 Good evening today to our public health seminar series. My name's Andrew Neumar. I'm an assistant professor of public health and sociology. And so I'd like to thank you for coming. It's my pleasure to introduce our speaker today, Dr. Raya Matar from the Vienna Institute of Demography and the International Institute for Applied Systems Analysis, which is also in Austria. And we call it YASA, usually. It has this awkward acronym, I-A-S-A. No, it's YASA. So Raya is a PhD or a D-Phil, I should say, in sociology from Oxford University. And before that, she got an MSc in sociology also from Oxford and a master's degree from Waseda in Tokyo and her bachelor's degree is from Chiang Mai University in Northern Thailand, which is where Chiang Mai comes from. And since getting her PhD, she has been a Mary Curie postdoc and a Max Weber postdoc at the European University Institute in Italy. And those are two very competitive postdoctoral fellowships. And today we're going to be hearing about the role of the financial crisis in smoking behavior. So welcome to Irvine. Thank you. Well, before I start, I would like to thank Andrew for inviting me here. It's my very first time at the DOI in college county. So I hope we have a chance to meet this afternoon. And I thank you all for being here. The reason why, oh yeah, this is a paper that I co-authored with Cyrano Gallus and Simone E. Galli. And the reason why I get interested to look at this issue is because we have come across so many literature that report that the kind of procedural relationship between bad political conditions and health. So basically, many literature say that when the economy goes bad, then health gets improved. This is kind of counterintuitive. And it's just what I'm saying that when the economy goes bad, the government has now the public spending of people's time in the pocket so they don't have to spend so much, and you're stressed because you miss that job, and maybe even though you're not losing your job, you might be stressed because of the way you lose your job. So it's kind of a point to the direction that if the economy is bad, one should actually expect to see that your health gets worse. So what we are trying to do is we look at the type of this current financial crisis on people smoking behavior and using the US data. Why it is important to look at smoking as you all know, in the US alone, it is estimated that 400,000 days is attributable to smoking. And because smoking is actually preventable, right? So it's quite an important public health issue. And why it is important to look at smoking especially at this time, because we might expect the government might cut down that spending on anti-smoking campaign. So it is sort of useful to look at how it looks like in the current time. And this is an article and from the report of the leadership of 2011. So the article from the New York Times, the head library, in hope to sign that the health spending is flattening out, the growth rate in this case in terms of growth rate of health spending was as low as millions of Americans lost insurance coverage along with their jobs. Worrying about job security, others may have fear taking time off for doctors visit or surgical procedures or skip from urgent care when money was cut. And also the report for the leadership of the culture relationship between health and this current economic crisis. The economic crisis is expected to produce secondary mental health effects that may increase suicide and alcohol death rates. So from these two courts that I took from the New York Times and the BOS show, it's kind of suggested we would expect to see the counter-servical relationship between health or health behaviors and the current economic crisis. But what kind of empirical evidence that is around there are lots of studies that look at the relationship between economic conditions and health. On the one hand, a set of literature suggests the counter-servical relationship. So they have found that when the economy goes bad, then health gets worse. So this is kind of at least me was expecting to see. So when the economy goes bad, they report the increased risk of overall cost of mortality, increasing suicide. And the recent article reported the reduction of consumption of healthy diet such as fruits and vegetables. But on the other hand, has been much more abundant set of literature that reported the proselytical relationship. So they found that when the economy goes bad, I found there's a decline in all-cause mortality, decline of alcohol-related deaths, decline of traffic deaths, the reduction in alcohol consumption, increase of physical activity. So people exercise more because they have more time, they lost their job, that kind of argument, and the decline of modulate. And with respect to smoking, it has been found that smoking prevents decline when the economy goes bad. But however, most studies, the previous studies that I was mentioning, they only focus on the normal business cycles or because generally the economy is fluctuated, like there's a period of good time and bad time, but they do not focus on the crisis per se. And during the period covered by the previous study from roughly from the 80s until the end of the 2000s, there has been no such a severe economic shock like the current one that we are experiencing. So in a way, the relationships found between health and the usual routine economic cycle might be different under the senior economic conditions like the one we are experiencing. I think my slide is skipping somehow. Should I do something again? Let's see how it goes. But maybe it's all right. Mark, okay. So now about the current relation, about, so then there hasn't been that much evidence on the current economic