The financial/credit crisis and our recent economic woes looms large in everyone's mind these days. Even in my counter-cyclical student bubble haven, there are daily reminders (both on the news and in the flesh) of people who have been hurt by the recent economic downturn (much more trivially, my own paltry investments have taken a huge hit these last 10 days, to the point where I've stopped logging into E-trade to view the carnage). All of this has deepened my resolve to try and understand how this crisis originated, how it can be prevented in the future, and what our country's economic prospects are for the next year or two.
My usual source for this kind of knowledge is The Economist. Lately though, I've found the stuff on the financial crisis to be a bit incomprehensible (is it just me or are they being a bit too jargony?). Thus, after a steady dose of the Englishmen, I felt up-to-date on all the news (like which bank went down most recently), but the hows and whys were still unclear.
Today, I did some browsing and found three pieces that really helped me gain a conceptual understanding about the financial system and what went wrong. The links are provided below. The first two couch the crisis in the form of hokey stories. It might seem a bit Mickey Mouse, but the insights are powerful and the analogies accessible. I would classify these as must reads for anyone who wants to understand what is going on without having to decipher all the usual Wall Street gibberish. The third is a guest post on the Freakonomics blog: also very good, and goes through the key questions and players and also provides some insights into what might happen in the future.
Finally, if any of you have found other sources/websites that explain the crisis in easy-to-understand terms, please post a link in the comments. Thanks in advance.
Links:
1. Interfluidity blog: Credit Crisis for Kindergarteners
2. CNN.com: Glenn Beck
3. Freakonomics: Diamond and Kashyap on the Recent Financial Upheavals
Thursday, September 18, 2008
Saturday, September 13, 2008
The Dubious Legacy of Candy Cigarettes?
I remember getting my first candy cigarette from a classmate at school when I was eight years old. I had no idea what it was, but it seemed pretty exciting and, despite my unusually high discount rate when it comes to candy (ask anyone in my family), I ended up saving it for when I got home.
This turned out to be a huge mistake. My mom saw me "smoking" and completely flipped out. From then on candy cigarettes were banned from my house (and, to this day, my mom gets mad if my sister and I pretend carrot sticks or pens to be cigarettes or cigars). Later, enough mothers were upset by the whole idea that candy cigarettes were banned from my elementary school, as well.
What worried moms across the country was the possibility that smoking candy cigarettes would induce children to take up the real thing a few years later. Is there good evidence that candy cigarette use has a causal effect on later smoking? I trolled around the internet and found some interesting links off the candy cigarettes Wikipedia page. As it turns out, there IS a study on the association between candy cigarette and nicotinic cigarette use (see here for a summary and here for the actual piece). Here are the findings (lifted from the abstract):
26.4% of respondents reported current smoking and 29.4% reported former smoking. Candy cigarette use was reported by 88% of both current and former smokers and 78% of never smokers (p ≤ 0.001). Logistic regression showed that the odds of smoking for those who used candy cigarettes was 1.98 (95% CI: 1.77, 2.21) for ever (current plus former) smokers and 1.83 (1.59, 2.10) for current smokers, compared to those who had not used candy cigarettes. Odds for current and ever smoking increased with increasing candy cigarette use.
The main issue here, of course, is whether the link between candy cigarettes and smoking is causal. For example, if individuals who really want to fit in use candy cigarettes to appear "cool" and later smoke for the same sense of social acceptance, one would see an association between the two behaviors, but this association certainly would not be causal. Even so, the results are pretty intriguing and worthy of further exploration.
This turned out to be a huge mistake. My mom saw me "smoking" and completely flipped out. From then on candy cigarettes were banned from my house (and, to this day, my mom gets mad if my sister and I pretend carrot sticks or pens to be cigarettes or cigars). Later, enough mothers were upset by the whole idea that candy cigarettes were banned from my elementary school, as well.
What worried moms across the country was the possibility that smoking candy cigarettes would induce children to take up the real thing a few years later. Is there good evidence that candy cigarette use has a causal effect on later smoking? I trolled around the internet and found some interesting links off the candy cigarettes Wikipedia page. As it turns out, there IS a study on the association between candy cigarette and nicotinic cigarette use (see here for a summary and here for the actual piece). Here are the findings (lifted from the abstract):
26.4% of respondents reported current smoking and 29.4% reported former smoking. Candy cigarette use was reported by 88% of both current and former smokers and 78% of never smokers (p ≤ 0.001). Logistic regression showed that the odds of smoking for those who used candy cigarettes was 1.98 (95% CI: 1.77, 2.21) for ever (current plus former) smokers and 1.83 (1.59, 2.10) for current smokers, compared to those who had not used candy cigarettes. Odds for current and ever smoking increased with increasing candy cigarette use.
