Coordinated by Aurelian GIUGĂL


Spatial polarization at the 2014 Romanian presidential election. A case study on the electoral geography of Bucharest

Giorgian-Ionuţ GUŢOIU

Ph.D. Candidate, Doctoral School of Political Science, University of Bucharest

Abstract: A presidential election was held in Romania on November 2014. The present paper explores the electoral geography of this election in the city of Bucharest. Using spatial statistical modeling our exploratory analysis reveals an important polarization existing at the geographical level. This spatial polarization is highly similar to the polarization that existed at the national political level. These findings suggest that more attention should be given to geographical contexts in understanding Romanian electoral behavior .

Keywords: 2014 Romanian presidential election, electoral geography, Bucharest, polarization.


Apresidential electiontook place inRomania in 2014, and was held in two rounds on November 2 and 16. Out of the fourteen candidates in the first round, the top two candidates qualified for the run-off, namely Klaus Iohannis and Victor Ponta. In the first round, Victor Ponta the leader of the Social Democratic Party and the current Prime Minister won around 40% of the vote while Klaus Iohannis the leader of the Christian Liberal Alliance and the mayor of Sibiu won around 30%. In the second round, Iohannis won the election with 54.5%. The outcome of this election was a surprise since it was the first time a non-Romanian ethnic candidate triumphed. Klaus Iohannis is a German ethnic of Lutheran confession. Also most of the pre-election polls gave Victor Ponta as the winner of the election. Our study proposes an electoral geography analysis of this election in Bucharest - the capital of Romania. By accounting for the spatiality of the vote we shall try to bring new contributions to the comprehension of this election, as well as the understanding of the Romanian electoral behavior in their geographical contexts. Our approach acknowledges the previous contributions in the study of electoral and political geography that had emphasized how the geographical contexts matters for the voter[1]. In this regard contributions were also made for the comprehension of how context matters for the Romanian electoral behavior[2].

 Such an approach understands political parties not just as merely electoral vehicles, although the Anglo-American literature often regards them as such. Instead we see parties as intermediaries between society and state, channeling resources from center to periphery and rewarding territories and territorial interests at the expense of others [3] . Parties are institutions that operate within certain geographical contexts. In the same contexts elections are held and voters make their decisions. For these reasons we expect a relation to exist between the behavior of the political parties and the behavior of voters. This relation is mediated through the geographical context. Our exploratory analysis shall account for the spatial dynamics of electoral behavior. In our analysis we shall also consider the national political context.

Our paper is structured in three sections. Firstly, we shall detail the political context at the time the election was held. Here our discussion equally concentrates on the candidates of interest and their political parties. The candidates and parties we shall refer are listed in the Appendix. In the second section we detail the methodology and the statistical tools we shall employ for our spatial analysis. In the final section, we analyze the spatiality of the electoral outcomes in Bucharest for the candidates listed in the Appendix and for the turnout.


In this section we briefly explain the national political context that preceded the election by placing the political actors within the party system and explaining the stances of candidates in the electoral campaign. Firstly we shall make references to the parties that had candidates in the election, namely the Social Democrat Party, the National Liberal Party, the Democrat Liberal Party, the Popular Movement Party and the Great Romania Party. References to Monica Macovei and Călin Popescu-Tăriceanu who competed as independent candidates shall be made further when we discuss the electoral campaigns.  It is important to mention that discussions regarding the doctrinaire space should be treated with caution. Nevertheless we shall make references to the left, center or right since parties identify with them (at least in their public discourse).

In 2014 the Social Democrats were the largest party in Romania. Since 2012 the party was the main co-partner in the incumbent coalition. The party self-identifies with the center-left, being affiliated to the European Socialists and the Socialist International. PSD resulted from the largest post-revolutionary movement (the National Salvation Front) and is regarded as the main successor party of the former Romanian Communist Party[4] . Since 2010 Victor Ponta is the party leader. Together with the National Liberals and the Conservative Party the Social Democrats formed in February 2011 the Social Liberal Union. The alliance governed until February 2014, when it dissolved and the National Liberals went into opposition. Victor Ponta ran for the Alliance PSD-PUNR-PC. The alliance was formed by three parties with PSD having the largest share of power. The traditional constituencies of PSD are composed by those lesser educated and poorer[5]. There is also an important contradiction between the principles of the Western Left and those of the PSD voters, who are rather conservative, religious, oppose secularism and live in small communities.

Klaus Iohannis was the president of the National Liberal Party since the summer of 2014 and ran as the candidate of the Christian Liberal Alliance, comprising the Democrat Liberal Party and the National Liberal Party. Both PDL and PNL position themselves in the center-right area of the political space. In most of the time this had resulted in a competition between the two parties. In the summer of 2014 both parties were members of the European Popular Party.

