Monday, November 19, 2012

Don’t Let Death be a Monkey Wrench


Levitt and Dubner highlighted a really interesting aspect of medical treatment, the fact that prescribing anything, whether it has been proven to be effective or not, acts as a powerful incentive. In their discussion of ontology treatment, L & D made it clear that the successes of chemotherapy are marginal at best – in one large study, only 63% of cancer patients survived, but chemotherapy contributed less than 2% of this success (I’m curious as to what factors were found to account for the much more impressive 61%). Despite this weak evidence in the efficacy of cancer treatment, such practices consumer a huge portion of the health-bill. Why? L & D respond that incentives such as high salaries of oncologists, and the average gain in two-months to live motivate doctors to prescribe these expensive treatments. However, they also point out a third reason, which I find to be the most compelling. They write that doctors find it difficult to tell patients that there’s nothing they can do – that modern medicine has no answer for them and they must face death without any tools with which to fight. I highlight this reason not only to defend the morality of doctors, but because I believe it reveals an important factor coming from the demand side of the health market. If cancer treatment is as ineffective as L & D suggest, even if doctor’s prescribe it, why would anyone do it? Consumers, (in this case patients), are not stupid; when they are diagnosed with a disease, they have the capability of looking up survival rates, treatment success, etc. Thus, it seems like cancer victims should opt-out of chemotherapy treatments much more frequently. While 30 % already do this, 70% do not. Their reason for suffering through the painful side-effects associated with such treatments must be hope – while chemo will probably fail in lengthening their life span, they understandably cling to the small hope that they will be the anomaly, and doctors are willing to oblige.
            This trend to delaying the reality of death is also illustrated in ER visits in last weeks of life. A study conducted by UCSF and Harvard colleagues found that over 50% of older adults who had died had visited the ER during their last month of life (this number rises to 75 % when extending the time constraint to 6 months). Furthermore, more than 75% of these people were admitted, and 68 % died there. These statistics highlight that people frequently use hospitals for comfort, which is a wildly ineffective allocation of resources. ER’s exist to diagnose and treat, whereas hospices are much more suitable to the type of care these elderly patients seek. These ill-advises visits may help explain why a quarter or more of Medicaid spending occurs in the last six months of life. In sum, similar to the chemotherapy battle, we as a society are guilty at throwing money at unfixable conditions in an effort to avoid accepting death – this fear acts a monkey wrench in the machine, preventing it from making optimal allocation decisions. This trend is extremely difficult (for those of you who disagree, I would say you probably haven’t faced this type of decision), but still clearly merits change from both the demand (the patient-side), and the hospital side, who are all too willing to appease fears and admit incurable patients.
Source:
http://newoldage.blogs.nytimes.com/2012/06/05/at-the-end-a-rush-to-the-e-r/

Lack-o-riginality

Super Freakonomics was clearly designed for the populace. While the bloggers below me raise valid points about Levitt's assumptions, transparency of methodology, and thoroughness (which I share), what I also noticed was how little of the book was original research.

Translation: why can't I google "examples of confounding variables" and publish the next installment in the "freako" series?

Chapter 1: Prostitution data is from Columbia's Sudhir Venkatesh.

Chapter 2: Data on life expectancy by month is from Douglas Almond and Bhashkar Mazumder, who also proposed the Ramadan solution;  data on athletes' birthdates comes from Florida State Professor Anders Ericsson; terrorist education data borrowed from Alan Krueger; the ER data was compiled by Craig Feied.

Chapter 3: Altruism research was conducted by economist Gary Becker; results from the Ultimatum simulations were conducted by "a group of preeminent scholars" and published in Foundations of Human Sociality.

Chapter 4: The proposition of the seat belt came in the 1950s from Robert McNamara; modern crash data comes from the Fatality Analysis Reporting System (FARS) (Levitt only summarizes the newer data to update McNamara's previous finding).

Chapter 5: The data linking locally grown food consumption and greenhouse-gases comes from Carnegie Mellon Professors Weber and Matthew; the idea that carbon dioxide is not a greenhouse gas comes from Nathan Myhrvold, the former CTO of Microsoft.

I could go on for each chapter, but the point is clear.

It seems a lot of the original contribution from Levitt is just insufficiently justified assumptions, as mentioned by the bloggers before me. Levitt's value-add (if you can call it that) seems to be statements like "let's say it takes an average of one minute to remove and replace your shoes...so...the tax is the time equivalent of 14 lives per year." (93)

I know for a fact Subin, Jen, Sarah and I can walk you through those sort of assumption-to-conclusion progressions -- we practiced consulting case interviews for hours on end over the summer.

