THE IMPACT OF ETHNIC HETEROGENEITY ON DECENTRALIZATION PERCEPTIONS IN ROMANIA

Abstract While most research on decentralization has focused mainly on the fiscal implications, there is also a need to assess the perceptions of citizens towards the process, perceptions which are likely to differ between different ethnic, religious or cultural groups. Although Romania is one of the largest European countries, research in this area has been very limited. The aim of this paper is to provide preliminary insights into the relationship between ethnic heterogeneity and perceptions of decentralization. Using an experimental research design, the paper examines support for decentralization in Bucharest, the mostly homogenous capital city located in Wallachia, and in Cluj-Napoca, the largest city in Transylvania with a relatively ethnically heterogeneous population. Results suggest that heterogeneous populations seem to be more sensitive to the rhetoric about decentralization.


Introduction
Over the past few decades, decentralization reforms have risen to prominence due to widespread belief that the one-size-fits-all policies imposed by central governments often fall short of improving conditions at the local level.In developing countries international organizations have often promoted it as a solution to increase the level of accountability and efficiency (Rondinelli, 1989).The varied benefits are attributed to the fact that, in contrast to standardized national policies, local governments understand better the local problems and can design policies that address local needs more quickly (Oates, 1972;Ahmad et al., 2005).In addition to attracting interest from practitioners, decentralization reforms have also generated considerable academic interest.However, the bulk of research has focused mostly on the fiscal implications.Nonetheless, as Diaz-Serrano and Rodríguez-Pose (2011) mention, the main objective of decentralization is to improve the delivery of public goods and services and respond quicker and better to citizen's needs; thus, research should also focus on analyzing the citizens' perceptions towards the reform.In this sense, ethnic heterogeneity has proven to be relevant to understanding the impacts of decentralization reforms, and several scholars have found a statistically significant relationship between the two.
Surprisingly there have been very few attempts to gauge the impact of decentralization in Romania, although this concept has been constantly discussed at the political level.This country constitutes a particularly interesting case not only because of its political importance, as the second largest European Union country in Central and Eastern Europe, but also due to its ethno-demographic distribution.While regions of Wallachia and Moldova have mostly homogenous populations, Transylvania, due to its history, is ethnically, culturally and religiously heterogeneous.
The aim of this paper is to address this important research gap by providing preliminary insight into how ethnic heterogeneity may influence citizens' perception towards decentralization using a survey experiment applied in the two largest Romanian cities which differ in terms of ethnic heterogeneity.The decision of using the experimental design is based on the fact that its results tend to resemble real world dynamics (Barabas and Jerit, 2010).The findings from this study can contribute to the continuous debate regarding the topic by demonstrating contextual factors that may catalyze or hinder decentralization.

Theoretical aspects
An increasing number of countries are decentralizing the administrative, fiscal, and political functions of the central government to lower-level governments in what has become one of the most important forms of contemporary policy reform and institutional redesign in the last several decades (Azfar et al., 1999;Ahmad et al., 2005).Decentralization can be administrative and/ or fiscal and/ or political and it can be done through deconcentration, delegation, devolution or privatization.The services that are most often decentralized are education, health, water, sanitation, public transport and infrastructure, road maintenance, fire protection, housing and social welfare.Macroeconomics, foreign affairs, defense and regional or national projects that generate economies of scale usually remain centralized (Ranis, 2012).