Elena Udrea ran as the leader of the Popular Movement Party. The party was formed in 2014 after initially being formed in the spring of 2013 as a political foundation comprising former members of PDL that were dissatisfied with the party leadership. PMP self-identifies with Christian-democracy, liberalism and the center-right and since September 2014 is a member of PPE.

The Great Romania Party was formed in 1991 by Corneliu Vadim Tudor. In 2014 Tudor was still the charismatic leader of the party. The party was stronger in the 90s when it had a salient nationalistic and populist stance. However, after early 2000s the party experienced a continuous decline whilst also it abandoned the nationalistic rhetoric trying to appeal rather as a center-left party.

The election was held in a political context that was highly polarized by the two mandates of Traian Băsescu, who served ten years as President from 2004 through 2014. The existing polarization in the first decade of post-communism had a continuity during the ten years he served as President. Băsescu described himself as a player-president (preşedinte jucător in Romanian), which implied an open-style and hands-on approach towards the state institutions and the general public. This type of behavior was in opposition with a more mediator and withdrawn position. Both Victor Ponta and Klaus Iohannis disavowed this type of approach and both assumed a style based on consensus and dialogue[6] .

Klaus Iohannis was regarded as the candidate from outside the existing political establishment. During the last decade he was the mayor of Sibiu.  His slogan envisioned a Romania of things well done (România lucrului bine făcut - in Romanian). The slogan was again a reference to his profile as a German ethnic, being known the appreciation Romanians have for the German people efficiency. His Romania of thoroughness uncovered a pragmatic liberal-conservative platform. Iohannis stance on the communist past was a clear cut one. He declared himself a fierce anti-communist and expressed his sadness about the fact many Romanians still regret Ceauşescu and live with the nostalgia of the former regime[7] . The stance on the foreign policy was a clear Western one. His profile of anti-establishment was highly attractive among the young voters. His strategy was also directed to appeal this age cohort[8] . The campaign on the social media was also highly effective[9].

Victor Ponta's campaign was dominated by patriotic themes. In this regard he stressed the message proud to be Romanians[10] . The entire electoral campaign of Victor Ponta was based on emotion. Those positive were enhanced by the messages like the ones described above, whilst those negative were directed towards his main opponent. Ponta tried to depict a negative image of Iohannis by associating him with Traian Băsescu.

Călin Popescu-Tăriceanu was a former Prime Minister and President of PNL who had a clear opposition towards PDL and Traian Băsescu. He departed from PNL soon after the party left the ruling coalition in February. Călin Popescu-Tăriceanu led the group within the National Liberal Party that sympathized a partnership with the Social Democrats. After the National Liberals departed from the ruling coalition Tăriceanu resigned from PNL and announced immediately the creation of a new liberal party, named the Liberal Reformist Party (Partidul Liberal Reformator - PRL)[11] . Nevertheless, Tăriceanu ran as an independent candidate. His slogan On your side, welfare and respect (De partea ta, bunăstare şi respect - in Romanian) had a liberal dimension. The message is a reference to the period he served as Prime-Minister, a time before the financial crisis that he considered to be marked by prosperity. Tăriceanu also positioned, as the other two candidates did, in opposition with Traian Băsescu and his conflictual style of politics[12].

The political program of Elena Udrea proposed the continuation of reforms began by Traian Băsescu. The new president had to be active and involved. According to Udrea the President is the most important political actor within the Romanian constitutional structure[13] . It was a direct reference to the legacy of Băsescu.  She conducted a negative campaign against Victor Ponta. The slogans she employed were that of Beautiful Romania (România frumoasă - in Romanian) and Good for Romania (Bună pentru România - in Romanian). It was a type of message that appealed to the low educated. Her message was focused on promises for eradicating poverty[14].

Monica Macovei ran as an independent candidate. She resigned from PDL, since she believed Iohannis was not a fit candidate to run for the right[15] . She employed a liberal, anti-corruption and rule-of-law platform, based on her record as Justice Minister and MEP[16]. Earlier that year, she also stressed her strong pro-Western choice noticing the importance of NATO and EU in the process of reform in Central and Eastern Europe[17]. Ponta negative campaign was also directed towards Macovei. Reference was made by Ponta campaign staff about her Greek-Catholic faith, to which she converted.[18] Macovei stressed a secular platform. This made Ponta's campaign staff to accuse her she was is not fully Romanian, since she fights against the Christian Orthodoxy[19]. Macovei responded and accused Ponta of xenophobia, chauvinism and intolerance [20] . As Iohannis did, she attempted to place herself as an anti-establishment candidate. In her slogan she stressed that she is Better than them. It was a message directed mostly to the young voters with the goal to mobilize this disenchanted social category. Her campaign on the social media was also highly intense and present. Her voters were mostly young urban and well educated[21].