It just seemed odd to me that such a popular book would only require skills most undergraduates already have. Paper A says X, but Paper B says X is wrong because of Y. Too large a portion of Super Freakonomics followed this format, a point with which the New Republic concurs.

To be fair, Levitt brilliantly synthesizes and presents the data cited in his book - and he certainly makes appropriate citations. However, I am a little peeved that Levitt gets another best-seller and furthers his reputation for cutting-edge economics while largely piggybacking off the research of others.

And yes, by 'peeved' I do mean 'jealous'.


In addition, I was very intrigued by the argument about birthdays and professional athletes. It seems that Levitt's thesis still holds up: 2005 data provided by Major League baseball revealed that October birthdays were far and away the most common. Sure enough, the Little League birthday cutoff is still July 31. A study on NHL hockey players revealed a similar trend for January births, since most international cutoffs are December 31.

The performance of these athletes once they reach the professional level, interestingly, has very little correlation with birth month.

Freako-not-eco


I admit it. I find the Freakonomics book series to be charming, and, damn, if it isn’t entertaining as hell. The economist-meets-journalist pairing of Levitt and Dubner makes for a witty, easy read. 

However, I have encountered skepticism about the accuracy of the claims in their works.  I first read a criticism of the Freakonomics franchise on the website Arts and Letters Daily.  ALDaily aggregates the “best” in long form journalism.  (I highly recommend it as a homepage.)

Leave it to a bunch of scholars and long-form nerds to throw down a one-liner like this:


"The Freakonomics formula. Anecdote-rich, contrarian narrative + speculative claims presented as fact = publishing phenomenon..."

There seems to be a bit of truth in this statement.  The actual article critiquing Freakonomics and Superfreakonomics, which appeared in the American Scientist magazine this year, is a little less tongue-in-cheek.   I revisited the piece this afternoon, and was happy to find it co-authored by Kaiser Fung and Columbia Professor Andrew Gelman. (I used Gelman quite a bit in my thesis.) Gelman and Fung, both statistics professors, acknowledge that it can be difficult to be engaging while teaching statistics, but note that Levitt and Dubner have a habit of making unfounded statements.

Gelman and Fung goes on to critique several of Levitt and Dubner's claims.  With regards to Superfreakonomics, Gelman and Fung specifically hit on three areas of the book: the opening anecdote about walking versus driving drunk, the section on predicting terrorists and the final chapter on climate change.  You can read their complete case by case critique here.

In the beginning of the article, Gelman notes:
“Levitt is celebrated for usin data and statistics to solve an array of problems not
typically associated with economics” (Gelman).


I find this quote to be particularly interesting because it makes me wonder -- why do we call it "Freakonomics"?  Is this truly a book about economics?  I do not mean to undermine Levitt's work as an economist, because he is incredibly well-credentialed.  But at its roots, isn't this more of a book on statistics than it is on economics?  What about psychology?  Many of the subjects discussed here were not touched upon in any of the other books we read for this class.  I'm certainly not trying to call Professor Blomberg's choices into question, I just find this trend quite interesting.  I was assigned Freakonomics for my Econ 50 class. 

Gelman and Fung concede that the original book Freakonomics was much more based around "Levitt's own peer-reviewed research."  I would definitely agree -- although I appreciate their linkage of street prostitutes and department store Santas.

It’s a fitting that this fall as the last book on our schedule.  At the beginning of this course, Heilbroner warned us about these “airport economics” books, which seek to please the reader, rather than to provide a thoughtful discussion of economic research.  Unfortunately, I think Levitt and Dubner's work falls into that category.  It’s a GOTCHA! kind of economics.  Sometimes their claims seem a bit outlandish and ounfounded.  If it sounds like the script for a Hollywood movie…well, it is.  Check out the trailer:


 

Privacy Invasion or Practicality - Behavior Prediction and Advertising


I enjoyed reading Superfreakonomics. Particularly, I found interesting the discussion of “Ian Horsley”’s model for using bank and demographic information to find those customers most likely to be terrorists. The success of this sort of endeavor has many interesting philosophical connotations – if our behavior is so reliable that government agencies or companies with access to sufficient data can readily predict what we will do, free will faces a significant challenge. However, another interesting implication of this form of behavior prediction is in its potential use for ad agencies and companies that want to target consumers.