crisis because it is quite new, but we have found three recent articles reporting the relationship between the current economic crisis and health. These three articles have been looking at different countries and they all reported that how to see the relationship. So health gets bad during this current crisis. They have found the increase in suicide rates, increase in report on medical needs and bad health in Greek, increasing depression and anxiety symptoms in Australia. So what kind of relationship between health, between smoking and this current economic crisis can we expect? We might expect to see the political relationship based on the previous literature that they found that smoking declined when the economy goes bad. Or we might expect to see the counter-sexual relationship because we, based on the recent evidence on smoking prevalence in the US, they have found that smoking increased between 2007 to 2008. The evidence in Italy say, report that smoking prevalence in Italy also increased between 2008 and 2009. And also another article, so report that has been an increase in the sales of tobacco products, which kind of suggests that people smoke more. We might expect to see no relationships at all between the current crisis and smoking. Based on the smoking epidemics model proposed by Lopez, Pito and colleagues, they said that the US has already entered the final stage of the smoking epidemic. So it's a period when smoking prevalence for both men and women has continuously declined. Whereas some other countries such as China, for example, it's still the pattern of the smoking rate for men is higher than women. So women is sort of increasing, trying to catch up with the men, but the US, thanks to the successful anti-smoking campaign, has already entered the final stage. So it's sort of, it has been continuously declined the smoking prevalence. So here I'm showing you what kind of relationship we would expect to see between smoking prevalence and an environment rate. So we are looking at the period between 1984 to 2010, and this is the percentage of current smokers. So in 1984 to 2010. So you can see that it has been continuously declining almost in a linear trend. Well, in some period in the survey, they also changed the question that they asked about smoking, so maybe you don't really see the straight line, but basically if you were to draw a straight line, then it's more or less continuously declining. But here is the unemployment rate, which we use as an indicator of economic condition, just to note that the scale is different here. So over the period, 84 to let's say 2006 to 2007, unemployment rate has been fluctuated, so there was a period of slightly higher unemployment rate, slightly lower, and so sort of falling the normal business cycle. But if we focus particularly on the period around the crisis, so from 2008 here, unemployment rate is about 6%. It jumped to almost 10% in 2009, and it carries over to 2010 and 2011 as well. So at the same time, you can see that smoking prevalence has continued to decline, right? But can we say that there's a relationship between the two trends? We cannot say yet from this two graph. So what we are going to do is we are using the data from the BRFSS survey, which is a cross-sectional data, and we focus particularly on the period around the crisis. So 2005, 2006 to 2007, we identified as a period before the crisis, and 2009, 2010 is the post-crisis period, and then comparing the smoking prevalence between the two periods. So we leave 2008, we omit it, so we allow some window for some change to occur. So we have a sample of two million respondents. We limit our analysis to the US citizen age 18 years and over. And of course, because we are looking at smoking, so we have to limit our analysis to only two people who have valid information on smoking behavior, and as well as people who have information on their age, sex, marital status, education, race, and employment status, which are all characteristics that are related to smoking patterns. And smoking prevalence is defined as the proportion of current smokers to the total number of population. And who are current smokers? Current smokers is people who report that they smoke every day or some days. Unfortunately, we don't have information on how much people smoke, so we cannot identify who are heavy smokers or light smokers. Is the previous slide, the graph? Yes. The US population, the same two million, no, no. Yeah, what was that population? It's also come from this survey. Same two million respondents. Two million is because we pull five surveys, we pull five to two million, so the previous slide, we would have many, many millions. Oh, yeah, so that's a total survey for US population. And this two million is a subset of that, but it's still representative of the US. It is representative of the US of the 15 states, including Puerto Rico, where we took it out. Also, another question about the previous graph. So my understanding is that the smoking prevalence for adults, men and women, but if you were to do that same picture for say, adolescents or younger kids, then you would have different things. Is that true? So this is just because you're looking at people over 18, but if you were actually to include a younger population, it wouldn't necessarily be... Yeah, that's much more of a variation, especially for young females. And the kind of technique that we are using to compare the smoking parents before and after the crisis, we chose to use the non-parametric approach, whereas other studies, mostly they used to kind of fix their progression and look at... Fix the state and fix the year, and then look at the coefficient, whether or not viewing the crisis smoking parents is higher. But we