Over the past month, American drinkers have experienced whiplash when it comes to alcohol and their health. First came the release of the NASEM report, which proclaimed that while the researchers could find a slight increase in breast cancer from alcohol consumption, they also found a decrease in cardiovascular risk. Then there was the surgeon general’s recommendation that a warning be placed on all alcoholic beverages that no amount is safe. Finally, there was the release of the controversial ICCPUD report which fully contradicted NASEM. It’s no wonder that consumers of alcohol in America are extremely confused.
All of the studies and recommendations are based on a set of data that scientists, researchers, and doctors have been analyzing for the last few years. What’s unclear to most people, however, in the release of these reports is how this data was collected, what the data is trying to do, and what actual takeaways and behavioral changes alcohol consumers should make based on the findings.
To try to cut through the noise, we sat down with economist Dr. Emily Oster to help us make sense of it all. Dr. Oster is the founder of ParentData and is on a mission to empower people by providing the data and tools they need to make confident decisions. After getting a PhD in economics from Harvard, Dr. Oster went on to pursue research in health economics and is now a professor of economics at Brown University. She is the New York Times best-selling author of “Expecting Better,” “Cribsheet,” “The Unexpected,” and “The Family Firm.” Dr. Oster uses her expertise in reviewing and analyzing data to help others navigate topics related to health and parenting.
What follows is a transcript of our conversation. If you would like to hear the full conversation, you can listen to it on the VinePair Podcast. Links to the episode can be found at the bottom of the page.
Adam Teeter, VinePair Co-founder and CEO: Thank you so much for taking the time to share your thoughts on these alcohol studies and announcements that are making real waves right now.
Dr. Emily Oster: Thank you for having me. I am happy to be here to talk about science.
AT: So there have been a lot of reactions to the studies that have come out recently, and while we know that most people have a general understanding that alcohol is not something that is ultimately good for them, many people right now are really shocked, because the articles and proclamations that are coming out around these studies are really alarming, and people don’t know what to think.
Individuals don’t understand the data that’s being used to come to the conclusions in these studies, or how that data is being collected, tabulated, and then analyzed. Can you start by explaining, at a basic level, what these studies are?
EO: Yeah. I mean, it’s a big question. I think there’s a lot of ways to start.
But I think one way to start to frame these studies for people is to say, if you wanted to know the health effects of alcohol, what would be the ideal way to learn about that?
And the ideal way to learn about that would be to take an infinitely large sample of people, randomly assign them to consume different amounts of alcohol over the course of many, many years, and then later analyze their health outcomes. And if you could randomly assign them and follow them forever, you could be very confident in establishing what we call a causal link between alcohol and health.
And there are a lot of reasons we don’t do that kind of study, but I think it’s worth keeping in mind that that’s kind of what we’re trying to replicate. We’re trying to understand the relationship between alcohol consumption and health, holding constant everything else about people. So sort of comparing two people who drink different amounts, but in other ways are identical.
And that’s what we’re trying to do when we’re talking about studies of alcohol and health.
AT: And why is a study like this so hard to do?
EO: There’s a few different reasons. One is that if you wanted to follow people for their whole life, you would be dead before you found out if they were dead. It’s also hard to get people to change their alcohol consumption behaviors in either direction. So I think there are some practical barriers to doing this in a randomized way.
The reality of what these studies do is they go out and ask people, how much alcohol do you drink? And some people say, I never drink. And some people say, I drink a little bit. Some people say, I drink a lot. And the researchers record this.
And then, in the best of these studies, they will then follow those people over time and see, do they die of heart disease or do they develop cancer? And so, what you see in the studies is that there are differences in cancer development, say, among people who drink not at all versus a little bit versus a lot. And there are differences in things like heart disease.
The fundamental problem is that those people are not the same in other ways. So, people who drink, who abstain, tend to be different in terms of their education, different in terms of their smoking behavior, different in terms of their dietary choices, different in terms of their income and their health resources. All of those variables are also varying across people in these different groups.