Decentralization and ethnic heterogeneity
According to the classic literature, decentralization should be implemented when preferences and demands, including the ones related to public policies, are heterogeneous across jurisdictions (Tiebout, 1956;Seabright, 1996;Bardhan and Mookherjee, 2005;Wibbels, 2006).In this regard, more heterogeneous states should be inclined to decentralize but when heterogeneity is absent, a centralized provision of goods and services for all jurisdictions tends to be more efficient (Oates, 1972).Ethnic, cultural or religious groups that are characterized by a different history and which have different needs and preferences from the rest of the country can be marginalized in a centralized system.Therefore, especially in multiethnic countries, decentralization can legitimize certain groups and satisfy some of their demands for self-governance (De Mello, 2000;Tranchant, 2010) and, thus, reduce potential ethnic conflicts (Brancati, 2006;Bakke and Wibbels, 2006).Nonetheless, in Tranchant's (2010) view, this scenario is realistic only if a certain ethnic minority constitutes a majority in its jurisdiction.Otherwise, decentralization could actually encourage this minority to fight against the local majority or against the central government in order to obtain self-administration (Cornell, 2002).Even when this is not the case, some suggest that decentralization can represent a potential threat to national unity (Ranis, 2012) as heterogeneity may increase the potential factionalism (Azfar et al., 1999).It should also be acknowledged that local heterogeneity does not imply that different groups have their own particular interests as people do not necessarily identify themselves only with ethnicity, religion or language (Azfar et al., 1999).
From an empirical point of view, Panizza (1999) was among the first scholars that observed that decentralization is positively and significantly correlated with ethnolinguistic heterogeneity; previously, Oates (1972) found a positive but not significant relationship in his regression.Panizza's (1999) results were later confirmed by Letelier (2005), Beramendi (2007) and Canavire-Bacarreza, Martinez-Vazquez and Yedgenov (2016).On the other hand, there are a couple of papers that suggest that, while significant, the relationship is negative rather than positive (Freinkman and Plehkanov (2009) and Nana (2009) in single country analyses; Arzaghi and Henderson (2005) and Tranchant (2016) in cross-country studies).As a further acknowledgment of the relationship, many other studies have used ethnolinguistic heterogeneity as a control variable in regressions that are measuring the impact of decentralization (Canavire-Bacarreza, Martinez-Vazquez and Yedgenov, 2016).

Decentralization in Romania
In Romania, the seventh largest country in terms of population in the European Union, ethnic Romanians make up almost 90% of the population.The largest ethnic minorities are Hungarians and Romani people, who make up 6.5% respectively 3.3% of the population.The country was ranked 86 th out of 159 analyzed countries in Alesina's et al. (2003) cultural diversity index, and it was ranked even lower on ethnic fractionalization (161 out of 215), linguistic fractionalization (137 out of 215) or religious fractionalization (162 out of 215) indexes.Fearon (2003) also considers it to have a relatively low ethnic fractionalization score and ranked it 112 th out of 159 analyzed countries.Nonetheless, there are large variations between the regions.While Wallachia and Moldova are homogenous, Transylvania has been always ethnically-diverse as, in addition to the ethnic Romanians, it historically had a significant number of Hungarians, Germans and Jews.After the Second World War, the Hungarian minority enjoyed a certain degree of autonomy but after 1968 the communist regime started pursuing a nationalistic policy and tried to homogenize the heterogeneous territories (Chen, 2003).
After the 1989 revolution, ethnic nationalism was constantly used in the public discourse especially in relation to the Hungarian minority (Linz and Stepan, 1996).Thus, in the early 1990s and especially after the ethnic clashes of Târgu Mureș which left six people dead and 300 injured in 1990, proposing any form of devolution has been characterized as anti-constitutional behavior and was banned from the public discourse for several years.Furthermore, the increased activism of the politically organized Hungarian minority, which requested self-governance and some form of territorial autonomy, strengthened fears of unintended consequences of decentralization (Horváth and Scacco, 2001;Csergo, 2007).In 2001 a group of Hungarian and Romanian intellectuals from Cluj-Napoca, which considered Transylvania a different region from the rest of the country in terms of historical cultural heritage, ethnic composition and even electoral behavior (Molnar, 2000), requested the central government to start decentralizing as this might improve the Romanian-Hungarian interethnic relations in the region.The central government on the other hand saw decentralization as a matter of national security and implemented only small and incremental reforms (Csergo, 2007).
While the first post-communist Constitution in 1991 laid out the principles of local autonomy, it ruled out any political decentralization.The 2003 revised Constitution introduced the concept of deconcentration and allowed ethnic minorities to use their own language in local administration in areas where they are a majority.Subsequent governments have decentralized, to a certain degree, various services like land management, local road infrastructure and public transportation, local cultural institutions, local public health units, territorial development and urban planning, water supply, sewage, public lighting, social assistance and protection for children, elders and victims of gender-based violence.There are also shared competences between central and local governments in areas like heating provision, social housing, primary and secondary education, public safety, social welfare, emergency situations, medical and social assistance services or population records.All other services are either handled centrally or deconcentrated under the supervision of the prefect which is the central government's representative (Profiroiu, Profiroiu and Szabo, 2017).
The research in the area is scarce.In a study from 2006, Profiroiu and Profiroiu suggest that most local elected officials tend to favor decentralization.In smaller communities, however, there was a more negative opinion.In a survey undertaken in 2008, patients did not see decentralization as a solution for the problems of the Romanian health system (Popa, 2011).Another study suggested that ethnic Romanians were concerned about devolving authority to the subnational level (Bako et al., 2006).A public survey administered by the CCSCC (Centrul de Cercetări și Studii Calitative și Cantitative) in 2013 suggested that almost 50% supported decentralization and most agreed that service delivery would be better administered at the local level.On the other hand, 85% believed that decentralization might lead to further empowerment of the local leaders, 75% considered that it would increase corruption and a majority thought that decentralization could increase regional inequalities.
As such, in light of the arguments outlined above the study proposes the following hypothesis: 'citizens from ethnically heterogeneous regions are more sensitive to rhetoric about decentralization'.

