War, unemployment and education: How they affect elections?

In our election forecasting model we combined election polls, previous electoral results and trends, and, last but not least, a number of socio-economic parameters that we uncovered to be important determinants of voter preferences. In our earlier text we already summarized how we did this and which variables we decided to put in:

“…we used data on local level unemployment, exposure of the community to the 1991-1995 war for independence, and the educational structure of voters in each electoral district (these three factors carry the greatest weight in predicting voting patterns of domestic voters).”

In today’s text we will briefly show how each of the stipulated socio-economic variables is correlated with previous electoral results, and how they could serve as a partial predictor of future votes for a given party.

Which is more important: the economy or the war? 

Much of our intuition on the variables we decided to use came from a paper by Glaurdic and Vukovic (2015) “Voting after War: Voting after war: Economic performance and legacy of conflict as determinants of electoral support in Croatia”. The hypothesis was to test to which extent are the economy and the legacy of the war important for electoral patterns of Croatian voters. The assumption was that the effect of the country’s war-stricken past will fade, while the impact of economic conditions (such as unemployment and income) will become stronger.

„While the findings suggest voters do respond to parties’ economic policies, the underlying pattern of electoral support demonstrates that competition is heavily constrained by the legacy of conflict, with the communities more exposed to the violence being more likely to vote for the principal party of the center-right which led the country into independence and throughout the war. This tendency exhibits a remarkable level of stability over time…“

So the experience of war is the most important determinant of voting patterns of Croatian voters. In communities more exposed to violence voters tend to vote for the center-right HDZ, while in communities less exposed (or without any exposure) to violence, tend to vote for the center-left SDP [1]. This effect is quite stable over time and shows no signs of abating. The reasons for this are many, although the most important one is the fact that the topics related to the war (Hague Tribunal processes on war crimes, the conditionality of EU entry on extradition of Croatian generals to the Tribunal, the issue of veteran pensions and benefits, the issue of fake veterans, and most recently the issue of introducing the Cyrillic script on public buildings in the city that suffered the most during the war), kept being reintroduced by the two dominant political parties, both riding the ideological wave of their voters. While the HDZ was successfully using the war topic and the veterans’ benefits issue to galvanize its own support in war-affected areas, SDP was using the issue of fake veterans and their benefits to raise support amongst their own voters. The strategies worked, and have even proved decisive in some elections.


Unemployment (%) per Croatian municipalities.


Vote share (%) for SDP candidate Josipovic in the first round of the 2014 Presidential elections


Vote share (%) for HDZ candidate Grabar Kitarovic in the first round of the 2014 Presidential elections

However the economy still is important. Municipalities with high levels of unemployment and low levels of average per capita income tend to show higher support for the center-right HDZ, while municipalities with low unemployment and higher incomes p/c tend to favor the center-left SDP (see the maps). This means that the standard left-right classification in Croatia is completely the opposite of the one in the West. In Croatia on average richer communities vote for the left parties, while on average poorer communities vote for the right parties. However, Croatia is hardly unique in this case. In fact, in the entire post-communist group of countries the political left and right are not necessarily the economic left and right. In almost every Eastern European country the political right is more socially-oriented, while the political left is more economically liberal (although not nearly as much as in the West). The distinction is purely ideological.

The reasons for this are also abundant, where the dominant explanation revolves around the socialist origin of these countries combined with the specific conditions within which a particular party found itself in while in power. Left parties, originating primarily from the former communist parties, had more to prove to their voters so they had a higher motivation for more liberal economic policies, including austerity measures in times of crises. The right parties, usually newly formed after communism, even though they started the privatization processes, showed a greater tendency to expand the welfare state (partially to offset the negative consequences of privatization-led layoffs and to protect certain voter groups, thus effectively buying greater voter support), and were characterized by, on average, larger government spending and greater deficits.

Linking war, the economy, and education

In municipalities highly exposed to the war there is a significant positive correlation between votes for HDZ and higher unemployment, and a negative correlation for incomes p/c. Many areas that suffered during the war (the eastern and southern parts of the country – particularly areas closer to the borders with Serbia and Bosnia and Herzegovina), have failed to even remotely recover from the war destruction. The reason is simple, the devastation of factories closed many jobs, the labor force was crippled with depressed wages, emigration (internal and external), and new welfare claimants (many war veterans, not being able to find work after the war, became retired and took veteran benefits), which all lead to lower tax incomes and an inability to achieve fiscal sustainability for the local governments.

Furthermore, in addition to having low wages and high unemployment, the population of war-stricken cities and municipalities is relatively less educated compared to the Croatian average. This means that in these areas the lack of educated and higher quality human capital is further deepening their relative stagnation and disabling any chance of breaking the vicious cycle of underdevelopment and welfare dependency any time soon. The direct consequence is mass emigration abroad.

It is therefore not surprising to notice the same relationship between education (and the lack thereof) and vote shares of the two dominant parties. The level of education (measured as total years of schooling) is our third most important predictor of electoral results. On average in areas with higher levels of education (more years of schooling) we observe more votes for the center-left SDP. And the opposite; in areas where the voting population on average has only a finished primary school or less, HDZ experiences greater electoral support (see figures below).

education and votes

The upper red graphs show the relationship between vote share for SDP (y-axis) and the % of people with finished primary school (first graph), high school (second graph) and higher education (third graph). The lower blue graphs show the same thing for HDZ. The figures show the trend on a countrywide basis, however there exists a considerable variation within electoral districts.

This, however, does not imply a causal relationship. It doesn’t mean that highly educated voters vote for SDP while the uneducated vote for HDZ. It simply points to a general trend on the level of cities and municipalities. We should therefore observe the levels of education jointly with all other socio-economic factors determining voter preferences, and based on this draw inferences on support for each party.

There is a strong correlation between incomes p/c and years of education in Croatian municipalities. Incomes in areas where the average years of schooling is 12 (meaning that the average voter has at least finished high school) are 4 times higher than incomes in areas with areas where the majority of the population doesn’t even have a primary school education. The similar correlation is for education and unemployment – lower levels of education are correlated with high unemployment. It is not therefore surprising to observe that less educated areas, which are also poorer, show a greater tendency to vote for the HDZ.

The interaction between the lack of education and consequential economic outcomes of low incomes and high unemployment is conditioning voter preferences in underdeveloped areas. Adding to this the relative exposure of the war, which very likely also affected the poor economic conditions and emigration of the more highly educated from underdeveloped areas, we get a complete picture of the drivers of voting preferences of domestic voters. Undeveloped areas, more exposed to war which partially generated the welfare dependency and a consequential vicious cycle of low incomes, high unemployment and the lack of human capital, express, on average, more support for the HDZ. The reverse is true for the SDP – areas with lower unemployment, higher incomes, relatively unaffected by the war, and with better human capital, express entirely different voter preferences. These findings enabled us to consider the entire picture behind voter choice formation in Croatia, and were thus crucial parts of our prediction model.

[1] The variable used as a proxy of war exposure was the total number of disabled persons (military and civilian) whose disability was caused by the war, per 1000 inhabitants. This variable almost perfectly paints the picture of where the war exposure was greater and where it was unaffected.


Total over-19 population (%) with primary school degree or less (red areas indicate higher percentage of low-educated)


Total over-19 population (%) with high school degree (red areas indicate higher percentage of better educated)


Total over-19 population (%) with higher education degree (university and other) (red areas indicate higher percentage of higher educated)

Share This