And so, looking at those, comparing those groups and saying, OK, any differences I see in health, I’m going to attribute to alcohol consumption, that’s actually a very, very hard claim to make. Because how do you know it’s the alcohol and not the smoking? How do you know it’s the alcohol and not a lack of access to resources?
How do you know it’s the alcohol? That’s really hard to separate. And it’s especially hard when we’re talking about moderate drinking.
So I think this is actually a really important distinction to make. There are a lot of reasons to think that drinking a lot is bad for your health. There are a lot of very strong correlations between drinking a lot and many health outcomes.
And we sort of know, at the limit, drinking too much alcohol causes liver disease, has all these negative impacts.
Where there’s more disagreement, both in the data and in the experts, is in what to make of the kind of small differences between someone who has, say, one or two drinks a day versus someone who has none. Those differences are very small in the data and it’s really, really hard to know if they’re about the alcohol or about something else that’s different across those groups.
AT: When people are looking at the surgeon general’s recommendation, are we to basically assume that the reason that recommendation is being made, is just because since there is some sort of a risk, the medical establishment has at this point decided they have to basically alert us to that risk, even if we can’t quantify how great of a risk it is to us?
EO: Yes, I think that’s right, although I think that the thing that’s a little weird about that argument, which I think is basically a correct description of why they’re doing this, is that if you take this kind of observational data, where you observe how much people drink and you observe their health, when you look at that data, you see that for cancer, drinking a little bit is associated with slightly higher risk of cancer relative to not drinking at all. When you look at heart disease, drinking a little bit is associated with a lower risk of heart disease than not drinking at all. And in fact, if you look at what we call all-cause mortality, so just dying, actually drinking a little bit is associated with a lower risk of dying.
Now, I personally do not think any of those correlations are particularly likely to be causal. I think they are almost all reflecting other differences across the groups.
But, what’s a little odd about some of the messaging is that if you were going to say, I really believe this correlation about cancer, then I think you have to also believe the correlation about heart disease, and all-cause mortality which go in the other direction.
And so there’s something a little confusing for those of us who are steeped in the data about saying, I’m going to believe one piece of this data but not the other piece. I’m going to believe this one piece because it’s consistent with some other set of beliefs I have, but not so much because I am learning from the data.
Joanna Sciarrino, VinePair editor-in-chief: That’s really interesting. I hadn’t thought about it like that. And actually the leading cause of death in America is heart disease, so do you drink moderately to protect your heart or not at all to prevent cancer?
EO: If you ask me in my heart of hearts, as someone who spends so much time in this data, what do I think the actual truth is? I think these effects are basically zero in either direction. That drinking a little bit is basically not positive or negative.
It’s just about zero and certainly well within the other risks we take every day in either direction. But I don’t think anyone wants to hear that. Somehow, no one wants to just be like, it doesn’t really matter. They want it to be one of the directions.
AT: But people want to know, how many drinks can I have, no one wants to just be told to moderate, because everyone defines moderation differently. Does the data give us any insights here?
EO: If you looked in the kind of graphs we are examining in these studies, which again are based on this observational data, I’m not sure how much we can take from them, but if you sort of look at where the correlations between, say, drinking and cancer sort of start to go up more, it’s kind of around three or four a day.
So the current government recommendation we have, which is the common recommendation, which is one drink a day for women, two drinks a day for men, is probably not an unreasonable definition of moderate. But, I think the data is completely incomplete to make a claim across the board like that. I do think there’s a piece of this which should be so much more personalized than it is.
There are people who would say, I have two drinks a day and I’m not comfortable with my relationship with alcohol. I feel too wedded to that relationship and I feel like I should cut down. And that is a reason to cut down on your alcohol consumption, even if two drinks a day was sort of fine from a health standpoint.
