Freakonomics and Complexity

The best-seller Freakonomics by economist Steven Levitt and journalist Stephen Dubner never mentions the word “complexity” in its 200+ pages of entertaining correlation analysis, but at its heart it is a book about complex adaptive systems, and about the consequences of our reckless passion for treating them as merely complicated. In fact, it’s more about statistics and social studies than it is about economics.As a reminder, the difference between complicated and complex systems is that only the former are completely knowable, analyzable, and subject to rigorous cause-and-effect analysis. Complicated systems are the left-brainer’s dream: To decide what to do all you need to do is identify all the variables, determine which causes which (using root cause analysis), draw systems thinking diagrams to depict the relationships, assess the possible points of intervention that could lead to a different and desired result (e.g. turn a self-reinforcing vicious circle into a virtuous one), recommend those interventions and collect your fee. Very scientific, and lots of fun. Unfortunately, in the modern world, complicated systems are fairly rare.

Complex systems are the rule, and they are not completely knowable or analyzable because the number of variables is essentially infinite, and hence the consequences of any particular intervention are largely unpredictable. You need to use a more sophisticated, less scientific approach when you’re dealing with complex systems, and be more tentative in your assessments. Freakonomics deconstructs some of the many erroneous and dangerous assessments we tend to make, and actions we therefore take, when we treat complex systems as merely complicated. Its authors tell us “look beneath the surface” to discover the complexity within, and the concept is represented on the cover as an apple with the (unexpected) insides of an orange.

Dave Snowden tells us that the approach to complex systems is “probe, sense, respond” in contrast to the “sense, analyze, respond” approach appropriate for complicated systems. Analysis is futile, but that doesn’t mean the probes beneath the surface can’t provide us with useful and compelling information that can allow us to act in a way that will most likely be helpful and positive. Dealing with complex systems requires pattern recognition, Snowden says. Our long-term memory has a capacity of about 40,000 patterns (models, archetypes, plans, idealizations and other representations of reality), and when we see, hear or otherwise pay attention to something we only perceive and internalize the 5-10% that resonates and is consistent with those patterns, that understanding of reality. There is evidence that until someone creates a mental pattern for a phenomenon, they are unable to ‘see’ it at all. And once a pattern has been set in the brain, it becomes very difficult to dislodge. So when we see what looks on the surface like an apple, we can’t even conceive of it being an orange inside. Freakonomics probes deeper than we normally do, challenges any assumptions about causal relationships (since those assumptions may be oversimplifying complex systems as merely complicated ones), and looks for the patterns that the rest of us can’t or don’t see. Some examples:

  • The cause of the recent drop in the US crime rate can be explained by four things, but despite scholarly works to the contrary, innovative policing practices, gun control laws, capital punishment, gun buybacks, a strong economy and an aging population aren’t among them. The largest contributor to the drop in crime rate was Roe vs Wade. Read the book to find out why (and contemplate the consequences if Bush stacks the supreme court to overturn it)..
  • Money doesn’t buy elections. While having money might secure you the nomination of a major party, once you’ve got the nomination the amount you spend against your opponent has no bearing on your likelihood of winning.
  • There is overwhelming evidence of cheating by teachers as well as students in No Child Left Behind standardized tests, and also overwhelming evidence of performance-enhancing drug use in many sports, including (although the book does not provide details) the Tour de France.
  • Your child is 100 times as likely to suffer harm visiting the home of a friend with a swimming pool than one with a gun in the house.
  • While the per-mile death rate of driving is much higher than that of flying, the per-hour rate is about the same.
  • Your young child is less likely to be harmed in the back seat with a seat-belt than in the front seat with a car seat; in fact, neither car seats nor cribs have any significant impact on the incidence of harm to children.
  • While who you are as a parent (your genes, and perhaps your passion and your example) has a significant effect on the chances your children will succeed in life, what you actually do with your children (including reading with them) does not.

Some of these findings are provocative (in fact the first and fifth have created a lasting furor). But the authors are not attempting to argue causality here, or even about the wisdom of doing certain things. They are probing, using correlation and regression techniques applied against huge amounts of data, to show what correlates with what, and in the process debunking a lot of myths about causes of and remedies for a lot of problems in our society. What they are doing is eliminating as many factors as possible, so that they can say with reasonable assurance that all other things being equal, there is a very high, significant correlation between X (e.g. availability of abortion) and Y (e.g. subsequent declines in crime rates) — or that there is not. Draw your own conclusions.

