Wednesday, June 17, 2009

For a large country, the US does alright

Via Infosthetics, the above graph (culled from the Flash visualization here):
...maps the base pay of the world's parliamentarians, ranked by country and expressed as a multiple of per capita GDP, versus a "Good Governance Index", which itself is a combination of the Democracy Index, the UN's Human Development Index and the Perception of Corruption Index.

In short, the further away the country dot is removed from the yellow cross, the more their MPs are being paid. The larger the (counter-clockwise) angle from the yellow line, the worse their corresponding governance.

Some commenters on the Infosthetics post have decried the visualization as unnecessarily flashy, but I think it's actually pretty useful, at least for relative comparisons. The best cluster of countries in this analysis are all countries with small, fairly homogenous populations such as Switzerland, Iceland, Luxembourg, etc. Not all such countries are in the same cluster (notably Austria and Belgium), but the biggest country (population wise) in the best cluster is Canada, with about 33 million people.

The second-best cluster of countries includes much larger countries, including the US, the UK, France, Germany, Spain, and Japan (Japan is the line that extends to the right out of the image). In general, these countries are more also more diverse (namely the US, the UK, and France).

So, in general, the trend seems to be that smaller, more ethnically homogenous countries get more for their money than large, ethnically diverse countries. This isn't all that surprising. But what was a little surprising for me was how relatively well the US does, considering that it is both the largest country in the upper-right quadrant and also the most diverse. Of course, it should be reiterated that the US' position on the graph is likely heavily influenced by the fact that the base pay is normalized per GDP, and the US' GDP is quite a bit higher than some of the countries in its cluster.

So that's what an A/B test is...

The other day I was reviewing this great Alertbox article on user experience methods, and I came across a method with which I was unfamiliar--A/B testing. Well, sure enough, just like when you learn a new word and then it suddenly seems that you hear it used a couple times within the same day, this 37 Signals post points to a real world example of A/B testing. In this case, Less Everything simply measured the number of "conversions" (I guess the number of people who tried the application for free?) when a button was placed on the left side of the screen versus when it was placed on the right side and found that putting the button on the right led to a conversion rate of 13.8% versus a conversion rate of 12.3% for when the button was on the left.

The study seems to follow the guidelines for A/B testing in that the sample size is large ("thousands of users") and it was performed on an essentially finished design where it was easy to change only one part of the design and observe the effect of that change in isolation.