they stand as proof that something must be true. We've all heard that 94% of statical citations are made up on the spot, and that makes intuitive sense because we've all had to best guess percentages throughout our lives.

- "About 30% of the folks there were women."

- "How come I'm taking out the garbage 70% of the time, but I'm only responsible for 30% of the trash that we throw away? Huh? Tell me that!!!"

- "I don't feel like this relationship is 50-50. At best, and I mean at best, it's 65-35, with a 3% margin of error."

We've all said each of these things many times in the last month, so we know that someone throwing out a statistic may not actually mean that much. But, what about a graph? A graph must be reliable, right? Graphs are bullshit too? Ah- but what about a graph based on research and a data set? Yeah, what about that!?!

Can't trust it. (Thanks, Chuck D.)
Don't believe the hype! (Calm down, flav.)

From this graph, it's clear what these numbers are communicating, but I have no idea what these numbers actually mean. I would like to see the regression that was done to determine whether the researchers accounted for factors beyond race in this study.

Hispanic is distinct from White and African-American in that Hispanic communities include a significant number of undocumented immigrant laborers. So, if these folks are included in the study, certainly that will skew the results. Are Blacks "not going to a doctor because of the cost" more often than Whites because Blacks as a group earn less than Whites, or are there other factors at play such that a Black person, a White person, and a Hispanic person each earning $35,000 a year would be just as likely to see a doctor? Are Black women (as a group) earning $40,000 a year more likely to look for home remedies than White women (as a group) earning $30,000 a year? If so, then this graph may be indicating something about racial disparities, otherwise it may just be communicating class-based disparities.

Are there differences in health status between racial categories? (Research has pointed to "yes" for a long time.) And, if these disparities exist, are the health problems common to Blacks more expensive than Whites' and less expensive than Hispanics', such that a Black person earning $40,000 a year would expect to pay $137 for a visit to the doctor, pharmacist, etc. while a White person earning the same wage would expect to pay $50 for seeking medical treatment? So, in addition to class-based differences, we may also be looking at the effects health status differences between groups. So, there's a second dimension.

Are folks across racial lines who earn the same wage, and have equivalent health status, showing differences in their willingness to carry health insurance? People without insurance go to the doctor less often because of cost than people who are insured. So, are we seeing that in these numbers?

If none of that has been accounted for, then the graph, although it's pretty, isn't really helpful. It's telling you that Blacks and Hispanics have greater health care insecurities than Whites, but that may not be "true". From a policy perspective, the response would be to target Black and Hispanics with health policy initiatives, when what is required would be a policy that addresses disparities in health status, and insured status, and income as an obstacle to obtaining health care.

In essence, I'm asking for a multiple regression analysis. I was going to post a link to some site that has a really tight cogent explanation of what regression analysis is, but all I could find was stuff like this or this. So, I figure that anyone who has read this far can take it upon themselves to google it, because honestly- no one is going to learn multiple regression from a blog site anyway. I mean, I had to take Stats II to get it, and I have sexy eyebrows!

So, that's what I want to know. What do these numbers mean? I looked up the original study from the Rockefeller Foundation to see if I could find my answers there. No haps.

The intention of the graph is to support universal health care coverage, which I'm fully behind. Socialized medicine my black ass all the way to fully insured with no co-pay while you're at it. But, the graph reads as propaganda to me. Even though I support its ultimate aim, it's an example of lying with facts. And, propaganda, even in service to objectives that I support, turns my stomach; not in a condemnatory way that's fueled by disdain, but in a, "I just stopped breathing for a second and didn't even realize it"-Orwellian kinda way.