And I think that we’re sort of missing in this conversation is that we’re looking for something that’s going to be driven by health, but really some of this should be about preferences and about sleep and other things which have nothing to do with what’s a correlation in some large data set.
AT: What about this idea that some people have that wine is better for you than other types of alcohol? Do we see that in the data and could we ever design a study that proves that?
EO: It would be technically possible to prove this. And I think there is enough data in these large data sets, that you could separate people based on how much wine do you drink, how much beer do you drink, how much spirits do you drink.
But the issue is that the people who are consuming these different kinds of drinks are themselves different.
As I said previously, we already know there’s an issue with people who consume alcohol versus those who don’t consume any. There’s also an important distinction between people who consume different types of alcohol.
One of the examples I often give is if you said, you know, what is the health and longevity of the people who drink 20-year-old Barolos with dinner every night?
The answer would be that those people are doing great, because those people are really, really rich. And they have a lot of resources to stay healthy. You could almost line up the sort of price of your nightly wine with your longevity, and it would have nothing to do with the wine. It would have everything to do with the resources and education and other things that people are, you know, experiencing. And I think that’s an illustration of why this study is extremely difficult to do.
AT: So basically what you’re saying is that if you have the means to go to the doctor, and you go to the doctor more, you’re probably preventing a lot of the issues that you would have if you didn’t go to the doctor, right?
EO: Yeah, exactly. And I think when we look at these kinds of patterns of alcohol consumption and different kinds of alcohol, you’d also see correlations with exercise, you’d see correlations with smoking behavior, things which we know themselves contribute to longevity.
Like you can’t ever have physiological constants to make this study work.
There are things you can observe in data, but there are really important differences between people that, as researchers, you can just never observe or never observe completely enough. And that is what makes this kind of evidence so difficult to learn from.
JS: So where do you think we go from here? Because obviously right now it feels like there’s a lot of the journalism that basically is just saying there’s no way this is good for you at all. Everyone should stop consuming.
What are our next steps?
EO: It’s a really good question. I would love it if we had better studies. So I would love to say, let’s run some randomized trials on this.
I don’t think it’s impossible, practically, to do. The fact that it has never happened suggests to me, though, that there are probably some reasons why this type of study isn’t likely to happen. I wish this conversation would shift into more trying to help people who are struggling with alcohol than to be this sort of zero one messaging.
Because, as with much other public health messaging, when you say every bit of alcohol is terrible for you, the goal of that seems to be to try to get everybody to drink a little bit less.
But the result of this is that people who are already only drinking a little will move to drinking a little bit less. And the health impacts of that are basically none.
And the people who are drinking a lot will not change their behavior at all. And that feels like those are the people you actually want to help. So we’re kind of focused on a message which is only going to reach the people who don’t need it.
And so it does feel like maybe a message that would somehow be more targeted to getting people to reduce if you’re having, you know, seven glasses of wine a day or a 6-pack of beer every day, might be more effective than the messaging we currently are working on.
AT: So speaking of messaging, how much does the data show us that messaging like the surgeon general’s proposed warning is impactful at all?
EO: It doesn’t matter at all. This is the other thing about the surgeon general’s idea, which doesn’t make any sense, which is the idea you’re going to put, like, warning labels on this and that’s going to matter. I mean, we know from many other settings that putting warning labels on things is just a completely ridiculous farce.
People will point to smoking, to cigarettes, and say warning labels work but the reality is that part of the reason the messaging about cigarettes worked when it did is that people truly didn’t know it. It was like they had never heard before that cigarettes were bad for you.
The surgeon general was the first one to be like, cigarettes are bad for you. People were like, oh my God, really? I didn’t know that.
But here, it’s like this messaging is just over and over and over again. People are not listening.
Listen Online
This story is a part of VP Pro, our free platform and newsletter for drinks industry professionals, covering wine, beer, liquor, and beyond. Sign up for VP Pro now!
