Dondena Seminar Series
Conventional wisdom tells us that women live longer than men because women don’t smoke. Every population, everywhere, we are told, women smoke less than men and thus live longer. But Marc Luy of the Vienna Institute of Demography didn’t buy it. And once he drilled down through previous studies and different data sets, he came to a different conclusion—one that reinforced how important it is to consider that there is a personal story behind every data point.
Give us a summary of your talk today. If people are going to miss it, what are they missing?
They miss new findings about the role that smoking plays for the differences in life expectancy between women and men. It is not as clear as it is commonly stated.
So, the message that you get when you hear someone speaking about gender differences in mortality, or the message that arrives everywhere, is that smoking is the main factor that explains everything, and that this is true everywhere. I once heard a presentation of a very famous demographer who spoke about mortality trends, and then at the end someone in the audience asked why it is like that, that women live longer than men. He said there are three factors that can explain this, ‘the first is smoking, the second is smoking, and the third is some other small things that also contribute a little bit.’ And this is the general picture that everyone has about the causes of gender differences in life expectancy. But the more I thought about this the more I had doubts that it really is so easy to generalize. First of all, we know that populations differ in the time and the speed that they go through the smoking epidemic. But smoking prevalence is different between populations. And second, when we speak about life expectancy, we mostly think of life expectancy at birth. But smoking starts to become effective for mortality from about age 50 onwards. We wondered whether this one factor can really be so strong that it outmatches all the other factors over the whole life.
So we started to look at that in a systematic way. First, we looked at the literature and found that the existing studies cover only a few populations and that they are very different in the age groups they analyze and the methods they use. It’s not so easy to get a comprehensive picture. So we did a lot of work to analyze as many populations as possible. We looked at trends in life expectancy at birth over 60 calendar years and we used the same data and the same methods for more than 50 populations. And at the end we found that the picture is in fact very heterogeneous. It can be misleading to give the message that smoking explains everything everywhere.
How is this message generally received?
The paper is not yet out, it is just submitted. We have presented the paper on a few occasions and, yes, people start to think ‘Why is it like that?’ They start to think about a specific population and they come up with some explanations. And this is exactly what we wanted, that you have to look at each population separately and think about their stories. There really are individual stories behind what’s going on in the populations. We cannot say that it’s smoking everywhere. Everywhere the situation is that the gender gap in life expectancy was increasing and then it started to decrease until today, and in most of the populations this change is due to smoking, that’s true. But the extent of the gender gap is only in the minority of the populations due to smoking, so there are many other factors that are important too. I think it can easily happen that these other factors somehow go unnoticed.
It is also interesting that when we look at the importance of smoking, it’s decreasing in each population. It’s becoming less and less important what is in line with the well-known model of the smoking epidemic. But when we look at the other factors that are not caused by nature, they are becoming more and more important in most populations.
Yeah, as well as alcohol consumption, stress, occupation-related mortality, etc. We also try to further disentangle these factors, but we haven’t done this yet. Until now we just compared smoking with the group of other factors that can be somehow influenced by human action, directly or indirectly. And these factors are becoming more and more important. As long as the message is going around that smoking explains everything, I think there is the danger that this is not realized.
I was at the doctor this week and the only lifestyle question she asked me was, ‘Do you smoke?’ I said, ‘No, of course not.’ And it was as if everything else was fine.
She didn’t ask about your coffee consumption?
Um…No. Should she have? Please don’t tell me that!
My wife is Italian, this is why I know a little bit about Italians and their coffee consumption. And she told me that doctors often ask how many coffees you drink in a day. I found this very funny because I was never asked that. But of course, knowing Italy better now, I can understand that. Italian coffee is really incredibly good.
Tell me about your data and tell me about how you did this research.