Corneliu Vadim Tudor, with his party strength diminished almost entirely from the previous decades, ran in this election relying only on his core of supporters. With a populist platform of left orientation, he declared the measures he will adopt in case of winning the election: decreeing the state of emergency in economy, the reduction of VAT from 24% to 10% and flat tax from 16% to 10%, as well as the modification of the Constitution[22] .

In the first round, at the national level Ponta scored 40% of the votes, while Iohannis gained 30%. In many instances, this difference between them two was considered insurmountable. The second round was scheduled for 16th of November. All the major candidates that failed to enter the second round expressed their support to one of the two candidates. Călin Popescu-Tăriceanu, Teodor Meleşcanu, Vadim Tudor and Dan Diaconescu all expressed their support for Victor Ponta. Yet Iohannis was not concerned to negotiate a transfer of electoral capital from the other candidates. However Monica Macovei[23] and Elena Udrea[24] publicly announced their support for Iohannis.

To win the election Iohannis needed an increase in turnout. Iohannis campaign for the second round was directed towards this end. The theme employed was that of the practices of Ponta's cabinet to inhibit the exercise of voting of those categories that would support Iohannis. These categories were the voters from the Diaspora, who traditionally voted against the Social Democrats. In the first round, Iohannis scored 46 % of the votes in the Diaspora, while Ponta gained only 16%.

During the Election Day, mass-media showed images from the polling sections abroad where people were standing in line to vote[25] . These have created an immense wave of emotion and sympathy with those standing in line, especially since the state officials refused to prolong the schedule in order for everyone to vote. These images from abroad found in social media the place where they would become an important catalyst for the negative attitudes against Victor Ponta. A full-blown domino effect was in place.

 The result of these two weeks of intense campaigning was the complete change of results in comparison with the first round. A change of 20% of votes occurred in favor of Klaus Iohannis who won the election with almost 10% in front of Victor Ponta. As we have depicted, the political climate at this election was highly polarized.


Elections are social phenomena that take place in contexts, which are continuously modified and shaped by particular geographical phenomena. To give evidence of these contexts in our thesis we employ spatial econometrics. We believe that spatial analysis through the use of the proper conceptualizations, operationalization and statistical tools can create the linkage between quantitative electoral geography and the social construction of space. The main argument that underpins this assumption can be formally explained through the interdependency of places as expressed in Tobler's first law of geography: Everything is related to everything else, but near things are more related than distant things[26] . Such an assumption views the individual observations as being correlated with each other on the basis of their spatial proximity.

Such spatial biases can result in similar values concentrated within adjacent territorial units at local scales, whilst at larger scales it can outcome in larger patterns and regional variations. Recognizing and assessing for such spatial biases is commonly referred as spatial dependence (also the term spatial autocorrelation is used interchangeable) at local scales and spatial heterogeneity for assessments at larger scales. The importance of these two geographical phenomena for elections was empirically tested in a collective work by John O'Loughlin, Colin Flint and Luc Anselin[27] . In our study we shall account only for spatial dependence.

Luc Anselin defines spatial dependence as follows: in general terms, spatial dependence can be considered to be the existence of a functional relationship between what happens at one point in space and what happens elsewhere[28] .Spatial dependence elucidates 'Galton's problem' of making statistical inferences when the observations are not statistically independent, since certain traits in an area are often cause not by the same factor operating independently in each area by diffusion processes[29].Spatial dependence occurs through two main phenomena: contagion and diffusion. As noted by political geographers Michael Shin and John Agnew, these two processes may capture local information, social networks or geographic organization of a political party [30] .

In order to use such spatial statistical models we employ a data format for geographic information system (GIS). We use the shapefile format where each location is displayed as a point, which describes vector features and has attached attributes that describe it: the geographical coordinates, the ID number, address, the sector, electoral records, etc. We conduct our spatial analysis using GeoDa, a free software package, developed by Luc Anselin and his team of researchers[31] .The software conducts spatial data analysis geovisualization, spatial autocorrelation and spatial modeling.

Depending on the scale at which we search for spatial autocorrelation the models we employ can be grouped as Global Indicators of Spatial Autocorrelation (GISA) and Local Indicators of Spatial Association (LISA)[32] . GISA are a form of measuring the overall geographic clustering of the data whilst LISA identify the actual clusters. To conduct our analysis at both scales we use the Moran's statistic as well as models derived from it.