The main issue here, of course, is whether the link between candy cigarettes and smoking is causal. For example, if individuals who really want to fit in use candy cigarettes to appear "cool" and later smoke for the same sense of social acceptance, one would see an association between the two behaviors, but this association certainly would not be causal. Even so, the results are pretty intriguing and worthy of further exploration.
Wednesday, September 10, 2008
Health Policy Seminar Series at Yale
If you are a Yale student or someone who happens to be visiting on a Monday between noon and 1:30 PM, be sure to check out this semester's Division of Health Policy and Administration's seminar series. The bill looks fantastic and represents a good mix of economics, health services research and management. It also features me, bringing up the tail end with a talk on a subject that is as of yet unknown (you'll hear more about it what I am doing in the next few weeks).
Hope to see you here!
Hope to see you here!
Tuesday, September 9, 2008
Public Nutrition Programs and Child Health
Two interesting working papers this week on the effect of major U.S. nutrition programs and child health. The first, by Douglas Almond and coauthors, examines the effect of the Food Stamp Program (FSP) on birth outcomes. Using variation in the roll-out of the program (Food Stamps became operative in different counties at different times during the 60s and 70s) to identify causal effects, the authors find that women exposed to the policy for three months or more of their pregnancy gave birth to heavier babies.
Daniel Millimet and coauthors study the School Breakfast Program (SBP) and the National School Lunch Program (NLSP), looking at the link between school feeding and childhood obesity. The authors find that the SBP works against the rising tide of childhood obesity while NSLP exacerbates it. Unfortunately, they offer little intuition for these opposing findings, especially when the two programs have similar inclusion criteria (i.e., they target poor students). One possibility might be the scope of the program: the NSLP serves twice as many students as the SBP. Another is the nutritional content of the meal: lunch is higher in calories (though similar in "healthiness").
Besides discussing issues of intrinsic interest to public health, these two papers are also noteworthy in their rigor in pursuing causal effects. The Millimet, et al study, in particular, uses a methodology where, recognizing that kids who enroll in SBP and NSLP are likely different than those who do not, they study the sensitivity of their results to different degrees of selection bias. Basically, the authors define a parameter for non-random selection into the program, and dial the level of this parameter up and down to see how their econometric results change. Very clever stuff. It's nice to see that, lacking experimental or even quasi-experimental program variation, such as that exploited in the Almond paper, it is still possible to say something about causality. The selection correction method utilized by Millimet, et al is simple and intuitive: I hope it finds it way into medical journals soon!
Daniel Millimet and coauthors study the School Breakfast Program (SBP) and the National School Lunch Program (NLSP), looking at the link between school feeding and childhood obesity. The authors find that the SBP works against the rising tide of childhood obesity while NSLP exacerbates it. Unfortunately, they offer little intuition for these opposing findings, especially when the two programs have similar inclusion criteria (i.e., they target poor students). One possibility might be the scope of the program: the NSLP serves twice as many students as the SBP. Another is the nutritional content of the meal: lunch is higher in calories (though similar in "healthiness").
Besides discussing issues of intrinsic interest to public health, these two papers are also noteworthy in their rigor in pursuing causal effects. The Millimet, et al study, in particular, uses a methodology where, recognizing that kids who enroll in SBP and NSLP are likely different than those who do not, they study the sensitivity of their results to different degrees of selection bias. Basically, the authors define a parameter for non-random selection into the program, and dial the level of this parameter up and down to see how their econometric results change. Very clever stuff. It's nice to see that, lacking experimental or even quasi-experimental program variation, such as that exploited in the Almond paper, it is still possible to say something about causality. The selection correction method utilized by Millimet, et al is simple and intuitive: I hope it finds it way into medical journals soon!
Thursday, September 4, 2008
Anyone Interesting in Working on a Small Project?
"Guess who's back, back again..."
That's right. I'm back to blogging after several weeks off - August turned out to be an exceptionally busy month (more on that in the next few days). It's great to come back to this space and resume writing about health economics and other topics of interest.