Method and data
The study compares the two largest cities in Romania, which are quite different in terms of ethnic heterogeneity.Bucharest (the capital) is located in historical Walachia and is the sixth largest city in the EU with a mostly homogenous population of around 2 million.The second largest city of Romania, Cluj-Napoca, is located in historical Transylvania and has a population of roughly 400,000 out of which 50,000 are ethnic Hungarians.The study uses cities instead of counties because, as Sujarwoto (2012) mentioned, the locus of decentralization reform is located at the local government level which is below the provincial administration.Bucharest and Cluj-Napoca are also relevant places to do the research since, as mentioned in the previous chapter, the debate about decentralization in Romania was mostly discussed in these two cities and most often, at opposite ends.
Using snowball sampling, electronic invitations to participate in the survey were submitted to inhabitants in each city.While this method has certain disadvantages like community bias, lack of control and anchoring, Noy (2008) suggests that this widely used technique can generate a unique, emergent, political and interactional type of social knowledge.299 responses in Bucharest and 252 responses in Cluj-Napoca were collected in October 2016.After removing those that were incomplete and those filled out in less than two minutes, while making sure that the completed surveys matched the characteristics of the target samples, 208 useful responses in Bucharest (0.01% of the population) and 153 in Cluj-Napoca (0.04% of the population) were retained.
The two samples are similar in terms of gender, education, professional status and income but rather different in terms of ethnicity.While in Bucharest 97.60% of the respondents are Romanian, in Cluj-Napoca, 11.80% are of a different ethnicity.These figures reflect to a certain degree the current demographics of the two cities since Romanians make 96.6% of the population in Bucharest while in Cluj-Napoca 18.5% are of a different ethnicity.The characteristics of the sample can be found in Table 1.
The study uses a scenario-based survey experiment that follows a 2 X 2 factorial, between-subjects design.Due to their unmatched capacity of establishing causality, experiments have been used extensively since the late 1960s to investigate a wide range of issues in social sciences as they offer accurate insights into trends experienced in real life (Mutz, 2011).In a study undertaken in 2006, Druckman et al. found 57 experimental studies, including survey experiments, published in the first 100 volumes of the American Political Science Review and suggested that their number had been growing rapidly.The decision to use this research design is based on the assumption that the treatments and the randomization can point towards the direction, if not the rough magnitude, of the effects that might be expected to occur in the real world (Gaines, Kuklinski and Quirk, 2007).In addition, randomized Internet survey experiments that analyze how the presentation of information alters beliefs are seen as a very valuable tool for studying public opinion (Lavine, 2002;Barabas and Jerit, 2010).
Despite their great potential, there have been very few attempts to analyze the phenomenon of decentralization using the experimental design, although it is a wellknown fact that standard empirical research on the effects of decentralization is influenced by endogeneity and omitted variables bias and many scholars have suggested expanding the research methodologies (Asatryan and Feld, 2015).Guth, Levati and Participants were invited via email through the Qualtrics computer software to participate in the study where they could follow a link leading to the stimuli and subsequent survey.At the beginning of the experiment, subjects in both cities were provided with instructions and a brief description of the experiment.After agreeing to participate, subjects were randomly assigned to one of the two treatment groups where they were provided with a different fictitious scenario within their city.Both vignettes present a press release from a candidate running for mayor in which he advocates either for more centralization or for more decentralization.According to the results of the chi-square test for homogeneity (p>0.05 in all tests in both cities), the randomly assigned respondents in sample are not significantly different in terms of age, gender, education, professional status, income or ethnicity.Thus, it can be assumed that the difference in responses is influenced by the manipulation and, due to randomization, any difference in answers can be interpreted as an effect of the treatment (Blom-Hansen, Morton and Serritzlew, 2015).
The treatments are in line with previous survey experiments and are reflective of the real world.After reading one of the two vignettes, all subjects were asked to respond to the same questions.They were asked to what extent they agreed that 12 public services should be controlled and financed by the local governments within their city and not by the central government.While the list of public services provided at the local level is much longer, these particular 12 were considered to be the most representative as they combined centralized services (higher education, emergency services, judicial system, building of social housing), decentralized services (local public transport, street administration, parks and recreation) and services with shared competence between the central and the local level (primary and secondary education, hospitals and medical care, police, social assistance and child protection, museums and public libraries).
All items were measured using a 7-point Likert-type scale where a score of 1 corresponds to 'strongly agree', and a score of 7 corresponds to 'strongly disagree'.Before launching the main experiment, a pilot study was run on a sample of 16 persons (4 in each treatment group).Initially the survey included two questions related to support for centralization and support for decentralization.Based on the feedback obtained, the first question was dropped from the experiment.No pre-testing was undertaken since this could have primed respondent and biased the causal estimates.The vignettes and the question survey are presented in Annex 1.
For the aggregated measure of support for decentralization, instead of averaging responses to the 12 services used to measure each of the constructs, the analysis constructed the composite measures using principal component factoring (DeVellis, 2016).This method is more accurate for measuring latent constructs as it weights the contribution of each item to the latent construct according to the degree of commonality it shares with the said construct (Porumbescu, 2017).Each of the latent constructs used in this study explained variance from 17% to 39% in Bucharest and 10% to 34% in Cluj-Napoca; details are provided in Table 2.