The authors are great believers in another principle of dealing with complexity — the importance and value of attractors and barriers (which Freakonomics calls incentives and disincentives) in bringing about desired actions or behaviour change. They believe you can learn a great deal by studying which attractors and barriers actually work in other situations (by ‘work’ they mean that the introduction or existence of attractors and barriers, whether natural or man-made, correlates powerfully, all other things being equal, with a subsequent desirable behaviour change. The attractors and barriers to jobs, for example, largely determine who goes into different fields, and two attractors (that it requires specialized skills you have, and that it is in high demand) and two barriers (excessive supply and unpleasant working conditions) correlate most with what the job pays. That’s the reason why, the authors say, most crack dealers still live with their mothers (excessive supply of applicants for the job) and why prostitutes earn more than architects (‘danger pay’, higher demand and relative shortage of supply).

But they also warn against the failure to consider all of the alternative variables that might have led to that condition or behaviour. For example, children with certain names tend to end up with significantly higher education and income than others, but that doesn’t mean giving your child one of these names will make their life easier. In fact, the propensity to give your child certain names correlates with a variety of other factors (such as your own education and income) which in turn correlates with your child’s success. Be careful about jumping to conclusions.

There are two remarkable quotes in the book. The first is this one by economist John Kenneth Galbraith about the follow of ‘conventional wisdom’:

We associate truth with convenience. with what most closely accords with self-interest or personal well-being or promises best to avoid awkward effort or unwelcome dislocation of life. We also find highly acceptable what contributes most to self-esteem. Economic and social behavior are complex, and to comprehend their character is mentally tiring. Therefore we adhere, as though to a raft, to those ideas which represent our understanding.

The next quote, on the very next page, is by Paul Krugman and provides a perfect example of Galbraith’s point:

The approved story line about Mr. Bush is that he’s a bluff, honest, plain-spoken guy, and anecdotes that fit that story get reported. But if the conventional wisdom instead were that he’s a phony, a silver-spoon baby who pretends that he’s a cowboy, journalists would have plenty of material to work with.

So, say the authors, we must be wary of conventional wisdom, skeptical until and unless the data strongly supports what we are told or what we believe. Thanks to the Internet, they say, some of the ‘information asymmetries’ that lead to unsupportable conventional wisdom are disappearing. But what Malcolm Gladwell calls ‘learned helplessness’ is still with us — our inability to go beneath the surface, challenge and debunk conventional wisdom or instinct leads us to dysfunctional beliefs and actions: That we’re safer in an SUV than another vehicle, for example, or that we should spend more money and effort trying to prevent terrorism happening in our countries than we spend trying to prevent common bacterial and viral infections, for example.

And if we want to bring about real change, we need to consider what attractors and barriers we can influence that will really affect behaviour. Those attractors and barriers can be economic (e.g. a tax shift that encourages domestic employment and penalizes waste of non-renewable resources), or social (e.g. an award or promotion, or incarceration for breaking a law), or moral (e.g. an appeal to our sense of fairness, or right and wrong). The authors also recommend what they call ‘bright-line’ attractors and barriers (those where the attractions and constraints are very clear) over those that have consequences that are ‘murkier’ (less clear or less certain).

Let’s suppose we want to bring about a significant drop in birth rates worldwide. First, we would need to probe to find out why they are currently as high, and as low, as they are. We could offer economic incentives to have smaller families — though we should start by studying whether we already have them (studies suggest that, worldwide, women have on average almost one child each fewer than they would like, and cite the cost of having children as the overwhelming reason for that decision). We might offer social incentives for smaller families — like awards or special opportunities available only to childless couples. Or we might offer moral incentives for smaller families and disincentives for large ones — by pointing out how much large families contribute to our unsustainable way of life, or by having leaders and the media publicly repudiate the reactionary pope and other religious leaders who encourage large families, and suggest that the followers of such religions are weak and irresponsible. Levitt and Dubner would have us believe that these would be far more likely to work than political or educational methods.

But alternatively we could look at the reasons why current birth rates are where they are now, and identify incentives and disincentives that might address those reasons rather than the decision on how many children people choose to have directly. The financial pinch motivation isn’t a helpful one — there are no attractors or barriers we can use to exploit it, short of deliberately trying to plunge the world into an economic depression (and Bush and Greenspan are working hard at that). In the third world, many women claim they have large families because, in the absence of a social safety net, it’s the only way they can hope to make ends meet in their senior years (and in many cases, the only way they can make ends meet period — children are the only assets they have). Now that’s something we can do something about: By providing incentives to third world countries to provide universal health care, education, old age pensions and social assistance programs for their citizens, we might dramatically reduce family sizes in the third world quite quickly. Of course, we’d need to confirm that such incentives actually work, but we could do that by studying planned and actual family sizes in countries that have significantly improved social services, controlling for other variables. And we’d need to look at the fact that, regardless of what they might say, the first and (to a lesser degree) second generations of immigrants to countries with comparatively good social services continue to have a large number of children, and understand why that is.

