Rabbiting On Some More

The Economist recently published a nice article and infographic detailing the average speech length for each country speaking at the UN. In this article we tweak the analysis a bit to look at one underlying variable, GDP.

The Economist graciously gave us access to their source data which we proceeded to augment with GDP data and then aggregate into counts of speaking engagements, rather than the average speech length used in the original article.

The data set extended from 2010 to 2015. Instead of taking average GDP across these years we took a simpler approach and obtained the 2013 GDP for each country, measured in 2016 US dollars. The final data set is available as a csv here.

We see a positive but insignificant correlation on the DataSplash Correlation tab. However, on visual inspection of the scatter plot we see that there are two large outliers in terms of GDP which seem to be influencing the trend line, the USA and China. On the left the data becomes cluttered, but we can still identify each country by hovering over the point with the mouse.

In DataSplash, we can easily removing these two outliers. Simply by right-click on the dot associated the data point you want to delete.

Having removed these outliers, we see a new relationship which is both significant and more important. The old relationship is shown with a dotted line and the new regression line is shown with a solid line.

After a brief exploration of data, analysts will often attempt a more causal explanation. Perhaps the poorer countries don’t feel empowered to yell at the richer countries to quit talking, or perhaps the richer countries have acquired extreme amounts of wisdom and therefore it just takes them a bit longer to communicate. Clearly in the case of the French we also have a cultural explanation.

Feel free to investigate other patterns. You may notice, for example, that GDP squared is significant in the opposite direction of GDP, indicating a decreasing marginal effect. Eventually this negative marginal effect drives a country under the simple linear trend line, so very wealthy countries may not rabbit on quite as much as otherwise expected.