To evaluate this global clustering we shall employ Moran's I statistic. We shall compute the Moran's I with the help of GeoDa software[33] . The spatial analysis at the global level outputs a statistic (Moran's I) that summarizes the entire studied area. The statistic takes values ranging from -1 indicating perfect dispersion to +1 indicating perfect clustering or correlation, whilst a zero value indicates a random spatial pattern. We also employ a test of the null hypothesis. The null hypothesis implies that the values of the variable are randomly distributed in space, meaning that we cannot predict the values of the neighboring observations having at base the knowledge of value for the center location that is neighbored. (we describe below how we perform this test).

A negative value of I and with the null hypothesis of no spatial autocorrelation rejected indicates a negative spatial autocorrelation meaning the clustering of dissimilar values among neighboring observations, while a positive value of I and with the null hypothesis of no spatial autocorrelation rejected indicates positive spatial autocorrelation. In our case, negative spatial autocorrelation appears when neighboring voters support different candidates while for the positive autocorrelation neighboring voters support the same candidate.

Moran's I is expressed as the linear relation between vectors of observed values, i.e. y, and the weighted average of the values that neighbors y. The weighted average values of y in location i are commonly known as the spatial lag of y in location i. GeoDa also uses this notation.

The formula for Moran's I is:


For our case study N is the number of observations and equals 275, ∑0 is the sum of all elements in the spatial weights matrix, y represents the vector of observations, while Wy is the spatial lag of y.

A matrix of weights is required since evaluating spatial autocorrelation imposes a framework of spatial interactions between the items and their values. The structure of the matrix imposes a restriction on the number of neighbors that can be accounted for. The spatial weight matrix is constructed as being based on the relation of vicinity between locations. We define distance as being based on Euclidean measurement. To construct the spatial lag of y we account for the four closest locations that neighbor the central one.

GeoDa provides a simulation test for the significance assessment of the I statistic[34] . We use this to test the null hypothesis of no spatial autocorrelation. This assessment is based on randomization through a number of permutations. The software limits the permutations up to 99,999 (the lower limit is 99). These are used to evaluate the possibility to observe a Moran's I value under the situation of spatial randomness. The permutations implies assigning to each observation a vector of randomly generated numbers, which are used to randomly reallocate each observation in space and each randomization computes the statistic with different sets of numbers. We use the notation "Z" for the randomization.

The output provided by GeoDa gives the Moran's statistic together with a scatter plot, namely Anselin's Moran scatter plot. This is a LISA and is designed to visualize the type and strength of the spatial autocorrelation since the slope of the regression line corresponds to the value of the Moran's I. On the x-axis of the scatter plot is displayed the standardized value of the variable while on the y-axis is deployed the standardized spatially lag of the same variable. In the output generated by GeoDa the scatter plot is sectioned in four quadrants each indicating four types of spatial association:

              Quadrant I: high values of y surrounded by high values;

              Quadrant II: low values of y surrounded by high values;

              Quadrant III: low values of y surrounded by low values;

              Quadrant IV: high values of y surrounded by low values.

The interpretation guide for the scatter plot is displayed in Figure 1. Electoral loyalties, strongholds, congregations of supporters of particular parties or mobilized voters can be identified with the help of the first quadrant. Since the values depicted in the quadrants I and III are the ones responsible for the spatial dependence in the dataset, these two can be further modeled for cross comparisons between variables. We employ such comparison in order to identify how votes have shifted between candidates in the same places or how different candidates have coexisted in the same milieu.

Figure 1. Anselin’s Moran scatter plot interpretation guide


Before the actual spatial analysis we shall interpret the descriptive statistics in Bucharest and at the national level. These descriptive statistics are displayed in Table 1 and can be further used as benchmark for the spatial analysis. The aggregated results at the both national and city level show different winners for the two rounds. In the first round the winner was Victor Ponta with a share of 31.78% votes in Bucharest and 40.44% at the national level, while in the second run Klaus Iohannis gained at the national level 54.43% and 57.3% in Bucharest. In Bucharest, for the first round the votes for Ponta and Iohannis account for two thirds of the valid votes. This weight is even greater at the national level. This ratio of votes refer to an important polarization existing at both levels between Victor Ponta and Klaus Iohannis.

Table 1 . Descriptive statistics of the electoral outcomes

No other candidate managed to gain neither at the national level nor in Bucharest at least 15% of the votes. In Bucharest the third ranked candidate was Monica Macovei. In comparison with the first round, in the second we see a major increase in turnout. For the both scales presented in Table 1 the turnout increases with approximately ten percentages between the two rounds. This increase depicts an important polarization of the political space determined by the competition between Ponta and Iohannis. For this reason understanding the turnout becomes important in explaining the polarization at this election.