chose our method of using non-parametric one, which I would explain later because we have the sample size is large enough, and also we don't want to rely on many assumptions, and our result could be, and it is replicable as well. And another thing behind the idea of using the method, which I would explain soon, is because we think that the crisis might not affect only the smoking parents, but it also affects the change in the number of smokers. Well, the reason is, as I showed you earlier, that unemployment rate jumped substantially from 2008 to 2009. So the data from the US population in 2008, there were 8.9 million people were unemployed. In 2009, the figure jumped to 14 million people were unemployed, and because different unemployment status have different smoking patterns, so we believe that simply because of the change in the number of unemployed people, then the number of smokers would have changed as well due to this change. Then as I showed you earlier, that was a declining trend in smoking parents. So we used the joint point regression analysis to detect what kind of trend we expected regression may be. Joining this picture is better. The idea behind the joint point analysis regression is very simple, so we just simply want to know whether this trend let me observe it. This is for men and total population and women. Let's see the total population. So we want to know whether this trend is linear or not, and whether it is the, whether there are different slope in the declining trend, but basically what the joint point analysis does is to identify whether the trend that we were observing if there are one slope, so it's declining like this and go up like this or maybe it's just one straight linear one. But what we found, it is simple, so it has been declining steadily for total population, for men and for women for the whole period that we focus on, so from 2005 to 2010. And then we analyzed our smoking parents by calculating the observed smoking parents and the expected one. How do we calculate the expected smoking parents based on different smoking characteristics, based on sex, race, education, marital status, age. So different group of people have different smoking behaviors, so we take that into account, so probably it's better to show you this slide. So our analysis of smoking parents, we take into account the trend that I was showing you earlier and how do we calculate our expected smoking parents? What we do is we stratify each individual into different combination of that characteristic, so the combination of that age, sex, race, education and so on, and then we calculate the proportion of smokers in each of these combination and average them up and then we would get the expected smoking parents according to that particular characteristic, so that's the average. So this is a pre-crisis one and then we do the same thing for the post-crisis one, assuming that the smoking parents is the same, but the characteristic change in the post-crisis. So in a way, if the distribution of these sex, age, race, education are the same in the pre-crisis period, if they are the same also, the distribution in the post-crisis period, then our expected smoking parents between the pre-crisis, post-crisis period would have been the same as the pre-crisis period. Probably it's easier to look at the results. So here, this is a, okay, yeah, I'm showing you first. So now we are showing the observed smoking prevalence in the pre-crisis period and the post-crisis period and of course, this is the observed one and this is the expected one already taken into account the different smoking behavior between different characteristics and the trend, the declining trend smoking prevalence and how did we estimate the trend? As I show you that it's a straight linear line so we estimated the slope of smoking prevalence for each month and what we get is smoking prevalence declined by 0.05 per month. So that's the observed. So the pre-crisis period, about 20% of Americans were smokers. In the post-crisis period, it has declined to about 18% and this declining trend applied to all employment status also. Those who are employed, that smoking prevalence declined. Those who are unemployed, that smoking prevalence declined and that's the same thing for those who are economically inactive. But if you look at the expected one, so we already taken into account the characteristics and the trend, we actually expect to observe the declining smoking prevalence even more in the post-crisis period but that wasn't the case. For the unemployed, this is even more interesting. We actually expected to see, because taking already into account the age, sex, education and so on, we expected to see only 24% of the unemployed should smoke in the post-crisis period but what we observe is as many as 33% of the unemployed smoke. So this kind of suggests that there is some characteristic among the unemployed that can increase their risk of smoking. Before we go too far, I'll ask now, but if the answer is better addressed after the bold thing then I can wait. But in the term that you used to produce these estimates, which is in the previous slide, one thing that happened during the crisis, the economic crisis is also that many states increased the tax on tobacco products to increase revenue. So cigarettes became more expensive and so it would be contraintuitive that people lose their job and they divide this product. So did you consider including in the period that you analyzed the average cost of smoking? We were absolutely right to point that out. So this is kind of the new analysis that we came up but the previous one, we used this big effect regression and looked at the difference between each state, smoking prevalence, unemployment