As the New York Times reported earlier this year, Target had begun to analyze customers’ buying habits with the hope of predicting when female customers were pregnant. The article explains that “new parents are a retailer’s holy grail” because new parents are one of a limited set of groups that are at a point in their lives when they’re willing to change their shopping routines. Thus, new parents would be more easily convinced that Target was the only shop they needed. As Target quickly found, the method of finding new parents was very effective – almost too effective. The company sent ads forbaby supplies to a teenager, provoking an irate response from her father. A few days later, the father called back to apologize and tell the store that they had it right – they knew his daughter was pregnant before he did.

Of course, most customers understandably found that sort of aggressive advertising creepy and intrusive. This “creepy factor” seemed to discourage them from considering Target for their shopping needs somewhat. In response, Target scaled back their advertising efforts and instead sent more subtle advertisements to the pregnant women – interspersing baby items with regular items, for example, rather than sending them a book of coupons for baby formula only. Interestingly, this case seems to suggest that some of the fears about targeted advertising are self-regulating – if people feel like an advertiser is extending into their life too much or being too forward, they will respond negatively, which will force the advertiser to change their approach. (That said, in this case, this incentive only changes how the advertiser or company appears to the customer, and not how much information the company gathers. Conceivably, though, in other circumstances the incentive could be structured such that less information would be necessary for the more subtle approach, and thus less information would be gathered.)

It’s interesting to see this sort of method play out (perhaps less successfully) in other circumstances. Facebook ads are, theoretically, a great opportunity for targeted advertising – the host company already holds a wealth of information about the audience for ads. Yet Facebook does not seem to have taken full advantage of its advertising potential, or at least does not use its data nearly as well as Target. Too often it posts random but specific or worse, contradicting ads – on many occasions I have had an ad for a Jewish dating site, an atheist organization, and literal Bible belts – with Bible quotes on them. (My religion was set to Pastafarianism at this point, for the record.) Yet Facebook manages to demonstrate some of the potential benefits of targeted advertising for customers as well as companies – when the ads actually match the customers’ interests, like about tickets for his or her favorite band on sale, it provides potentially valuable information that overall increases utility for the customer. Understandably, our gut reaction to companies gathering private information about us (though obviously if we post it on the internet it’s not quite so private) might be concern about our privacy, risks of identity theft, and so on. Yet it seems that, given enough safeguards to protect information and a sufficiently subtle approach, targeted advertising could serve as a very useful tool for facilitating business between companies and customers that might not otherwise meet.

Of course, I’ll still find it a bit creepy every time Facebook perfectly anticipates my Domino’s craving with a well placed ad. But if I get $5 off and a yummy pizza, who cares?

Sunday, November 18, 2012

A Closer Look at Cable Television

I found SuperFreakonomics to be super interesting and entertaining. I was, however, frustrated at the lack of detail and methodology in the book. I know that this book is aimed for pop audiences and so this it to be expected, but I wanted to look every article up as I went. Since there were dozens of articles used as the basis of this book, I just chose one: The Power of Cable TV: Cable Television and Women's Status in India. Specifically, Levitt and Dobner's discussion of this issue stood out to me because there seem to be so many other correlates of access to cable television that and discrepancies in villages over time that I was not sure that this was feasible to prove. 

There were several things, before checking the article, that I wanted to make sure were controlled for: (1) income of the village, (2) closeness to urban center, (3) prevalence of other types of media (satellite television, internet usage).

Income was somewhat controlled for. In one of the two states used in the article, income was available and tested. It was found that income correlated with getting access to cable. Thus, it could be that as rural Indian families get more money, their attitudes towards women change AND they get cable access. 

Furthermore, access to cable was also associated with the timing of receiving access to cable. Since villages closer to towns were more likely to get cable, they are more likely to be modern anyways. Worldwide this seems to be the trend; there is a great dichotomy on social issues between rural and urban areas in the US.

This article does not control for access to other types of media. It could be possible that one town never gets access to cable because a competitor, satellite for example, got there first. This is not mentioned in the paper.

Furthermore, the results on women's status and cable were not as consistent as they were portrayed to be in the book. In fact, those with cable access thought  it was acceptable for a husband to beat his wife in more situations than those who never had cable. While getting cable access decreased the acceptance of abuse, there was a significantly larger differential between before and after cable and those who never got cable and those who did. This signals that there may be significant, unmeasured differences between the villages in question.