Further information regarding the electoral polarization between Ponta and Iohannis we can find from Table 2. Here we see listed the candidates who gained the most votes for every location of the 275 ones. For both rounds the winner were either Ponta or Iohannis. In the first round Victor Ponta gained more votes than Iohannis in 168 locations, whilst Iohannis in 106 locations. A tie was also recorded in the first round. For the second round the results are completely different. Ponta achieved higher scores than Iohannis in only 7% of the polling locations. The second run meant a landslide victory for Iohannis throughout Bucharest. It is important to note that our spatial analysis does not account for the variations that may exist between the municipal administrative subdivisions named "sector".

Table 2. The Candidates with most votes per location

We continue our exploration by using the Moran’s I statistic and the models we derive from it. The statistics for Moran’s I are presented in Table 3. The Z scores confirm the existence of spatial autocorrelation for all variables of interest. In all cases this spatial autocorrelation is positive and varies from a medium to high intensity.

Table 3. Moran's I for the candidates and the turnout

For both rounds, Klaus Iohannis and Victor Ponta had similar Moran's I, namely 0.67 (first round) and 0.62 (second round). We presume this situation is the result of having two electoral geographies constructed in opposition one with the other. These statistics indicate a high spatial autocorrelation pertaining to electoral geographies built upon important spatial patterns and clustering. The following exploratory analysis will confirm whether the territories where Klaus Iohannis is dominant are the same territories where Victor Ponta recorded his weak performances and vice versa. Also for both Macovei and Tudor the spatial autocorrelation is at a high intensity. For Macovei the Moran’s I is 0.66, whilst for Vadim Tudor the statistic is 0.59. The candidate with the lowest value is Elena Udrea, namely 0.25. For the president of PMP the Moran's I indicates a rather low to medium spatial autocorrelation. For Tăriceanu the Moran's I is 0.40, which pertains to a geography with more clusters than the one of Udrea. Udrea appealed to the highest number of social strata than any other candidate did. For the turnout the value of the statistic is 0.43 for the first round and decreases to 0.33 for the second round.

These values of Moran's I captured the existence of important geographical clustering and spatial patterns within the city. At this presidential election the city was divided in its electoral sympathies. Further we explore these spatial patterns at a more local scale. In this regard we shall employ the Moran's scatter plot. Firstly the scatter plots will help us identify the direction of these spatial patterns and see for each candidate if there are positive or negative clusters that dominate the geography or it exists a consistency within the spatial distribution. Secondly, we shall make comparisons between scatter plots in order to explore similarities or dissimilarities between the spatial patterns for different candidates. The Moran's scatter plots are displayed in Figure 2 and Figure 3.

Figure 2. Scatter plots for the First Round

Figure 3. Scatter plots for the Second Round

As we mentioned above, the first and third quadrants include the asocciations of similar values. For the first round, except for the scatter plots for Iohannis and Ponta we find for the other scatter plots a inconsistency between the first and the third quadrants. For Macovei and Călin Popescu-Tăriceanu prevalent are the territories with positive clusters. It is in the scatter plot for Macovei that we find the highest number of observations located in the first quadrant, namely 124. This distribution recognizes Macovei as a candidate with well-defined rhetoric and image, making her appealing to particular social strata. Inconsistencies between the first and third quadrants we also find at Elena Udrea and Vadim Tudor. However, these two candidates have higher shares of observation with negative values. Elena Udrea is the candidate with the lowest number of observations located in the first quadrant, namely 84, whilst the third quadrant depicts 99. This distribution pertains to the type of campaign Udrea promoted as she tried to appeal indistinctively in as many social and geographical milieu as possible. For Vadim Tudor the distribution of observations in quadrants I and III is specific to a niche candidate. Vadim Tudor is the candidate with the highest number of observations located in the third quadrant, namely 124. This indicates Tudor as a candidate that was particular unattractive for many categories of voters.

The turnout also presents an inconsistency between the two quadrants. Its number of observations located in the third quadrant is higher than the one in the first quadrant. The difference between the two quadrants is 15 (first round) and 21 (second round). However, the difference of ratio between the two rounds is low. This indicates the existence of a temporal stability in what regards the geographical structure of voting.

For the first round Klaus Iohannis and Victor Ponta are the candidates with the most consistent relation between the two quadrants responsible for spatial autocorrelation. For the two scatter plots the differences between the two quadrants are low. However, the same stability is not present for the second round since the scatter plots are constructed in opposition. The decrease in consistency between the positive and negative clusters we explain through the polarization of political space generated by a confrontation between two candidates. However, for each candidate a clear stability exists between the two rounds. This implies the existence of well-defined spatial patterns for each candidate. As the political space becomes more polarized these patterns also become more salient.

