Can the world be divided up into civilizations?

Political scientist Samuel Huntington predicted in his 1996 book that conflicts in the future are going to be due to cultural and civilizational as opposed to ideological or economic differences. He then proceeded to divide the world up into some major civilization groups.

In this post, using answers to the World Values Survey I check whether the data supports Huntington’s civilizations.

Huntington divided the world up into nine main civilizations. The map based on this can be seen below (source: Wikipedia). This division of the world has already received some verification from data. For instance, State et al. (2013) confirm the groups using international email flows. Now I attempt to test with World Values Survey data whether the existence of different civilizations can be explained on a more fundamental level: cultural values.

Huntington: Clash of Civilizations map

The way I test whether data supports this categorization is by looking at the World Values Survey. I use 122 questions from the survey collected in 77 countries. I have already worked with this data set in previous posts. For instance, I developed an online questionnaire that compares the respondent’s values to all countries, and also compiled a cultural distance matrix that shows the cultural distance between any two country pairs.

To see whether countries can be grouped into civilizations in a way similar to Huntington’s classification I use k-means clustering. This is a statistical technique that basically divides the data points into k clusters, where k is a number chosen by the researcher (i.e. me).

The main idea behind this method is that it creates k clusters in a way that minimizes the within-cluster variation. Within-cluster variation is basically a measure of how similar the data points (in our case, countries) within a cluster are. The aim is to minimize the differences within each cluster while also separating the data into exactly k clusters.

Now, it is tricky to decide what the number of clusters should be. I experimented a little and I found that 7 clusters provided the most readily interpretable result. I found the following clustering.

Culture / Civilization clusters (k-means)

Let us compare this to Huntington’s results.

There does seem to be an Eastern/Oriental civilization, though I do not detect the distinction between Sinic, Hindu and Buddhist cultures. It must be noted that the number of countries in the sample from this region is small.

The Western civilization largely checks out but it should exclude former socialist countries from Eastern Europe, none of which are in this cluster.

There are indeed clear Muslim, Orthodox, African and Latin American civilizations.

Finally, there is a residual group including various countries. Chief among them are former socialist Eastern European countries, which seem to form their own group, and interestingly the more developed Latin American countries of Chile, Argentina and Uruguay; the Southern European trio of Italy, Cyprus and Spain and Taiwan and Japan from Asia.

This last group is interesting, I labeled it “transitory”. While I’m almost certain that Taiwan and Japan are just statistical artifacts in this group, the other countries could be ones that are en route to becoming part of Western civilization, but have not quite arrived yet. Indeed these countries are almost all transitory economies or have become advanced very recently. Even though Spain, Italy and Cyprus are normally considered advanced, they might be in this group due to high social conservativeness relative to the rest of Western civilization.

To check the robustness of this clustering, I will turn to another statistical method for grouping data points into clusters: hierarchical clustering. The idea behind hierarchical clustering is simple. Start with putting each data point into a cluster of its own. Then join the two clusters closest to each other. Repeat until there is only one cluster. Two issues arise in this process: first, we will still have to decide at what point we want to stop joining clusters; second, how to calculate the distance between two clusters.

Using one particular measure of distance (complete), I plotted a cluster tree below. This basically tells us how the countries were grouped together. Remember the algorithm starts at the end (at the leaves) where each country is in its very own cluster. And then it starts joining them together. This process can be very well followed using the diagram (actually it’s called a dendrogram) below (click to enlarge).

Culture / Civilization clusters: dendrogram

Hierarchical clustering also supports the results of k-means clustering. From left to right, we can identify the Transitory group, the Western group, the Muslim group, the African group, the Latino group, the Orthodox group and finally the Oriental group.

An advantage of representing the results using the figure above is that one can see the distances between various groups more clearly. For instance, the Transitory and Western groups are indeed very close to each other as predicted above. We can also verify that Japan is a very standalone culture. And that if we cut the tree quite high up to form only three clusters we’d have a Western-Transitory cluster, a Muslim-African-Latino cluster, and an Orthodox-Oriental cluster.

Broadly speaking therefore, I would argue that Huntington’s civilizations groups were pretty much correct and are supported by this data set. An interesting question is what the long-term trends in these clusters are going to be. As I said before I am pretty sure that the Transitory cluster will slowly become part of the Western one, but what about the others?

The Muslim, African and Latino clusters seem to form a group of the most socially conservative clusters. As/If these regions develop, they might abandon some of these values slowly and merge into other clusters. I could particularly imagine the Latino cluster merging into the Western one, but this probably won’t happen before the end of the century.

The Orthodox cluster might lose some countries to the Western one, but overall I think it will stay intact. The Oriental cluster will probably stay intact as well.

Finally, note that this analysis is not really asking whether the cultural differences between each of these clusters is so huge that it could lead to potential conflicts along cluster-lines. It might be that even though we have these differences, they are rather small or at least getting smaller. But if the world were to be completely “clusterless”, then we should not see such clusters that have so clear boundaries. I.e. in a clusterless world our results would be almost completely random clusters that are hard to make sense of or label in a straightforward way (such as Western, African, etc.). For this reason, my opinion is that indeed the world can be divided up into several distinct civilizations in a more or less robust way.

Advertisements

2 thoughts on “Can the world be divided up into civilizations?

  1. This is an interesting article. With China’s expansiveness, it’s really hard to place the country in one area considering the technological east and very poor and varied ethnological strata. Japan is also definitely in its own area as of now. Being on an island and having created its own self sustaining language based on Chinese gives it its own values.

    Huntington’s model seems to hold true now, but whether it holds true 500 years from now will be something to see.

  2. Very accurate. In latin america, the guyanas are very isolated, they speak a language other than spain or portuguese. And Japan is very standalone indeed In culture as well as in language.

Leave a Reply

Fill in your details below or click an icon to log in:

WordPress.com Logo

You are commenting using your WordPress.com account. Log Out / Change )

Twitter picture

You are commenting using your Twitter account. Log Out / Change )

Facebook photo

You are commenting using your Facebook account. Log Out / Change )

Google+ photo

You are commenting using your Google+ account. Log Out / Change )

Connecting to %s