rate, plus the increases of tax, the cost. And we look as well as the change in smoking policy because some states still burn except the smoking ban in public, right? So it's still changing in different states. So we found some kind of expected direction of the effect of corporate tax. But I would like to hear more how we can incorporate this into the kind of non-fermetry approach. But generally it doesn't change the overall conclusion but that is absolutely important. Then after we analyzed the smoking prevalence, then now we're moving to looking at the change in the number of smokers. So how do we calculate the number of smokers in the US? We took the data from the US Census Bureau. So just taking the data of total US population in the period observed and then based on our analysis of smoking prevalence. I'll jump to this slide, which is easier to understand. So what is the change in the number of smokers? It's very simple. So it's the number of smokers in the pre-cuis... The difference between the number of smokers in the post-crisis period and the pre-crisis period. How do we obtain this number? So we take the total population from the US Census Bureau and then we take the smoking prevalence from our own calculation. So we get the number of smokers in the post-crisis period and the number of smokers in the pre-crisis period. So based on this equation, we can further decompose how much of this change in the number of smokers are due to the change in population and how much of this change are due to the change in the characteristics. So for the change in population, of course, because naturally the US population grows and if the population grows, smoking prevalence stays the same, then naturally the number of smokers increase. So we can identify how much of this change is due to the simple change in the growth in the population and how much of this change is due to the change in smoking prevalence. And from this formula, we can further decompose how much of this change in the prevalence is due to the change in the trend. So following that regression like I was showing you earlier, the declining trend, so we can capture that and we can as well explain how much of this change is due to the change in the characteristics. Like I showed you earlier, the expected smoking prevalence based on the characteristic in the post-crisis period and the difference between this expected smoking prevalence between this characteristic in the post-crisis period and the pre-crisis period. So we can capture how much of this change is due to the change in the distribution of characteristics alone. Whatever is left unexplained is an error. So we can also as well identify that what we cannot explain from the change in the trend and from the change in the practice, it might be the effect of the financial crisis. But one has to consider also that the effect, the fact that the crisis itself changed the composition of population by their employment status. The fact that the crisis increases the number of people who are unemployed and decreases the number of people who are in employment. So in a way, the impact of the crisis is the fact that they change population composition. So we can say that the impact of the crisis is also the change in population composition and different group of population have that different smoking behavior. So that is also the impact of the crisis which come from the change in population in each employment status and the error term. So here's the result. This is our calculation of the number of smokers in the post-crisis period. So in the pre-crisis period, there's on the number of US population and our own calculation of smoking prevalence. In the pre-crisis period, 59 million of US people are smokers and this number reduced to 53.9 million in the post-crisis period. So that's from the total population. And of course if we decompose it, so the US population increase, so we can say that within this change, part of it comes from the increase in the US population, part of it is due to the change in the prevalence and because as I show you that there is a declining trend, so most of this change actually, the declining in the total number of population is due to the change, is due to the declining trend in smoking prevalence and in general. If we look at the people who are unemployed, the number of people who are smokers amongst the unemployed have declined substantially from 37.9 million to 29.6 million. And part of it could be explained by the fact that the number of employed people declined in the post-crisis period. So that automatically contribute to the declining in the number of smokers among the employed people and part of it is due to the declining trend and part of it is due to the change in the distribution of practices of people who are employed. So basically amongst the unemployed, due to the crisis, the number of smokers reduced by 2.2 million. For the unemployed people, so there were five million smokers among the unemployed in the pre-crisis period and that number increased to 8.6 million in the post-crisis period. And simply we can explain this increase of course as the declining trend, the change in the characteristic, but most of this change in the number of smokers among the unemployed is due to increase in the number of people who are unemployed in the post-crisis period. So now come to the conclusion. So for the smoking trend, we have found no significant change in the linear slope of the smoking trend in the U.S. over the period of serve. So there has been no acidulation in the pre-crisis and the post-crisis period. It's sort of the continuous straight linear line. And for smoking prevalence, we can largely explain the reduction in smoking prevalence