The above exploration has helped us identify the existence of spatial patterns within the electoral geographies for the candidates and the turnout. Further, we shall compare the scatter plots and try to identify the extent to which these spatial patterns overlap or oppose. For this we shall quantify the number of shared observations located in the first and third quadrants between the scatter plots. This last section of our exploratory analysis can confirm the existence of geographical polarization.

Table 4 depicts a crossing between the first and the third quadrants of the scatter plots discussed above.  The percentages displayed in the table are computed from the number of the observations located in the scatter plots listed on the rows. These figures represent an indicator of the weight spatial patterns of the geographies listed on the columns have within the maps displayed on the rows. For the scope of the present paper we chose to present the overall conclusion we can draw from this table instead of detailing each cell. As we see from Table 4 in the first round we can notice how spatial patterns oppose or overlap depending on the existing relations within the political space. Klaus Iohannis and Monica Macovei have almost entirely similar patterns. These are constructed in opposition with the spatial patterns of Victor Ponta and Vadim Tudor. Both Ponta and Tudor display electoral maps in complete opposition with the geography of the turnout. On other side, two-thirds of the spatial patterns of both Iohannis and Macovei overlap the existing patterns of the turnout. Here we identified a polarization with two given poles: one that comprises the geographies of Iohannis and Macovei and another one that comprises the geographies of Ponta and Tudor. For each pole we can give an ideological label, namely a liberal one for Iohannis and Macovei and a center-left one for Ponta and Tudor.

Table 4. Cross comparison between scatter plots for the 1st and 3rd Quadrants in the First Round


Adjacent to the midpoint of this polarization are the geographical patterns of Călin Popescu-Tăriceanu and Elena Udrea. However, for each of these two candidates there are certain patterns in their electoral map that account for a position closer to a certain pole. Hence Tăriceanu displays an electoral geography that shares more than two-thirds of its clusters with the maps for Iohannis, Macovei and the turnout whilst with maps of the other pole Tăriceanu shares aproximately a tenth of his clusters. If Tăriceanu is closer to the liberal pole, however the geography of Udrea places her rather closer to the pole of the center-left candidates. Her geography is surprisingly given the center-right position of her party. The structure of her spatial dependence is much similar to the one displayed by Ponta and Vadim. Two-thirds of the clusters in Elena Udrea's map are also to be found within the maps of Vadim and Ponta. Nevertheless for Udrea there are still some similarities with the spatial dependence displayed by the maps for Iohannis and Macovei. For example, 15% of the observations depicted in the first quadrant and 20% of the observations depicted in the third quadrant of her scatter plot are shared with the scatter plot for Iohannis.

Table 4 has shown us the existence of an important polarization in the first round of the election. This polarization indicates the existence of two poles with two structures of geographical patterns with each of this structure in opposition with the other. One pole is represented by the spatial patterns of Klaus Iohannis, Monica Macovei. At the other pole there are the spatial patterns of candidates who self-describe themselves as center-left, namely Victor Ponta and Corneliu Vadim Tudor. For each pole both maps hold almost a complete consistency of spatial dependence. The spatial association of the turnout is constructed in opposition with the one for the candidates of the center-left, whilst also being closer to the liberal pole of Iohannis and Macovei, although it doesn't overlap with the spatial patterns of these two. Meanwhile, neither Călin Popescu-Tăriceanu nor Elena Udrea identify with the spatial patterns of solely one pole, although their propensity towards a particular pole is evident for each. We can associate Tăriceanu with the liberal pole of Iohannis and Macovei, although in his electoral map there are also areas that could be associated with the geography of Ponta or Vadim Tudor. The electoral geography of Elena Udrea is closer to the one of the center-left candidates, albeit her party being of a center-right orientation. This situation is explained by the catch-all rhetoric Udrea displayed at this election. Nevertheless, there are also some similarities for Udrea with the geographies of Iohannis, Macovei and the turnout.

We conclude our exploration with an analysis of the spatial and temporal dynamics established between the geographies of the two rounds. In Table 5 we can observe the replacements and realignments within the spatial context between the candidates in the first round and the candidates in the second round. Here the geographical dynamics formed between candidates are similar to the ones we described for the first round. Both Iohannis and Ponta are consistent with their performance in the first round and both manage to preserve the loyalties with the geographies associated with their poles: namely Iohannis with his territories, the ones of Macovei and the turnout and Ponta with his territories and the ones of Vadim Tudor. Iohannis shares approximately 60% clusters of his clusters with Tăriceanu, whilst Ponta shares approximately a half of his clusters with Elena Udrea. However, in the second round for Iohannis the weight of the clusters shared with the turnout decreases in comparison with the first round from 72% (Q 1) and 88% (Q 3) to 64% (Q1) and  88% (Q 3). The conclusions we draw from Table 5 serve to confirm the persistence of geographical patterns and polarization.