The point is, the means to bring about change is hiding there in the information in that orange beneath the apple peel. And while many readers have found Freakonomics either entertaining or outrageous, and are focused on the specific examples in the book, the real importance of this book is that it lends credence to complex adaptive systems approaches to understanding why things are the way they are, and how they might be made better — through mechanisms that, when we fail to look below the surface and allow ourselves to be blinded by conventional wisdom, we might never have considered.

This entry was posted in Our Culture / Ourselves. Bookmark the permalink.

10 Responses to Freakonomics and Complexity

  1. Elizabeth says:

    Do you have a cite for the studies that say that women have fewer children than they’d like? I’ve heard the claim before, but never found the studies it’s based on.ThanksElizabeth

  2. Some of this can be misleading to people who don’t understand what statistics and averages are. For example, if we take: “what you actually do with your children (including reading with them) does not.” It doesn’t mean that with your particular children it won’t have any effect (positive or negative), even if in a statistically relevant group is has little incidence.

  3. dave says:

    Actually, Levitz’ “abortion cut crime” theory has been thoroughly trampled. That generation of post Roe v. Wade kids went on the biggest murder spree in the history of western civilization. Read here : http://www.isteve.com/abortion.htm

  4. Dave Pollard says:

    Elizabeth: I talked about this in this earlier article and cited the studies I used at the bottom of that article. I’m aware of the fact some studies say the opposite, but they are largely studies on the availability of contraception, and though well-intentioned are suspect for some of the same ‘conventional wisdom’ reasons Freakonomics describes.Mike: There are always exceptions, and every individual is different. The point is, don’t beat yourself up if you read to your children and they became axe murderers, (or if you didn’t, and they became axe murderers)– chances are what you did or didn’t do with them did not affect that outcome.Matt: I think it’s Steve Sailer’s argument, which you cite, that has been thoroughly trampled. Here.

  5. lugon says:

    Dave – maybe you have some insight on how to aproach pandemic preparedness regarding a possible (even likely) influenza pandemic? http://www.fluwikie.com http://effectmeasure.blogspot.comThanks!

  6. bdot@bill-wilson.net says:

    Hi Dave, I just wanted to say that complex systems do not defy causation or analysis. It’s just that typical analytical techniques are focused on direct-line causes, which doesn’t work for complex systems where system conditions (contributing causes, latent weaknesses, etc.) are much more important. In a complex system, an initiating event just provides a spark to get the fire going. Typical root cause analysis using a technique like “five whys” will focus on how and why that spark came into existence… which has almost no relevance on what the consequences of an event will be. Just a thought. Regards…

  7. Dave Pollard says:

    Lugon: The advice we got repeatedly during the SARS outbreak here was (1) wash your hands with soap and water often, and (2) stay away as much as possible from hospitals, doctors’ offices and other places frequented by sick people.Bill: Thanks. I continue to use root cause analysis as a technique where the number of variables is finite and definable. It can be used as well to provoke areas of inquiry in complex situations, to ascertain what correlations to investigate. What I was trying to say was that such causation is impossible to prove (just as you can’t prove the flap of the butterly wing in one country ’caused’ a hurricane in another, or that el nino ’causes’ various climate phenomena. You can’t get further than finding a compelling correlation, and then deciding what to do in the circumstances. That may lead you, if you’re in law enforcement, to work for liberalized abortion laws (it would me). Or it may lead you, if you’re a parent, to fill your bookshelves or name your child one of those ‘successful’ names (also highly correlative, but in my opinion not causal and hence not worth doing).

  8. lugon says:

    Dave, I didn’t just mean “ways to slow down transmission”, but also (perhaps even more importantly) “ways to cope with social disruption”. Some fear it might be a bit like a global war.

  9. Elizabeth says:

    Thanks for the link — I’m not entirely convinced, but I appreciate the background info.

  10. Hi Dave, I use the barrier and attractor approach to intervention design. For example, I was running a workshop last week using this technique to tackle a staff morale issue. One of the practical things I’ve noticed using the technique is the difficulty in using the term ‘barrier’. While the barrier referred to in the approach is essentially the container of the phenomena of interest, the word ‘barrier’ is translated by participants as ‘a problem.’ I’ve talked about this problem on my blog (http://www.anecdote.com.au/archives/2005/07/barriers_and_at.html). I now use the less accurate but more useful term, ‘boundaries’.

Comments are closed.