by the change in the smoking trend and changes on the distribution of socio-economic characteristics related to smoking behavior. And for people who are employed, the smoking prevalence actually declined more than we expected. So that resulted in two million smokers less among employed people in the U.S. For the unemployed, normally the unemployed people smoke much, much more than people who are employed in the pre-crisis period as much as 35% of people who are unemployed are current smokers. In the post-crisis period, of course, as I showed you, that the smoking prevalence declined from 35% to 33% due to the declining trend in smoking prevalence in general. But because the crisis doubled, the number of people who are in unemployment, so what we observed is the crisis is responsible for the increase of more than four million smokers in the U.S. I haven't talked much about the people who are economically inactive, so that could be people who are not, that could be housewives, people who are not looking for a job. That smoking prevalence also declines, but not as far as people who are in employment. So overall, with all this calculation and different employment status, together we found that the economic crisis is responsible for an overall increase of about 20 million smokers. Thank you. Thank you. Questions? Why I have them, you know? Okay. I was wondering if you collected data on race and ethnicity, but you were to break that down at all and were there like certain ethnic groups that were more likely to smoke because culturally acceptable versus other groups? Yeah, so we group people, maybe it's not as fine as we should do, so it's a self-reported ethnicity, so we have white, American, black, Hispanic, and then we group the rest to others, and they obviously have different patterns, especially white and American smoke more than ethnic minorities. Do you say ethnic minorities then? Yeah? Because I have to always be careful with the term. Yeah, it's ethnic minorities. The data on employment, however, suggests a huge disparity along ethnic lines. So for example, unemployment affected, I think African-Americans much more than other ethnic categories. So if that's the case in this population that you sampled, and I don't know the trend with just general smoking across the ethnic categories, but it may, I don't know how it will affect your data, but accounting for the fact that the transition to unemployment was not evenly distributed across the ethnic groups. So we might, yeah, that is a good idea actually, and then we break down into different ethnic groups, and given they have different smoking periods and different streets. Before and after. So that could be, we call this an implication of which groups should not be, and I would guess maybe women, or so it can be more than women. Yes. Well, so forgive me if you mentioned this, and I missed it, but so there's the question of, unemployed versus not employed, right? So if you're unemployed, you have to be in the labor force, meaning, or the dominator of the unemployment rate is people in the labor force, so people who are employed and people who are looking. And so discouraged workers are people, I believe that's the term, who are not looking for work, but they also do not have a job, and they may in the future drop back into the labor force by starting to look for a job again. So do you have data on, and when you say unemployed, you mean? They're looking for a job. Yeah, so, I mean, do you have a sense of how this would be affected if you looked at? Yeah, actually. Workers versus? I haven't showed you the total, total, so you have to consider also, I would say people who say that they are making an active, so they're not looking for jobs, that would be the discouragement. Plus, we have to calculate it for the retirees and students, but I won't show you, because basically a conversation of people who are in retirement and students, they don't change, but they're looking. Are you referring to the adversary this morning? No, I was just wondering if that affected your results, because from a public health point of view, like you really wouldn't care whether someone's a discouraged worker, and therefore not in the labor force, or if they were unemployed and looking. I mean, you don't want, if your goal is to understand better, to reduce smoking, I mean, if you see what I mean. So I just was wondering if your results were robust too, like the, taking them, because there's all sorts of people churning in and out of the labor force, right? Who? I think for this current period, so we really have people involuntarily lost their job, right? Probably 2019, 2010. So that, I would assume that in that period, they were looking for a job, but the article that I was mentioning that I just read this morning is, now the retirees, although the people in retirement age in the US, so they voluntarily left their job, so that would sort of increase the space for people who are probably these college workers and coming in, but also the increasing number of these college workers, one has to take into account different combination, I guess, especially at this 2011-2012. So I guess I am safe for the 2019, the same time. But, yeah. My other question has to do with sort of selection effects of smokers into unemployment for them. I guess what I'm mostly thinking of here is is that employers may sort of subconsciously or at least not overtly, be more likely to let go smokers because they're worried about health insurance costs going up. And some jobs expressly say, I think that the smoking is not allowed. So I know like for a lot of police officers, they're not allowed to smoke. And so, so it's sort of discrimination against smokers and maybe legal discrimination, but