Main findings
The mean values of the four treatment groups are presented in Table 3 while the detailed mean values of all the 12 services are shown in Table 4.
A one-way analysis of variance was used to assess whether there were significant differences between the two treatment groups in each city.The results reveal that participants in Bucharest seem to not have been significantly influenced by the manipulation since there is no significant difference in responses between the two treatment groups as F(1,206) = 1.644, p =.201.This suggests that Bucharest participants' attitudes toward questions of decentralization or centralization of public service provision are not very sensitive to political rhetoric on the issue.However, contrary to the expectations, for nine public services respondents that received the centralization manipulation seem to support greater decentralization than those that received the decentralization manipulation.
In Cluj-Napoca, on the other hand, there is a significant difference between the treatment groups as F(1,151) = 4.73, p = 0.03.In other words, it appears that the attitudes of participants from this city are more sensitive to decentralization rhetoric than to centralization rhetoric.Moreover, there is descriptive evidence for all 12 services apart from police that participants that received the decentralization manipulation are much more inclined to support the reform than those that received the centralization manipulation.Furthermore, the Cluj-Napoca sample is much more inclined to support decentralization compared to Bucharest.Apart from parks and recreation, the respondents in Cluj-Napoca appear stronger in their support of decentralization for all the proposed services with significant differences in areas like education, medical care and law enforcement.

Discussion
Taking into account the differences between the two samples, we can conclude that the proposed hypothesis is supported.People from Cluj-Napoca are much more sensitive to rhetoric about decentralization compared to those from Bucharest and they are also much more inclined to support the reform.While this difference is not really surprising, taking into account the different history of the two cities, the author expected to see much more interest for decentralization in Bucharest for a number of reasons.The capital city is by far the most developed city in Romania contributing with 18% to the national GDP, while Cluj County, which incorporates six additional cities and towns and 75 additional communes, contributes with only 4%.Furthermore, the GDP per capita in Bucharest is twice as high as in Cluj-Napoca (19,700 EUR versus 9,400 EUR in 2014) and according to the National Commission for Prognosis (2016) this difference will continue at least until 2020 as both GDP levels are expected to grow at the same pace.Bucharest is also an economy powerhouse at the European level having a GDP per capita of 130% of the European Union average.Combined with the fact that it is the sixth largest city in the European Union, it could have been expected that people living in Bucharest to be more interested in devolving power to the local administration as several other European capital cities like London or Brussels have requested.
Another reason for this reluctance to embrace decentralization is related to the corruption problems of the local administration in Bucharest since two former mayors have already been investigated for corruption.Taking into account the CCSCC survey results from 2013, when a large majority believed that decentralization leads to increased corruption and local capture, it might be assumed that people from Bucharest might be skeptical to empower the local administration even more.This is also portrayed by low turnouts during the local elections (25% in 2005, 30% in 2008, and 33% in 2016).Furthermore, some citizens might not consider being fully represented by the local government due to the first-past-the-post electoral system through which the current mayor won the election by gaining the support of only 13% of the entire Bucharest population that is eligible to vote.
In Cluj-Napoca, on the other hand, there has been a continuous battle for devolution since the Middle-Ages when the Romanian majority continually requested more rights from the Hungarian or Habsburg crowns.Therefore, it is not surprising that its residents are more sensitive towards this reform as decentralization, in this context, may be of greater cultural significance.It should be mentioned, however, that the political tendency in Cluj-Napoca is to vote for parties that are right-of-center.Leftof-center parties, that have been more inclined to promote decentralization in the past decade, have never won the local elections since the 1989 revolution.This is related to the fact that people might perceive the major left-of-center party in Romania to be somehow connected with the communist party, which adopted an excessive centralization policy.
Since the results are not statistically representative for the whole country, the current study does not recommend more or less decentralization.Nonetheless, the results might suggest that an extended prolongation of the status quo might involve political costs for the subsequent governments, at least in certain regions of the county.In this regard, a possible hypothetical solution would be the implementation of the Spanish model which uses an asymmetric decentralization system where some regions are more autonomous than others.Nonetheless, the application of this framework in Romania would be difficult to implement using the current administrative divisions that lack a regional layer.Another hypothetical solution would be to devolve more authority only to certain large and prosperous cities, while not implementing a full decentralization reform.
Regardless of the way forward, the results of this study suggest that practitioners need to find a way to satisfy various heterogeneous preferences as a one-size-fits-all approach cannot be a viable solution on the long term.

