[MUSIC] Hi, in this module, I'd like to introduce the concept of the ecological fallacy. We've previously talked about how studies are conducted at different levels. We have micro-level studies, perhaps of relationships in variables among people, or families, or small firms. We have studies that are conducted at the macro-level by comparison of countries. And then somewhere in between, we have meso-level studies that look at relationships among characteristics of units like communities and neighborhoods. Now, one important phenomenon that we have found out is that the relationships of superficially similar variables may actually differ across levels. For example, one thing that we're very interested in, the relationship of education to health may differ according to whether you are looking at it at the individual level. That is, whether an individual's own education influences their health. In fact, in almost every society, education is an important influence on individual health. You may get a different picture of that relationship if for example, you look at the average level of education in a country and its influence on these average level of health in a country. Relationships may be weaker at that macro-level because there are so many other influences on population, level of health. Now, that's just an example. Now, an ecological correlation refers to a claim or a statement about a micro-level relationship based on the observation of a relationship at some broader macro-level. The original reason that people conducted studies that calculated ecological correlations, or made use of them, is that, especially in the past, individual level data that would support micro-level studies were actually fairly scarce. So a lot of classic early studies, especially in things like epidemiology, trying to assess the sources of disease risk. Because they didn't have individual level data, they looked at regional or country-level statistics on the prevalence of particular kinds of diseases.. And then look for associations with country-level measures of the prevalence of certain kinds of behaviors or diets. And then tried to draw links at a very macro-level, from which they made claims about micro-level relationships, that perhaps if certain behaviors were more common in certain countries. And those countries were characterized by higher levels of a particular kind of disease, maybe at the individual level those behaviors were increasing people's risk of disease. However, this turns out to be very risky. And we have a term called the ecological fallacy to refer to essentially what happens or what goes wrong when we look at a macro-level relationship. And then from that, try to make some claim about a detailed micro-level relationship. So I'm going to start by talking about a rather abstract example, and then I'll move to a more concrete example later. So imagine we have two types of people in our population, we have red people and blue people. Some people are happy, some people are sad, and we're trying to figure out if being red makes people happier. And assume initially that we actually don't have access to the individual data, we can only look at two groups of people. And then compare their average proportions happier or average proportions sad. So we have one group of people, which is 40% red and 53.3% happy. In a second group of people we have 47.1% red and only 47.1% happy. So it seems like comparing these two groups, not looking at the individual data but just looking at the percentages in the groups and comparing them, the population with more red people is the sadder population. Now, that turns out to be what we refer to as an example of a ecological fallacy. If we look within groups, so we get access to individual data where we can break things down and look at people one-by-one. It turns out that in the first group, among the blue people, 44% of people are happy. And among the reds, 66.7% are happy. Similarly, in the second group, 33.3% of blues are happy versus 67.5% of reds. So in both groups, reds are happier than blues. And in fact, overall, it turns out that reds are happier than blues, 64.3% overall versus 38.9% overall. So, again, this is a abstract example of a ecological fallacy. Where by comparing relationships at a group level, we make an erroneous claim about relationships at a micro-level, that would be revealed if we actually got down and looked at the individual level data. Now there's other examples of ecological correlations that people often cite. One is with party preferences in the United States as a function of economic standing. There are differences between party preferences and its relationship to economic standing depending on whether you look it at the state level or at the individual level. It turns out that if you look at a map of the United States, that poor states are more likely to vote Republican. But, if we break it down and look at individual level data from surveys, it turns out that within most states, within each state, actually, wealthier people are more likely to vote Republican. And what's going on is that there are very big differences between states in the percentages supporting one party or another that are overwhelming these state level differences in economics. Another example that you may have heard about, or seen claims made, and this is just an example, is recently the claims made about The Mediterranean Diet. So, a few years back, people observed that in countries that bordered the Mediterranean, they had diets that were high in olive oil, pasta, vegetables, other ingredients that we tend to think of as healthy. It was also noticed that, or claimed, that these countries had lower risks of heart disease. So, based on, again, this ecological correlation, it was claimed that this Mediterranean Diet was actually a way of reducing the risk of heart disease. Now, it's not clear that this is a ecological fallacy, but it's certainly an ecological correlation. And people are conducting micro-level studies to follow this up and figure out whether this really works at the individual level. And finally, quite often we see, and you may have seen, cultural explanations of individual and family behavior. So, people, in comparing different countries, they notice that certain norms, beliefs, preferences, as a result of culture are more common in certain countries than in others. There are big differences in the traditions, the cultural beliefs between for example, east and west. And then they observe differences in the rates of particular types of behaviors between East and West. It might be divorce, marriage, etc. And then claim that the culture is driving the differences in the rates. This used to be claimed as an explanation for the lower rates of divorce in East Asia up through the 1980s, 1990s, or the 2000s, people observed that divorce rates were low in East Asia. And that East Asia was also characterized by a specific configuration of cultural norms and traditions that made divorce unfavorable. Well, that claim turned out probably to be incorrect. Divorce rates in East Asia have skyrocketed, clearly for other reasons. So, culture at the macro-level is not somehow driving individual level divorce behavior in these situations. So, overall, I hope that in this module I've sensitized you to the issue of the ecological fallacy. And I hope that you'll be aware, when you see claims made in the media about differences between countries as evidence of differences in the risks of particular disease linked to particular kinds of behavior and so forth. That you'll treat those claims skeptically and look for more detailed data based on micro-level studies.