We used macro data from the WHO: number of deaths, the living population, and causes of death. And from the information on causes of death we estimated how many of the deaths are due to smoking. There are some methods proposed for doing that. Several of them are based on lung cancer mortality and smoking prevalence in a population. And then there are methods that are a little bit different, but in the end they do not differ that much in their results, so estimating smoking-attributable mortality seems to be quite reliable. We used one of these methods based on the causes of death data to estimate the smoking-attributable mortality of women and men for each population, and then we translated that into life expectancy differences between the sexes.
In other words, we decomposed the life expectancy difference between women and men into years caused by smoking and other factors. The very new aspect in our study is that we isolated the non-biological part of the gender gap. There are biological factors that contribute to the differences between women and men, such as genetic factors and hormonal factors. For estimating their impact we used the data from our “Cloister Study”. I like this sample of Catholic order members. It’s a fascinating group of women and men. The original idea of the Cloister Study was that here we have a group of women and men who are very similar with regard to mortality risks. They have the same daily routines, they have the same beliefs, occupational risks are not so different, there are no socioeconomic differences, there are no marital status differences, so they are very, very similar.
And in fact, the difference in life expectancy between women and men among Catholic order members is only about one year. This gives quite a good idea that biology doesn’t play such an important role for male excess mortality, because if it were mainly biology then monks could not escape this burden. We used this data to estimate the effect of biology and from what remained we isolated smoking, and this is how we came up with the results I am going to present today. There were some smoking studies before, but not that comprehensive for so many populations and for such a long time span and with the same data for all populations.
How does this fit in with the overall picture of what you do in your research broadly speaking?
Like most if not all aging and mortality researchers, we want to find and understand the mechanisms behind aging, healthy aging and longevity. My research philosophy is that we get the best answers by looking at differentials in health and mortality, such as differences between the sexes, between social groups, between regions etc. For example, when we understand why women live six years longer than men then we understand a lot about the basic mechanisms behind aging in general. And this whole process is very complex. There are so many factors coming together and interacting with each other, and in different populations they do this in different ways at different times. The gender differences are probably at the heart of all these phenomena of differentials in health and mortality, because everyone belongs to one of the two groups. But as always, it starts with very easy questions and then, the more you think about them the more complex they become. It seems you never get an answer to what you are thinking about. But we are optimistic.
What do you like about what you see going on in your field right now, and what do you wish you could see more of? Either in the types of research people are doing or the kinds of data they’re using to do it?
I like that we deal with the most important problems of the people, health, and that we might help to make lives better. This makes our work really important.
What do I wish to see more? From my perspective, in the research of health and mortality and their differentials, the problem is that we are still too focused on the risk factors. What is missing, according to me, or isn’t yet developed enough, is that we are looking more intensively at risk groups. For example, in the gender differences. In general, women live around six years longer than men. But it’s not that this applies to every man and to every woman. There are specific risk groups that are causing this difference and the trend in the difference. I think we should do more to identify these risk groups and study why they are having such a high or low mortality. Regarding the gender differences we have just published a paper in GERONTOLOGY in which we hypothesize that it’s mainly specific risk groups of men with very high mortality who are responsible for the extent of the gender gap. At a first glance it seems that these risk groups have nothing in common. But I think at the end their high mortality is based on the same mechanisms, and I think the key factor behind all these differentials might be socioeconomic status.
And then it would be necessary to combine our research on risk factors or risk groups with health behavior theories. We know now that most of the differences in health and mortality are caused by health behaviors. Biology is not that important. There are of course also environmental influences, but most behavior is done by humans directly or indirectly. And when we think of the behaviors, there must be some reason or motivation why people behave in a specific way. Everybody knows that smoking is harmful, but still too many people are smoking. There is some decline in smoking prevalence but this decline doesn’t really reflect the knowledge about smoking. This means that for many people there must be something that is more important than knowing that what they are doing is really unhealthy and most likely shortening their lives. Health behavior theories might help us to better understand this. This is what I’m missing a little bit in the field. Our research is still too much data driven.
Learn more about Marc Luy here. Learn more about the Dondena Seminar Series here.
Last updated 10 December 2016 - 05:39:04
Dondena - Università Bocconi