Table 6. Cross comparison between scatter plots for the 1st and 3rd Quadrants in the First Round and Second Round


In this paper we have used spatial models to account for the electoral behavior in the city of Bucharest at the 2014 presidential election. Our exploratory analysis has revealed the presence of an important polarization existing at the geographical level. We found that this polarization is constructed as having two main poles. The weak areas for the candidates at one pole are the same with the strong areas for the candidates at the other pole. At one pole there are the geographies we defined as being 'liberal' or 'center-right', namely the ones of Klaus Iohannis, Monica Macovei and the turnout, whilst at the other pole the geographies we defined as being of 'center-left', namely those of Victor Ponta and Corneliu Vadim Tudor. In between these two poles there are the geographies of Călin Popescu-Tăriceanu (closer to the center-right pole) and of Elena Udrea (closer to the center-left pole).

Our findings suggest that these two poles are determined by the actual structuration of the political system. How candidates and parties place themselves within the party system also influences their actual electoral geography. This geographical polarization depicts dynamics that are similar to the ones displayed by the political context at the national level. We suggested that this relation is in many stances determined by the existence of political parties as mediators between state and society. Further research should identify the variables that are responsible for the existence of a correspondence between the political space and the electoral geographical space.

Appendix. The candidates of interest


AGNEW, John, Mapping politics: How context counts in political geography, Political Geography, Vol. 15, No. 2, 1996, pp. 129-146.

AGNEW, John and SHIN, Michael, The geography of party replacement in Italy, 1987–1996, Political Geography, Vol. 21, 2002, pp. 221-242/ p. 226.

ANSELIN, Luc, Spatial Econometrics: Methods and Models, Springer Science & Business Media, 1988.

ANSELIN, Luc, Local indicators of spatial association - LISA, Geographical Analysis, Vol. 27, 1995, pp. 93-115.

COVACI, Mihai, Factorul Facebook în alegerile prezidenţiale din 2014, Sfera Politicii, No. 183, 2015, pp. 85-91.

GIUGĂL, Aurelian, Geografia electorală a Dobrogei postcomuniste: 1992-2012, Editura Fundaţiei pentru Studii Europene, Cluj-Napoca, 2013.

GUŢOIU, Giorgian, Realinieri partizane româneşti în context geografic. O explorare a geografiei procesului de substituţie între partide în Transilvania, 1996-2008, Studia Politica. Romanian Political Science Review, Vol. 14, No. 4, 2014, pp. 515-538.

JOHNSTON, R.J. and PATTIE, Charles, Putting Voters in Their Place: Geography And Elections in Great Britain, Oxford University Press, Oxford, 2006.

LEIB, Jonathan and WARF, Barney (Eds.), Revitalizing Electoral Geography, Ashgate, 2012.

O'LOUGHLIN, John, FLINT, Colin and ANSELIN, Luc, The Geography of the Nazi Vote: Context, Confession and Class in the Reichstag Election of 1930, Annals of the Association of American Geographers, Vol. 84, No. 3, 1994, pp. 351-380.

MATEI, Silviu, Romania at voting age, Lambert Publishing, 2011.

MITULESCU, Sorin, Evoluţii în electoratul românesc: de la FSN la PSD şi PD-L, Sfera Politicii, Vol. 17,  No. 40, 2009, pp. 53-60.

POP-ELECHEŞ, Grigore, Separated at birth or separated by birth? The communist successor parties in Romania and Hungary, East European Politics and Societies, Vol.13, No. 1, 1999, pp. 117-147.

POP-ELECHEŞ, Grigore, A Party for All Seasons: Electoral Adaptation of Romanian Communist Successor Parties, Communist and Post-Communist Studies, Vol. 41, No. 4, 2008, pp. 465-479.

SHIN, Michael and AGNEW, John, Berlusconi's Italy: Mapping Contemporary Italian Politics, Temple University Press, 2008.

TOBLER, Waldo, A computer movie simulating urban growth in the Detroit region, Economic geography, Vol. 46, No. 2, 1970, pp. 234-240.

TUFIŞ, Claudiu, Ataşamentul faţă de partidele politice în România, in COMŞA, GHEORGHIŢĂ and TUFIŞ (Eds.), Alegerile pentru Parlamentul European: România, 2009, Polirom, Iaşi, 2010, pp. 19-54.