To start off September, I'd like to use this post to solicit help on a small project I am starting. Basically, I want to look at published studies of industry sponsorship bias in drug trials (a very hot topic: see here, from the recent issue of JAMA) and understand what we can learn with the existing statistical evidence, and need a research assistant to do some data collection and extraction. You will also be able to help out in designing the analysis. Here is what you get in return: authorship if the work gets published, and experience on what should be an interesting, fun, non-trivial and short project.
If you are interested in learning more, write me at this e-mail address.
Finally, I realize I'm being especially vague here about what the project is actually about. This is because I recently had a bad experience getting scooped and do not want to go through that again. However, you can trust me on the fact that this will, indeed, be interesting!
That's right. I'm back to blogging after several weeks off - August turned out to be an exceptionally busy month (more on that in the next few days). It's great to come back to this space and resume writing about health economics and other topics of interest.
To start off September, I'd like to use this post to solicit help on a small project I am starting. Basically, I want to look at published studies of industry sponsorship bias in drug trials (a very hot topic: see here, from the recent issue of JAMA) and understand what we can learn with the existing statistical evidence, and need a research assistant to do some data collection and extraction. You will also be able to help out in designing the analysis. Here is what you get in return: authorship if the work gets published, and experience on what should be an interesting, fun, non-trivial and short project.
If you are interested in learning more, write me at this e-mail address.
Finally, I realize I'm being especially vague here about what the project is actually about. This is because I recently had a bad experience getting scooped and do not want to go through that again. However, you can trust me on the fact that this will, indeed, be interesting!
Thursday, July 31, 2008
Work Hour Limitations for Interns and Residents: Good or Bad for Patients?
A great piece by Sandeep Jauhar in Slate explores the effects of work hour limitations for medical residents on patient care. The article highlights trade-offs inherent in these policies. On the one hand, sleepy residents make more mistakes. On the other, work restrictions that reduce the total number of hours worked among all residents encourage the use of cross-cover mechanisms (the article talks a lot about night float) which may reduce the continuity of care. This, too, may adversely affect patient care if continuity is important. As Jauhar points out, the ultimate impact of work hour restrictions comes down to an empirical question, and the available evidence appears mixed.
The discontinuous introduction of residency work hour restrictions nationwide, as well as the random assignment of interns and residents to regular and night-float shifts, can be exploited to study the impact of both work hour restrictions as well as different cross-cover mechanisms. With such a transparent identification strategy available, I'm surprised that only two major studies have been done on this issue (even more so since the meta-evidence is inconclusive). Perhaps it is difficult to find data on outcomes or on how the time of residents is allocated?
I'd be curious to start looking at some new data on this if anyone is interested.
The discontinuous introduction of residency work hour restrictions nationwide, as well as the random assignment of interns and residents to regular and night-float shifts, can be exploited to study the impact of both work hour restrictions as well as different cross-cover mechanisms. With such a transparent identification strategy available, I'm surprised that only two major studies have been done on this issue (even more so since the meta-evidence is inconclusive). Perhaps it is difficult to find data on outcomes or on how the time of residents is allocated?
I'd be curious to start looking at some new data on this if anyone is interested.
Tuesday, July 29, 2008
Air Pollution and Infant Health
A recent working paper by Janet Currie and collaborators finds the following:
We examine the impact of three "criteria" air pollutants on infant health in New Jersey in the 1990s by combining information about mother's residential location from birth certificates with information from air quality monitors. In addition to large sample size, our work offers three important innovations: First, because we know the exact addresses of mothers, we select those mothers closest to air monitors to ensure a more accurate measure of air quality. Second, since we follow mothers over time, we control for unobserved characteristics of mothers using maternal fixed effects. Third, we examine interactions of air pollution with smoking and other predictors of poor infant health outcomes. We find consistently negative effects of exposure to pollution, especially carbon monoxide, both during and after birth. The effects are considerably larger for smokers than for nonsmokers as well as for older mothers. Since automobiles are the main source of carbon monoxide emissions, our results have important implications for regulation of automobile emissions.
See here for an earlier post on the subject, discussing a paper looking at the relationship between pollution from forest fires and infant mortality.