Limitations of the study
The geographical distribution of the analysis, certain particularities seen within the two samples and certain limitations related to the use of the research design limit the results of the analysis.Although the results are assumed to be representative for a significant part of the country's population, perceptions in the regions of Moldova, Oltenia or Székely Land might be, and probably are, different or significantly different from those presented in the current study.Although the samples represent a diverse pool of participants, the author acknowledges that it uses a non-probability voluntary sample with strong internal validity but low external validity making it not statistically representative for the whole country.Nonetheless, as mentioned by Remler and Van Ryzin (2010), most behavioral studies using experiments rely on voluntary samples and the aim of the experimental design is obtaining internal rather than external validity.In addition, as mentioned by Mutz (2011), there is no empirical proof for championing one research setting over another for purposes of external validity.
A further limitation is the reduced response rate from people that are over 51 years old, are unemployed or are retired and the significantly large participation of respondents with higher education.While both cities are relatively younger, more educated and with a larger workforce than the rest of the country, the author acknowledges that is more representative for working people or students under 50 that obtained or are pursuing an academic degree.Nonetheless, these are inherent limitations based on the fact that the surveys were administered online using a computer software.
Another limitation of using the survey experiment is the fact that people may not give honest responses, wish to appear consistent with previous answers or to appear to comply with the status-quo of the society (Wenzel, 2005).In addition, it is assumed that the manipulation of one feature from the vignette will generally change respondents' beliefs about other features of the scenario as they are using their knowledge from real life (Guth, Levati and Sausgruber, 2005).

Conclusions
The aim of this paper was to provide preliminary insights into the relationship between ethnic heterogeneity and perceptions of decentralization.Results suggest that there is a relation as people from Cluj-Napoca are much more sensitive to the rhetoric about decentralization compared to those from Bucharest.They are also much more inclined towards the reform although in the past 25 years decentralization was implemented with little planning or analysis and without increasing the capacity of the local governments (Profiroiu, Profiroiu and Szabo, 2017).As mentioned by Profiroiu and Szabo (2016), there is no unanimously agreed solution on how decentralization should be implemented, especially in developing countries.However, Roma-nia should not fall into trap of transposing a successful practice from the developed countries in the West and should frame it in accordance with its complex institutional realities.
If Romanian decision-makers want to adopt some good practices in terms of decentralization, they might look into the case of Poland, a country with a similar history that has strongly empowered subnational governments in the past 25 years (Szabo, 2015).Regardless of the adopted solution, the central government needs to start looking more into the separation of powers between the different layers of government in order to find a right balance that, on one hand respects the principle of subsidiarity, and on the other satisfies to some degree the needs and preferences of all ethnic, linguistic and religious groups in the country.Since further studies on this topic in Romania are much needed, scholars should replicate the analysis in other regions of Romania and try to include the whole structure of the population.

Table 1 :
Samples statistics necessarily related to decentralization, there have been several experiments related to tax compliance and willingness to pay taxes, conducted in the USA and in Europe.

Table 2 :
Total variance explained in the two samples

Table 3 :
Mean values of treatment groups in each sample

Table 4 :
Mean values of the services analyzed in each treatment group