Online resources;;;;;;;;;;;;;;;;;;;


[1] John AGNEW, Mapping politics: how Context Counts in Electoral Geography, Political Geography, Vol. 15, No. 2, pp. 129-146; R.J. JOHNSTON and Charles PATTIE, Putting Voters in Their Place: Geography And Elections in Great Britain, Oxford University Press, Oxford, 2006, pp. 46-54; Jonathan LEIB and Barney WARF (Eds.), Revitalizing Electoral Geography, Ashgate Publishing, 2012.

[2] Aurelian GIUGĂL, Geografia electorală a Dobrogei postcomuniste: 1992-2012, Editura Fundaţiei pentru Studii Europene, Cluj-Napoca, 2013; Giorgian GUŢOIU, Realinieri partizane româneşti în context geografic. O explorare a geografiei procesului de substituţie între partide în Transilvania, 1996-2008, Studia Politica. Romanian Political Science Review, Vol. 14, no. 4, 2014, pp. 515-538; Silviu MATEI, Romania at voting age, Lambert Publishing, 2011; Claudiu TUFIŞ, Ataşamentul faţă de partidele politice în România, in Mircea COMŞA, Andrei GHEORGHIŢĂ and Claudiu TUFIŞ (Eds.), Alegerile pentru Parlamentul European: România, 2009, Polirom, Iaşi, 2010, pp. 19-54.

[3] Michael SHIN and John AGNEW, Berlusconi's Italy. Mapping contemporary Italian politics, Temple University Press, 2008., p.52.

[4] Grigore POP-ELECHEŞ, Separated at birth or separated by birth? The communist successor parties in Romania and Hungary, East European Politics and Societies, Vol. 13, No.1, 1999, pp. 117-147; Idem,  A Party for All Seasons: Electoral Adaptation of Romanian Communist Successor Parties, Communist and Post-Communist Studies, Vol. 49, No. 4 , 2008, pp. 465-479.

[5] Sorin MITULESCU, Evoluţii în electoratul românesc: de la FSN la PSD şi PD-L, Sfera Politicii, Vol. 17, No. 40, 2009, pp. 53-60.

[6][], accessed at 08.06.2015.

[7] Klaus IOHANNIS, Pas cu pas, Curtea Veche, Bucharest, 2014, p. 23.

[8]See in this regard his speech held at the candidature launching on 27th of September [], accessed at 09.06.2015.

[9] Mihail COVACI, Factorul Facebook în alegerile prezidenţiale din 2014, Sfera Politicii, No. 183, 2015, pp. 85-91.

[10][], accessed at 10.06.2015.

[11] [], accessed at 08.06.2015.

[12] Ibidem, p.7.

[13] The political program of Elena Udrea, available at [], accessed at 11.06.2015. p. 3.

[14] [], accessed at 11.06.2015.

[15] [], accessed at 11. 06.2015.

[16] [], accessed at 11.06.2015.

[17] [], accessed at 11.06.2015.

[18] [], accessed at 11.06.2015.

[19] Ibidem.

[20] [], accessed at 11.06.2015.

[21] [], accessed at 11.06.2015.

[22] [], accessed at 11.06.2015.

[23] [], accessed at 11.06.2015.

[24] [], accessed at 11.06.2015.

[25][], accessed at 11.06.2015.

[26] Waldo TOBLER, A computer movie simulating urban growth in the Detroit region, Economic geography, vol. 46, no. 2, 1970, pp. 234-240.

[27] John O'LOUGHLIN, Colin FLINT and Luc ANSELIN, The Geography of the Nazi Vote: Context, Confession and Class in the Reichstag Election of 1930, Annals of the Association of American Geographers, Vol. 84, No. 3, 1994, pp. 351-380..

[28] Luc ANSELIN, Spatial Econometrics: Methods and Models, Springer Science & Business Media, 1988, p. 11.

[29] John O’LOUGHLIN and Luc ANSELIN, Geography of international conflict and cooperation: theory and methods, in M.D. WARD (Ed.), The New Geopolitics, Gordon and Breach, London, 1992, pp. 11-38/ p. 17.

[30] John AGNEW, Michael SHIN, The geography of party replacement in Italy, 1987–1996, Political Geography, Vol. 21, 2002, p. 226.

[31] GeoDa is available for free download at the software's site:, accessed at 25.05.2015.

[32] Luc ANSELIN, Local indicators of spatial association - LISA, Geographical Analysis, Vol. 27, 1995, pp. 93-115.

[33] Steps for computing Moran's I with GeoDa at [], accessed at 27.05.2015.

[34] Steps for performing a significance assessment at [], accessed at 25.05.2015.