We examine the impact of three "criteria" air pollutants on infant health in New Jersey in the 1990s by combining information about mother's residential location from birth certificates with information from air quality monitors. In addition to large sample size, our work offers three important innovations: First, because we know the exact addresses of mothers, we select those mothers closest to air monitors to ensure a more accurate measure of air quality. Second, since we follow mothers over time, we control for unobserved characteristics of mothers using maternal fixed effects. Third, we examine interactions of air pollution with smoking and other predictors of poor infant health outcomes. We find consistently negative effects of exposure to pollution, especially carbon monoxide, both during and after birth. The effects are considerably larger for smokers than for nonsmokers as well as for older mothers. Since automobiles are the main source of carbon monoxide emissions, our results have important implications for regulation of automobile emissions.
See here for an earlier post on the subject, discussing a paper looking at the relationship between pollution from forest fires and infant mortality.
Wednesday, July 23, 2008
Do Better Trained Physicians Provide Better Care?
A while back, I blogged about how medical care varies in quality and quantity across different areas in the United States, as well as the reasons why such "small area variation" might exist. This post looks at the same topic but from a slightly different perspective: to what extent can differences in physician quality (where they were trained and where they currently work) explain the variance in medical care seen across groups of patients and regions?
This isn't an easy question to answer: getting at causal effects of physician quality is really hard. For example, an association between patient outcomes and physician quality could simply reflect the fact that wealthier or smarter patients, who are better able to translate directions from their physician to better health outcomes, happen to choose better doctors themselves. As such, conventional estimates using observational data may be biased.
In a very clever study, Joseph Doyle and co-authors get around this selection issue using a natural experiment methodology. In particular, they utilize the randomization of patients in the Veterans Affairs (VA) system to clinical teams from two academic medical centers. One center happens to be very highly rated and the other not so much. Their results suggest that physician quality matters in terms of costs and length of stay, though not so much for ultimate health outcomes:
Those treated by physicians from the higher-ranked institution have 10-25% shorter and less expensive stays than patients assigned to the lower-ranked institution. Health outcomes are not related to the physician team assignment, and the estimates are precise. Procedure differences across the teams are consistent with the ability of physicians in the lower-ranked institution to substitute time and diagnostic tests for the faster judgments of physicians from the top-ranked institution.
Interesting stuff.
This isn't an easy question to answer: getting at causal effects of physician quality is really hard. For example, an association between patient outcomes and physician quality could simply reflect the fact that wealthier or smarter patients, who are better able to translate directions from their physician to better health outcomes, happen to choose better doctors themselves. As such, conventional estimates using observational data may be biased.
In a very clever study, Joseph Doyle and co-authors get around this selection issue using a natural experiment methodology. In particular, they utilize the randomization of patients in the Veterans Affairs (VA) system to clinical teams from two academic medical centers. One center happens to be very highly rated and the other not so much. Their results suggest that physician quality matters in terms of costs and length of stay, though not so much for ultimate health outcomes:
Those treated by physicians from the higher-ranked institution have 10-25% shorter and less expensive stays than patients assigned to the lower-ranked institution. Health outcomes are not related to the physician team assignment, and the estimates are precise. Procedure differences across the teams are consistent with the ability of physicians in the lower-ranked institution to substitute time and diagnostic tests for the faster judgments of physicians from the top-ranked institution.
Interesting stuff.
Tuesday, July 22, 2008
Contest Winners
Hi all. This summer's 'Dar he Blogs winners are Christi H and James Hudspeth, for the professional eating and p-value contests, respectively. I thought Christi's answer really hit it on all cylinders as far as pointing out the physical and mental strength required to succeed in competitive eating. James ran unopposed in his contest. While I didn't quite get the answer I wanted, he did alert me to "The Lady Tasting Tea," which is turning out to be a great book (anyone interested in a breezy account of how statistics has contributed to the practice of science in the 20th century must absolutely read this book!).
Christi and James: Congrats, and get in touch with me about your prize, which will be a popular economics book of your choice.
Christi and James: Congrats, and get in touch with me about your prize, which will be a popular economics book of your choice.
Thursday, July 17, 2008
Contest Reminder!
You have until Saturday (the 19th) morning to submit your entries for the "Competitive Eating" and "P-value" contests. The prizes are excellent and are outlined in the second link (and see here for proof that I actually deliver on the prizes).
Bring it.
Bring it.
Subscribe to:
Posts (Atom)