For the data on pregnancy, the results were even more volatile. The discrepancy between years for the non-cable villages was much greater than the downward trend in the villages with access. This signals that this paper may have been damaged by an insignificant sample size - approximately 180 villages.



In summary, the book was entertaining, but each point would have to be examined individually for this to be credible. For one, cable access is not a panacea for gender disparities.

Superfreakonomics- Heating up the global warming debate


Superfreakonomics- Heating up the global warming debate

Superfreakonomics was a very interesting read. Building on the Freakonomics formula of “this is why the conventional/intuitive view on _____ topic is wrong and here is the data to back up what we view as the correct understanding of _____ phenomenon,” Superfreakonomics continues to look at data in very interesting ways. The most interesting part of these books for me was the ability to use clever naturalistic experiments to examine interesting causal links.

When Superfreakonomics came out it caused a bit of ruckus. When rereading the book, I focused a bit more heavily on the sections that I had read some criticisms about. For instance, Ezra Klein challenges the drunk driving anecdote: arguing that one of the key assumptions in Levitt/Dubner’s calculation “If we assume that 1 out of every 140 of those miles are walked drunk -- the same proportion of miles that are driven drunk” is completely unjustifiable. Ezra goes on to point a number of reasons this statistic might be skewed (more people substitute away from driving drunk towards walking drunk, the sort of miles traveled (rural/urban), and a host of others http://voices.washingtonpost.com/ezra-klein/2009/10/the_shoddy_statistics_of_super.html). While perhaps a justifiable claim, I found this line of criticism a useful reminder not to necessarily be swept up in a statistical story just because it was interesting, but to remember to look closely at both implicit and explicit assumptions in the model. Superfreakonomics is very good at telling interesting stories with data, but it is easy to get swept up and forget to challenge what might be implausible assumptions.

Unsurprisingly, a possibly skewed or misleading assumption in drunk driving was not what caught the media’s attentions. Rather, the passage on global warming raised quite the stir amongst environmental economists and a few global warming advocates. The case Dubner and Levitt seem to be making in the broad sense is that is difficult, if not impossible for a large spectrum of people to change their behaviors with little or no incentives to do so. This is consistent with their broad message, and does a good job of illustrating why the current incentive structure regarding the environment seems suboptimal. Where the critics seem to have taken issue with the work comes through in a few important ways. First, and perhaps most importantly, seems to be a general challenge to the way the facts are presented. While many of them are factually correct, taken together these facts present a misleading picture. For instance, the following sentence received a great deal of criticism: “When Al Gore urges the citizenry to sacrifice... the agnostics grumble that human activity accounts for just 2 percent of global carbon-dioxide emissions, with the remainder generated by natural processes like plant decay...” DeLong and other critics point out that this fact is misleading for a number of reasons. First it underplays the importance of human impact on global warming (http://delong.typepad.com/sdj/2009/10/yet-more-superfreakonomics-blogging-yes-i-know-i-know.html). How the critics would like that sentence to read places more accurate emphasis on human impact: “Of course the agnostics are misleadiing you: the right way to think about it is that already 1/3 of the CO2 molecules in the atmosphere are the products of human actiivity, and the fraction and amount are growing very rapidly indeed…”
Other criticisms focus on misrepresenting scientific studies. Not being an environmental scientist, it is rather difficult to critically say whether these studies were presented fairly. Some certainly seem to claim otherwise: http://thinkprogress.org/climate/2009/10/12/204787/superfreakonomics-errors-levitt-caldeira-myhrvold/ and http://krugman.blogs.nytimes.com/2009/10/17/superfreakonomics-on-climate-part-1/. While, from what I have read many of the general criticisms are a bit overblown and really fall into the category above of being factually correct but at times a little misleading: http://www.freakonomics.com/2009/10/18/global-warming-in-superfreakonomics-the-anatomy-of-a-smear/.

Another area of criticism is drawn around first the criticism of solar panels and the inaccurately rosy picture of geo-engineering. First the solar panels. Superfreakonomics notes that the color of the solar panel makes a difference. For the same reason painting all the roofs white helps to lower the temperature, placing black solar panels all over the place would have the opposite impact. Most of the critics seem to note this as a rather minor impact, and argue that solar power won’t solve the energy crisis for a host of other reasons (http://delong.typepad.com/sdj/2009/10/hoisted-from-comments-nicholas-weaver-on-solar-vs-nuclear-myhrvold-dubner-and-levitt.html). One commentator compared saying the color of solar panels causes global warming in the same way photons from stadium lights cause a curve ball. While it seems solar power is slightly better than the book would have you believe, geo-engineering seems to be a bit worse. There seem to be a host of potential problems with geo-engineering, the least of which being its untested, unknown nature.

