# Does age matter in politics?

**Posted:**April 1, 2011

**Filed under:**Own projects |

**Tags:**age, many eyes, vaalikone, visualization, yle 3 Comments

I did a small analysis on the Yle vaalikone data tonight. I wanted to find out if young and old politicians differ in their opinions. Can I, as a fairly young voter, expect to find my political soul mate among my peers?

Well, that seems to depend on what questions are important to you. By calculating the correlation (I just used the correl-function in Excel) between the age of the candidate and the answers given in the voters advice application you get an idea of how well political opinions and age are linked together. The bigger the correlation, the wider the gap between young and old:

The strongest correlations here are between 0,2 and 0,3. What does this mean in practice? Lets look at one of the questions more closely. The biggest difference between old and young candidates appears on the question about whether registered homosexual couples should have the same rights heterosexual couples. I did a visualization in Many Eyes to illustrate the difference: (click to open interactive full size version in Many Eyes)

So, do you think gay rights are important? Start looking among the young candidates for your pick. Is Nato, the length of the work day or the right to strike the most burning questions? In that case age is really just a number.

Very interesting visualization. For someone who has forgotten everything about statistics it would be nice to hear how strong a 0.3 correlation is? It would also be good to have a horizontal scale on the diagram.

Great. Thanks

I think the graph below gives some sort of idea of how big the difference is. If I remember my statistics courses from university right, one could say that a correlation of 0,3 explains 9 percent (0,3*0,3) of the difference (variance) between the two variables. That means 90 percent of the difference between opinions are explained by other factors, such as party affiliation.

This is interesting graph, but there might be very difficult statistical problem here.

The correlation is so low, that it barely has any significance in this matter.

Furthermore: the statistical test Excel uses for correlation is Pearson correlation. If I remember correcly from my studies, the variables should be at least interval scale. Age is that, but opinion is of nominal scale. (You can’t add or subtact the answers on nominal scale, nominal scale is like comparing apples and oranges.)

So, if I remember correctly, the results you have for correlation are meaningless. You have numbers in your data, and that’s why Excel gives you any results. If the results would be coded in alphabet, like A, B, C, there would be no result at all.

There might be a possibility to code the results as dummy-muuttuja (dummy variable, one that gets values 0 and 1). But I’m not sure, if you could use Pearson correlation even for this test.

The many eyes -visualisation has nothing to do with correlation. So that is valid and very interesting.