discrimination against smokers, could increase the number of smokers among the unemployed if they're the first to be let go because the employer has explicit or implicit concerns about having smokers on the job. So you sort of explain it, that's why we find. Well, it could be. But your explanation fits with the finding. Yes, it does, yeah. We actually found that we employ people as smokers and it has been defined in the modern. And that fits because you stay in job and then you smoke and then you may even try to stop smoking if you haven't started. Because you'll be at it, you might be in job. That's perfect, that's good. We haven't found a good explanation for that, thank you. So I have a similar question to Bailey about this. During this period, governments are also changing policies in terms of the size of cigarettes changing. And so, one thing just would be nice to see, you could just sort of estimate the results by states that haven't changed at anything versus those states that have increased the price of cigarettes to see if there are many differences in your results. But then one thing that might help you, so I was thinking, to me at least, I don't know if this is true, but my prior would be, it would be the states who are in most dire straits that would probably be increasing the price of cigarettes during these periods. And those would also be the states that have the highest share of unemployed people. And because you're actually finding that the unemployed are, the prevalence of smoking is actually increasing, that could help your story against the story that I'm telling. So, but anyway, it was just something to think about. Most people would think that the simultaneous change of price of smoking would affect these people. But you're right as well, and I think particularly when we look at the US, and the US is a big, it's like Europe as a continent in a way, and you always have considered differences by state and by city. Like you said, maybe one thing we could do is to split it up using the same type of methodology but split them up. Just to see how they change, the estimates change. So, there has been many high-profile movies about the tobacco industry and how sneaky they can be. And I find it, I have to believe they don't know this information, they use this as a way to increase the number of smokers, but it would be very difficult to accomplish. But are there, they don't publish that data, of course. I mean, if you work for the tobacco industry, you might know this without publishing it. But this is not the fourth crisis we've had, right? So, I'm wondering how robust this is over the different, because I think in 1992, for example, we had a similar economic crisis at least in California. And going back in history, you showed some of the references, but was it really that predictable that with economic crisis, high-risk behaviors go off? Yeah, in fact, the fact is that I was showing the focus of that particular period, and they found that it goes down, actually, with the risk in here, but most of the evidence of that size and the explanation, most of it, they come from economic literature, and they kind of explain that when you lose your job, you have more free time, so you sort of be drinkless, because normally, even if you're in employment and you're stressed because of jobs, you use more and more smokers, but that is kind of ridiculous, but kind of robust, really, for me. So, in fact, the idea behind this research is sort of trying to figure out, but not replicating the period that you mentioned, but looking at this period, the idea was that the job was higher, because in fact, in 1990, it was high unemployment, but still, the job, but the job was in that time. Yeah, yeah. Are you looking at all other tobacco products? Because I know there has been a jump in some, and smokeless, for instance. Also, there's been some change in rolling your own versus branded cigarettes, and even when those taxes go up, moving to rolling pipe tobacco, I've just seen some recent. But overall, I think the smokeless would be where you would see, there might be some change in this period that might indicate that overall, tobacco use is up versus just smoking. Right, of course, yeah. This one refers to anything you smoke, but you were right in maybe, there's some variation within what kind of things. Well, I'm thinking, yeah, if people who are unemployed maybe are still using tobacco, even if they're not smoking, so maybe the overall tobacco use is even higher because they're starting to use smokeless as an alternative, or maybe the employed are, so I don't know how that would fit into the whole picture, but certainly there's been a lot more marketing of smokeless over this period. Yeah, I think we know what kind of thing this works, that could be something we might not know. But that wouldn't be smoking necessarily, it would be smokeless, so. Yeah, but I think that ties into what Andrew was saying. If smoking is not allowed, yes, then there may be other ways people choose to indulge without an obvious. Yes. So it might be interesting to look at that. I have to say that smoking is a public health issue all over the world, but the U.S. is probably the best in anti-smoking campaign in the world. In Austria, 35% of people in general smoke, and everybody go, in restaurants you can smoke, and that is quite horrible, because in Italy it's doing good actually, it's more or less the same level as in the U.S. I think California, when you refer to U.S., I think California is way ahead of some other states, and we've seen them decline in cancer rates, from lung cancer in California. Okay, so this is not representative. Probably not. Yeah. Oh, thank you very much. Thank you. Thank you.