The economist criticizes the work by saying: “it contains little in the way of economics, other than a brief discussion of externalities and the observation that the hose system would be much cheaper than building an entirely new low-carbon energy infrastructure for the world.” This quote does a pretty good job of summarizing by feelings on this particular chapter of the book. The economics is generally pretty good; I agree with their thoughts on incentives and the difficulties surrounding policy solutions. I think their focus on one particular plan rather than a market structure aimed at fixing some of the underlying incentives is a bit weak.

For a quick summary of the debate and some of the sources please see:
http://economistsview.typepad.com/economistsview/2009/10/superfreakonomics-on-climate.html

To be or not to be....a Prostitute or a Wife?

In the first chapter of Superfreakonomics, Levitt and Dubner provide a cost-benefit analysis of the choice to be a street-level or a high class prostitute in Chicago. What really interested me about this chapter is the last few pages – about Levitt and Dubner’s description of Allie, a ‘high-class’ prostitute in Chicago who spent her nights being wined, dined, and pampered by wealthy businessmen. Allie makes an important observation about herself in her interview with Levitt and Dubner -  she could be the younger, more sexually adventurous “ideal wife: beautiful, attentive, smart, laughing at your jokes and satisfying your lust”(53).  This, coupled with Levitt and Dubner’s point that prostitutes have a harder time finding a husband than a non-prostitute, made me wonder whether or not is was economically better for a woman to be a prostitute or a wife, as well as why men choose to get married rather than continually engaging with prostitutes (or doing both). In order to discuss this question, we first have to assume that selecting a mate itself is a market, and that marriages only occur if it is profitable for both parties involved.

Because marriage can be an important (and sometimes only) source of income for women, prostitutes themselves face an opportunity cost when deciding to pursue this career. Prostitutes are paid so much higher than women who work in other fields, or women who are wives, because they need to be compensated them for the marriage-market earnings they gave up when they decided to be a prostitute.
Because prostitutes are undesirable as wives, buyers must pay a sort of ‘no-husband’ tax that a man normally pays if he is married. Moreover, the opportunity cost for becoming a wife in comparison to a prostitute is all the lower because wives can easily divorce and become a prostitute if married life is not suiting them.  However, while they are more limited in the above respect, prostitutes only provide (and are paid for) one specific service by men. Wives, on the other had, deliver a variety of services – they cook, clean, and provide sex. A study conducted by Lena Edlund and Evelyn Korn compares the two different careers and determines the cost-benefit analysis of each. Assuming that both wives and prostitutes are sellers of the same product (and thus are interchangeable), this study determines that wives can offer more than a prostitute at a lower cost. While prostitutes can only offer non-reproductive sex, wives can offer their husbands both non-productive and reproductive sex (ie children), and they offer it at a lower cost (ie, husbands do not have to pay for sex, and, more often than not, wives provide other services like cooking and cleaning, that prostitutes do not). However, although wives are a low-cost alternative to prostitutes (ie you get more bang for your buck – yes, pun intended), there is an opportunity cost to becoming/engaging with a wife – it is a much longer contract than if becoming/engaging with a prostitute. Men and women who engage in marriage have much different and greater responsibilities to each other than a man and a prostitute would. Also, because marriage is a longer contract, we must also take into account that wives age, and because men cannot buy services from multiple wives like they can prostitutes, aging does impose an additional externality on the buyer. This makes wives not have the same guarantee for your money as with a prostitute.  This lack of guarantee could be the reason for why married men decide to engage with prostitutes as well as their wives. Moreover, some economists consider wives to be superior to prostitutes because the consumption of the wife increases as income rises – like fine wine. This basically means that the wealthier you are, the more likely you are to consume champagne (wife) than beer (prostitute). Essentially, from a male perspective, the greater your income, the more benefit you will receive from choosing to marry a woman instead of engaging with a prostitute every evening.

There are other ways for women to get income besides simply being a prostitute or a wife. However, as Levitt and Dubner show through this first chapter, the prostitutes of Chicago (whether high-class or street-level), often make more money than they would working in a job in another field. Moreover, based on the anecdotal evidence provided through Allie’s story of being a high-class prostitute, women who are prostitutes have much more flexible schedules, and work about half as many hours as women making a comparable wage, making it seem like being a prostitute is a more economically desirable career than working in a different field.