A Curmudgeon's and Improv's Guide to Outliers: Introduction

So I picked up Malcolm Gladwell's newest book Outliers: The Story of Success the other day, as I'm sure many of you will be doing on your next trip to the airport (where stands of Gladwell's hardcover book, marked down thirty percent, block your every exit through the already cramped airport bookstores.) Gladwell's books are fun, but I find myself often disagreeing with his analysis, so I thought it would be entertaining to take my time reading his latest and jot down my thoughts as I progress. Well "entertaining" in that "holy shit dude you are pedantic" sort of way. Note that I really do like Gladwell's books, and indeed for me, reading with critical eyes is exactly the reason I like his books. Ah, the life of a curmudgeonly pedant, revealed before your eyes, here on these there intertubes!

To balance things out, I've also included some thoughts from the improv part of my brain: the part that takes ideas at more than face value and tries to run with them.

SPOILER ALERT: Dude, I can't talk about the book without giving away what the book is about, so if you don't want the book's main ideas to be spoiled, don't continue reading.

IDIOT ALERT: I'm in no way qualified in most of the fields Gladwell will touch on, so please, a grain of salt, before you start complaining about my ignorance. Yes I'm an idiot, please tell me why!

Outliers: Introduction

Don't have a heart attack: just move and integrate into Roseto???

Gladwell begins Outliers with an introduction describing the work of physician Stewart Wolf and John Bruhn on the town of Roseto, Pennsylvania (for the story as told by these authors see The Power of Clan: The Influence of Human Relationships on Heart Disease.) Roseto is a town which consisted largely of an close knit group of Italian immigrants who, during the 1950s and 60s exhibited a remarkable lack of cardiovascular disease (more specifically heart attacks.) The conclusion of Bruhn and Wolf was, after examining many different factors (hereditary, diet, exercise, and geography) which might cause this effect, that none of the normal factors could explain Roseto's good health, and that the only way to explain the effects in Roseto was that the Rosetians had good health because of the community they lived in. In particular Bruhn and Wolf noted that the community was very close knit and had a very strong family (three generations under one roof) and civil structures. Further studies (The Roseto effect: a 50-year comparison of mortality rates by B Egolf, J Lasker, S Wolf and L Potvin (1992)) evidence that as the community began to lose its strong social structure, the health of the inhabitants decreased. This effect, that community could lead to better health, has been termed by some the Roseto effect. This isn't a view which one hears too often, and indeed Gladwell tells us that Wolf and Bruhn had some work to do: "Wolf and Bruhn had to convince the medical establishment to think about health and heart attacks in an entirely new way: they had to get them to realize that they wouldn't be able to understand why someone was healthy if all they did was think about an individual's personal choices or actions in isolation."

So this, then, is where Gladwell begins his book: describing a town which was an outlier in terms of medical health, and describing how an unusual hypothesis, that community mattered, was largely responsible for this effect. And this, says Gladwell, is what he wants to do, but instead of explaining outliers in medical health, he wants to do this for success.

Curmudgeonly pedant's view: First of all, like most stories Gladwell chooses to focus on, this one is compelling upon first reading. And I'm inclined to give great artistic license to the introductory chapter (a philosonomicon, if you will) where a tone and main thesis are to be hinted at, but not defended in great detail. But that's no fun! Let the pedanting begin!

One issue that most bothers me when reading Gladwell is exactly the thing that makes him most enjoyable to read: argument by reference to research without the details of the research. So let's think a bit about the story being told. A physician finds a small town with a low rate of heart disease that also has a tight knit community. Surrounding areas with similar environmental hazards don't show these low levels of heart disease. Should I be surprised? Well, if you examine the relevant data you find that Rosero had about half the age adjusted mortality rate as that of neighboring Bangor. A closer look at that paper, it shows that this data it gives 95 percent confidence levels for the ratio of age adjusted mortality rates of Roseto and the nearby Bangor. For heart attacks in three decades the data shows that the ratio is statistically significantly at this level.

