Tag Archives: statistics

The Best Example of Bad Graphs

Those of you in the UK (or elsewhere) who read the Telegraph may have noticed an article recently called “How the Fed triggered the Arab Spring uprisings in two easy graphs”. If you want the gist of it, here is the first graph:

Now, as most economists learn, correlation does not equal causation. But let’s assume for the author’s sake that he has a point. What should we think when we see commodity prices rising (starting June 2010) and the Fed’s Treasury purchases rising after the fact (around Aug-Sept 2010)? Yes, if we want to read causation, then we’d say that the rise in food prices caused the Fed Treasury purchases, and not vice versa (as the article claims).

The second graph, which we show below, displays the correlation between, well, we’re not sure. Continue reading


Few things annoy economists more than anecdotal evidence. Just think of “mockumentaries” and various news headlines that don’t document “how many” or “how much”, but rather “what happened to a certain person”. It is easy to capture people emotionally with one, drawn out story, but this also goes against an economist’s training in statistics and probabilities. See how Dan Ariely complains about the $48,000 which was spent to rescue one terrier left aboard a ship, only because the dog was featured on local news channels.

Ultimately, and despite Stalin’s morbid quote that “A single death is a tragedy, a million deaths is a statistic.”, economists are actually correct in wanting to work with concrete data rather than tear-jerking anecdotes, at least for their purposes. This is the main reason why economists have tended to look down upon the humanities, which always seemed to be characterized by fluff.

This, however, might be changing, thanks to Google and Jean-Baptiste Michel of Harvard University. Michel has been digging into the data available thanks to Google books and has been able to start putting quantitative values to cultural studies, such as linguistics. You might not think much of interest could emerge from this, but thanks to this digital collection, we can analyze how often words appear, as well as group of words, and we can study the evolution of words and phrases throughout time. In fact, for the first time we can really estimate how many words truly exist in the english language (apparently slightly over 1 million). We can also find the occurrence of typos, how long it takes new inventions and phenomena to become used in everyday language, how quickly celebrities and pop-culture references fade in and out of existence, among many other things.

We think this is fascinating, and we also welcome the addition of concrete data to fields which seem to abound with anecdotes and personal stories. We do, however, think the term “culturomics” sounds off-putting. Maybe “culturnomics” might work better?

And this also gives us the excuse to show this awesome Google lab “Books Ngram Viewer”: http://ngrams.googlelabs.com/graph?content=computer%2Ccar&year_start=1800&year_end=2000&corpus=0&smoothing=3

I can see many economists, statisticians and mathematicians taking a renewed interest in the humanities in the near future.

Using Economics for Real Life

Forbes recently had an excellent article on Alvin Roth entitled “Un-Freakonomics“. Alvin Roth has created almost a cult following by doing the opposite of what Dubner and Levitt achieved with Freakonomics. Rather than use ‘the dismal science’ to find out whether Sumo wrestlers are cheating or not, he uses economics to do things like save lives.

This is a rarity in itself since most theorists do not also work in the field, but Roth isn’t scared of rolling up his sleeves and applying his theories. One example was with the New York city high school match. Roth himself was a high school dropout from Queens, and knew that the New York system of matching students with schools was a complete mess. In fact, what happened was you would rank your top 5 high schools and mail them in, and these lists would then be mailed to each high school. So each school knew if they were your number 1 choice or not. He took a variation of an algorithm which matches men and women who might want to marry, and applied it to students and high schools. This same system was then used for Boston primary schools.

Another accomplishment was designing a system for kidney donors. Many times a loved one will want to donate a kidney but not have a matching blood type. His system allows them to exchange with another relative of another patient who is in the same predicament. Even Dan Ariely, of Predictably Irrational, states “I’m his biggest fan.”

We would also like to take this opportunity to remind everyone that our new book “Bringing Sexy Back to Economics“, uses economics to analyze serious matters such as gun laws, providing water to drought stricken areas and finding ways of feeding the world. The more Economics can do to change the world, the more sexy it will be.

Zipf’s Law

Zipf’s Law is a statistical curiosity that has been observed time and time again, although there is no determined reason for it. It was first observed with word occurrence, and simply put, it states that the frequency of a word is inversely proportional to the word’s ranking among all other words. In other words (sorry), the word with the highest frequency will occur twice as often as the word with the second highest frequency, three times as often as that with the third highest, etc.

It seems that the larger the text, and the more fluid the language (if it is, say, a book with complete sentences, rather than a bullet-point how-to guide), the more this law holds true, independent of which language the text is in. Even more interesting is that it seems to hold true in other fields as well, such as city population, income distribution and size of corporations.

Zipf’s law appears regardless of what the initial plans were for any city (or text or income level), and appears to be completely natural. Also, interestingly, it has only been observed in human-related phenomena, but has yet to be seen anywhere in nature (e.g. with tree growth or size of ant colonies, etc.).

Hattip to the nytimes blog for sparking the interest.