Systematic and people-oriented
methods for strategy, improvement and innovation
Freedom of Expression and Absence of Corruption: a winning cocktail?
April 14th 2022, by Michael Ohler
What this study is about
In a previous study, I have shared an analysis of data gained from the “State of Democracy” website. Among the many indices the organization tracks for all countries and all years since 1975, I have merely explored the freedom of expression. In this study, I am looking at two indices together: the freedom of expression and the absence of corruption.
- These two indices are good at telling apart forms of government (democracy, autocracy, …)
- Improvement, and degradation, of the two indices is correlated
- “Static” reports like this pdf-document are “clumsy” at analyzing these data for a large number of countries.
1 Freedom of expression and absence of corruption
If you are lucky like I am, then you live in a country with strong institutions and a strong civil society. In that case you may not even notice how valuable clean elections, civil liberties, or the absence of corruption are - to name but a few attributes of democracy that matter to people. Some of my friends aren’t that lucky. They live in countries where elections are rigged, civil liberties trampled and corruption is rife.
In this post, I want to extract insights by looking at two indicators together. I do that with three intentions:
- Talking to people who probably know and decades of reading newspapers lead me to the assumption that corruption paired with the suppression of free expression can become “toxic”. I want to understand what countries are most at risk. I also want to test how valid my assumption is.
- I’d like to explore what graphical tools can be used to visualize the status and the evolution of multiple, if not “all” countries, and this for any combination of two indicators.
- On the mid-term, I want to provide others with data-visualization tools to explore themselves the attributes of democracy and compare different countries.
Let’s go.
2 First trial: two attributes of democracy, compare many countries
As a first attempt, I build on the results from my previous study. In that study, I have developed an algorithm that allows exploring trends over a selected span of years. The idea here is to recycle that toolkit and to
- plot the status of the two indicators on x- and y-axis
- color-code the resulting dots with the identified trends
- use the speed of change (i.e. the combined slope of the two trends) for the bubble-size and
- attach a little arrow so the user can read in what direction the evolution of the past years predicts the country may move next.
For the color-coding, I use
- red for “degradation of both indicators”
- green for “improvement of both indicators”
- orange for “undecided” (one indicator improves significantly and the other deteriorates significantly).
I also color as green if one indicator improves (green) and the other shows no trend (which would otherwise be colored as grey). Predictably, the resulting graph turns out to be quite complex and hard to read:
Figure 2.1: Freedom of expression in Russia
As we see, apparently the two indices hang together: if one goes up, so tends to do the other. On the other hand, we are probably seeing here the limits of what can be explored in a “static” report like this: we are displaying here four attributes for each each country: absence of corruption in a given year, freedom of expression in the same year, the direction these two indices have taken over a selected period of time and the speed of that change.
3 Two indicators, many countries - towars a “dynamic” graph
What is required instead is to allow the user to explore the data dynamically: We want a slider-bar to change the years and then watch what happens to the dots, i.e. the countries. Creating the underlying graph is relatively straightforward. We also want to have a graph that displays, for a list of selected countries (eventually all countries)
- the population of the country coded as bubble size
- the continent or regime status coded as color
- the x/y position for two indicators.
A prerequisite for this graph is that it must be “fast to calculate” to allow for a good user-experience. We are also pulling population data from the Worldbank database for all years and countries. Unfortunately, that involves manual work to match the country names. In this report, we want to build such a graph and explore its properties. Building it into a dynamic environment will be one of the next steps. Here it is:
Figure 3.1: Freedom of Expression versus Absence of Corruption, 2018
We see how diverse Asia is: yellow bubbles are found all across the graph. However, while it is interesting to see what country ends up where, the color as such doesn’t provide much additional insight. Interestingly, the “State of Democracy” database also tracks other attributes for all countries, namely the “regime type”. When we color-code the countries by that attribute we obtain the following result:
Figure 3.2: Freedom of Expression versus Absence of Corruption, 2018
What we see now is that these two indicators alone are telling apart the type of regime (!). The State of Democracy tracks almost 200 indicators. These two indicators, absence of corruption and freedom of expression, are able to reasonably predict the regime type. Quite visibly, a high freedom of expression and a high absence of corruption is invariably linked to a democratic state. By the same token, a low absence of corruption, i.e. high levels of corruption, and a low freedom of expression are (almost) invariably liked to an authoritarian regime. The sole exception is Turkey, the yellow dot among a “red sea” of authoritarian regimes. We have performed the same exercise for other pairs of indicators, with similar results - something to be explored better.
4 Next steps
I find these results promising. At the same time I notice how number-crunching becomes the bottleneck for insights: only what a data scientist decides to look at is what people get. When I explore my customers’ data, that is fine and in a way it is also wanted: to some extent it is what I get paid for. Here it’s different, though: it is important that many people can explore the data, ask their own questions and draw their own insights:
Analytics capabilities must be “democratized”.
That is not an unusual task and something I have done for a living during the Corona-times: A production manager couldn*t easily go and see how production was doing. She may have a Gigabyte of data per day, arranged in 400-800k rows and some 100+ columns of attributes. Good luck crunching that in Excel, by the way. What the production manager needs is an interactive “dashboard” that allows her to visualize the data and extract insights: What machine is underperforming? Where sits the inventory? And so on.
As for the data here, first I want to build a tool that allows for dynamic, interactive exploration of this rich source for insights. I think I should also perform a principal component and cluster analysis on the data: how many “independent factors” are there in the data? If already the two indices I happened to choose here, and a few more for which I have tried the same, allow telling apart regime types, how many of the indicators are truly independent? I’d also like to understand how we can cluster countries using data. We could THEN relate these clusters to established country groups, such as OECD, BRICS, PIGS or others: are they truly hanging together? Would other groups make more sense? I also have a few more graphs in mind that can help understand what is going on.
All that is quite tactical, though.
5 The bigger picture
I see a broader mission here: We need to “democratize” the analysis of data. When a dataset is available, such as the one put at the disposal of the world on the State of Democracy website, its analysis can’t be left to skilled data scientists. We can’t allow them to be the bottleneck for insights. What needs to be created is a toolkit that allows everyone, including schoolchildren, explore the data and gain insights. Ideally, it is a toolkit users can contribute to and improve over time.
Ping me if you are interested in joining in.