But this is where my ears perk up a little, not enough to really make me question the Roseto effect, but enough to put me in a bit of a fowl foul statistical mood. How many cities are there in the United States (somewhere in the neighborhood of 20000. How many of these are smaller than 10000 people? Well as of 1998 the number of towns in population great than 10000 was 2500. So even working backwards to the relevant time periods of the study one would guess that there is somewhere in the neighborhood of 10000 small towns which might potentially be a candidate for a study such as the one performed in Roseto. If we assume that age adjusted mortality rates for heart disease are normally distributed among cities, we thus might expect somewhere around 250 towns with abnormally low ratios like that of Roseto to Bangor (that's just a rough guestimate given the 5 percent significance, working with one side of the distribution, this means the number is perhaps a bit too high.) Of course the beauty of the story is that Roseto is a small town with a unique community culture, and putting a guess on how rare this occurs, is a hard task to put a number on. But if this number is anything like one in a hundred then I would be surprised if a place like Roseto doesn't exist, just in the United States.

Of course what is even more interesting is that the effect persists over three decades. It's a bit unclear to me what time period one should take for this effect. For example if we assume that these should be thought of as three separate instances of the Roseto effect, then the above statistical observation about the rarity of such deviations isn't as strong. My guess, however, would be that the effect should be binned approximately on a generational time scale, which is around 25 years in the United States.

This is, of course, not to say that I don't buy the overarching hypothesis of Wolf and Bruhn. But the work of just Roseto is not enough to convince me. So has there been other studies that take a larger approach to how community can effect community in the way Wolf and Bruhn hypothesize? Well as I said in the introduction, I'm not at all qualified to answer this, but a little searching did produce some interesting results. In particular I certainly learned a lot from reading Social capital and health promotion: a review by Penelope Hawe and Alan Shiell (2000). That paper has a ton of great references, but let me just quote from the conclusion:

We have suggested that the science of social capital, by which we mean its empirical capacity to explain health patterning, is relatively weak at present. The
concept is too broad relative to more precise, alternative constructs.
...
Progress will only be made, however, with cross-disciplinary literature reviews and more rigour in the translation from theory to measurement

Which is, to me at least, a nice way of saying there could be something here but we're not quite convinced yet. Note that this is fairly far from Gladwell's interpretation which is that this is well established, and by caveat, important.

And here, at the word "important", we get into a question of quantification of the size of the effect. Because the study of Wolf and Bruhn is limited to one city we can't get an idea for how large of an effect this is in comparison to the other effects listed by Gladwell. An effect can be unexplained by correlations with other factors, but still be small in size. And this, it would seem to me is an essential point to Gladwell's argument. Gladwell says

Living a long life, the conventional wisdom at the time said, depended to a great extent on who we were-that is on our genes. It depended on decisions we made-on what we chose to eat, and how much we chose to exercise, and how effectively we were treated by the medical system. No one was used to thinking about health in terms of community

But from my cursory understanding of the Roseto work, it isn't at all clear that one can draw a conclusion as to the relative size of any community effect versus all of the other factors, like genetics, diet, etc that are listed. For example, I've seen references (here's one) to studies where dietary changes account for a drop in death rate by heart disease of around twenty percent. What, if any, is the size of the Roseto effect compared to these factors? If one is going to "explain" how to live a healthy lifestyle and one leaves out the effects of diet, genetic, income, and exercise, would that be a good explanation? Somehow I'm a bit skeptical.

Improv's view: If a curmudgeon looks all ideas dead in the eye and says "I am skeptical", an improv's job is to take the ideas presented, not disagree with them, and then run with where ever they might lead.

And now, your highness, we will discuss the location of your hidden rebel base...

Take the Roseto effect, then, and multiply it a millionfold. What if, all of your health, were dependent exclusively on social structure of the community you belong to? I think what I'd like to know about such a society was how it dealt with rebels. You think you've got it bad, young rebellious high schooler, just imagine if your rebellion caused your friends to die! Get your ear pierced, give your dad a heart attack (okay, well, this one may hit too close to reality.) This reminds me a bit of these lines from Roger Waters Pink Floyd's "The Final Cut":

Tell me true, Tell me why, was Jesus crucified?

Was it for this that daddy died?

Was it you, was me, did I watch to much T.V.?

Is that a hint of accusation in your eye?

Of course, watching too much T.V. is exactly the kind of activity that does sort of kill you, since it tends to lead to an unhealthy lifestyle, but what if it wasn't this effect, but the fact that you were alone watching television instead of being out on the streets talking with your neighbors that killed you? I'm guessing that the pull of American Idol on the world would be a lot less powerful in such a universe, but then again, the power of Simon and Paula is pretty damn strong.