De Econometrist neemt een statistische kijk op de wereld.
When you were still a little child, you were probably told that all babies were delivered by storks. Later, when you got older, you found out that this idea is total nonsense. However, imagine that you are still in doubt whether this idea is true, and to discover the truth, you want to use your statistical abilities. As it turns out, from a statistical point of view it seems that storks really do deliver babies. Or does it not?
In the year 2000, Robert Matthews, professor at Aston University, published an academic paper on the correlation between the birth rate in a country and the number of storks. The paper is even called “Storks Deliver Babies (p = 0.008)” where the p stands for the p-value. And as econometricians, such a low p-value indicates to us that the effect of storks on the birth rate is statistically significant. So how did he get to this seemingly strange result?
In his research, Robert Matthews took a look at 17 European countries. From these countries, he collected the birth rate and the number of stork breeding pairs. Using this data, he performed a linear regression where he used the birth rate as the dependent variable and the number of stork breeding pairs as the only explanatory variable (next to the intercept). When replicating his research, I created the following graph to illustrate this effect.
In this plot, we see the linear regression line through the number of data points. It is definitely not a perfect fit, however, it does not perform poorly. There are no countries with a very high birth rate and a very low number of stork breeding pairs and vice versa. This explains the high statistical significance. However, there is of course a crux to this story. When we add a variable like the size of the area of a country, the statistical significance of the number of storks disappears. Hence, there is some omitted variable bias here. When a country spans a very large area, there are in general more storks.
This example of the correlation between storks and birth rate can also be found in the book “How to Lie with Statistics” by Darrell Huff. It is one of the best-sold books ever in the field of statistics and it shows that we should not just focus on the numbers but also use our heads. More about his work can be found in this article. As you might have heard earlier, correlation does not imply causation and this example shows that clearly. More funny examples from this can be found at https://www.tylervigen.com/spurious-correlations. Here, we for example also see a strong correlation between the number of people who drowned by falling into a pool and the number of films Nicolas Cage appeared in.
So, the next time that you receive a birth announcement card with a stork on it, remember that correlations are useful, but you should use them with care.
Source: Matthews, Robert. (2000). Storks Deliver Babies (p= 0.008). Teaching Statistics. 22. 36 – 38. 10.1111/1467-9639.00013.
Dit artikel is geschreven door Stan Koobs