How stubborn data changed the way we think about money and health
The ‘Hispanic health paradox’ shows that scientists should follow the data, not their biases
Scientists have a tendency to want to discredit and scoff at unexpected results. The gold standard in research is a result that can be replicated repeatedly over time, so we enter into research with expectations that can sometimes shut out the possibility for discovering something new.
This bias occurs across scientific fields: experimental evidence published as recently as late 2016 dispelled the dogma that women cannot create new eggs after sexual maturation. Another example is the several times that (chicken) eggs have been condemned as bad for our health, most often falsely attributed as raising serum cholesterol levels due to their high levels of dietary cholesterol. Both of these assumptions shape the way our society and markets operate and create a possibly faulty foundation for future research.
Health and wealth
But this isn't a new issue for researchers: we've been grappling with our desire to prove our hypotheses for decades. A foundational example was Kyriakos Markides' findings that Hispanic Americans flouted all working theories of the relationship between health and wealth. And it remains a powerful reminder of the fact that truth often comes with asterisks.
In the mid-1980s, Markides, a professor at the University of Texas doing research on the local population of Mexican Americans, was certain there was a strong connection between socioeconomic status and health. The more money and social resources you have, he reasoned, the better your health would be. He even wrote a paper in 1977 criticizing methodology for studies looking at the relationship between socioeconomic status and infant mortality.
Despite other research indicating that the relationship between the two was weakening, he maintained that it was in fact stronger. He made this argument by condemning studies' use of small cities and small populations that made using of certain statistical methods impossible. Attempting to prove his complaint, he published his own analysis that concluded a strong relationship continued to exist. He remained firmly in the “socioeconomic status predicts health” camp – as infant mortality is a popular litmus test for overall human health.
He had already stuck his head out in publishing the paper that defended the strong correlation between socioeconomic status and infant mortality, so he continued to orient his research to assume the theory was correct. But something wasn’t adding up. According to his theory, low socioeconomic status should equal high infant mortality rates. But the data he was collecting was not producing the strong correlation he expected. He dug deeper and finally realized that if he separated out the data by race (known as stratification), he could recover some of that perfect line. For black and white Americans, the expected relationship rang true. It was data on Hispanic Americans that were throwing off the results.
The Hispanic health paradox
In 1986, in the face of compelling evidence, Markides changed his focus and published a paper that made waves and coined the term 'Hispanic health paradox.' Drawing on dozens of studies, he showed that Hispanics paradoxically had similar, and sometimes even better health than white Americans, despite on average having a much lower socioeconomic status. This went beyond infant mortality, with similar outcomes between the two groups for total mortality, cancer, and cardiovascular disease.
Other scientists immediately pushed back, unwilling to let go of the conventional wisdom Markides himself had long defended. They offered a host of alternate reasons for Markides' findings:
- Healthy migrant theory: only healthy individuals are able to immigrate into the US, biasing the health data
- Salmon bias: sick immigrants tend to go to their home country before death, thereby omitting their death counts from the data
- Additional data bias: perhaps Hispanics are more likely not to be in official government databases, perhaps especially those that are in worse health
These rebuttals represented a standard, healthy part of the research process. Except that none of those theories have held up when they've been investigated. To this day, most of the findings from the 1986 study persist (it is worth noting that a few diseases, such as diabetes, are worse for Hispanics – and that Puerto Ricans do not follow the same health pattern). By following the evidence and keeping an open mind despite his initial opinion, Markides changed the very core of how we look at the interplay between ethnicity, socioeconomic status, and health.
So, make space to explore paradoxes – your inconsistent data could become new conventional wisdom for future generations.