The Price of Decentralization: Linguistic Polarization and the Provision of Schooling*
*** Preliminary version – please do not cite or circulate without the authors’ permission ***
Francesco Cinnirella† and Ruth Maria Schueler§
Abstract In this paper we argue that the decentralization of education policies in Prussia resulted in under- provision of public education in areas characterized by conflicting and divergent interests. Using rich county-level data on public schooling from the education census of 1886, we show that linguistic polarization in East Prussia had a negative effect on local school spending. We obtain this result while holding the municipal tax base, state expenditures, and the agricultural and industrial structure constant. Instrumental variable estimates using distance to the eastern border and the geography of Prussian annexations suggest that the negative relationship between linguistic polarization and public expenditure on education can be interpreted as causal. Our findings suggest that linguistic polarization has a negative effect on local school spending, which results in fewer teachers and school buildings as well as lower teacher wages.
Keywords: human capital, polarization, public goods, decentralization, Prussia. JEL classification: N130, N330, H410, I220
July 31, 2015
* We are grateful to Sascha O. Becker, Pawel Bukowski, Luigi Pascali, Rita Peto, Ralf Meisenzahl, Stelios Michalopoulos, Alfred Reckendrees, Justin Tumlinson, Nikolaus Wolf, Ludger Woessmann, and the participants of the Economic History Talk in Mannheim, at the European Historical Economics Society Summer School in Berlin, the First German Economic History Congress in Muenster, the Conference of the Association for Religion, Economics and Culture in Boston, and at two internal seminars at the Ifo Institute for their comments and suggestions. † Ifo Institute for Economic Research at the University of Munich, Poschingerstr. 5, 81679 Munich, Germany; [email protected]. § Ifo Institute for Economic Research at the University of Munich, Poschingerstr. 5, 81679 Munich, Germany; [email protected]. 0
1. Introduction
Nineteenth-century Prussia is well-known for its advanced educational system. For example, by 1816, the average enrollment rate in primary school was already about 65 percent. In 1871, the share of persons older than 10 years old able to read and write was about 70 percent. Protestantism is believed to be a crucial factor in this country’s high level of human capital (Becker and Woessmann 2009), a belief bolstered by the high levels of literacy in economically poor but highly Protestant Sweden in the 19th century. From a political economy perspective, Lindert (2004) convincingly argues that the success of the Prussian education system was mainly due to its high level of decentralization, which allowed it to respond more efficiently to a local demand for education. At an early stage of development, highly centralized systems, in contrast, would be unable to detect the local demand for education and therefore tend to under-provide public primary education (Hanushek et al. 2013). In this paper we argue that for a given demand for education related to the economic structure, a highly decentralized educational system may result in a comparatively lower spending on primary education in presence of a heterogeneous population. We argue that the comparatively low educational performance of the Prussian eastern provinces can, to a large extent, be ascribed to the linguistic heterogeneity of those areas. More specifically, we hypothesize that the divergent interests of different linguistic groups prevented reaching a consensus on local expenditure on primary education. The eastern part of Prussia had always been a melting pot of German and Slavic culture and the foundation of the German Reich in 1871 fueled national feelings on both sides. Becoming part of a German state increased the minority’s awareness of its own ethnic identity and exacerbated conflicts of interest. We use language as a salient measure for diversity as language policy played a major role in the Germanization policy of the Prussian state and because linguistic identity was the only ethnic identifier that Prussian statistical authorities provided (Heinrich 1992a). As school policy played a major role in the enforcement of Germanization policies, we look at the provision of public school funds as a means of investigating the effect of diverging interests. As decisions on school funding were taken at the municipal level, that is, by the local school boards, we expect there to be conflicts over the provision of public education in linguistically polarized counties. Using data from the first comprehensive education census of 1886, published by the Royal Statistical Office of Prussia, we show that the presence of two linguistic groups of similar size, that is, high polarization, led to lower expenditure on public primary education. Specifically, our
1 rich dataset allows us to look more deeply into the allocation of school funds and investigate the impact of linguistic polarization on school density, student-teacher ratios, and teacher wages. The educational census of 1886 provides a unique opportunity to study the impact of linguistic polarization on the local provision of public education as the state’s contribution to primary education was still very limited at that time. In fact, decisions on school funding were at the discretion of local authorities; no statewide law regulating school funds was passed until 1906 (Hühner 1998). We address the issue of causality using an instrumental variable approach. The arguably exogenous variation in polarization comes from the geography of the Kingdom of Prussia’s annexation policies. The successive annexations toward east, for example, the Partitions of Poland, increased the share of ethnic Polish in the population and thus the linguistic heterogeneity. Thus, each province’s distance to the eastern border is used to identify variation in linguistic polarization. Instrumental variable estimates confirm the negative effect of linguistic polarization on local expenditure on primary education. We find a substantial negative effect of polarization on primary education expenditure and one that affected several aspects of school-related spending, including school density, child-to- teacher and student-teacher ratios, and teacher salary. Therefore, we are able to shed light on why East Prussia underperformed in education and, consequently, in economic performance, as basic education played an important role in the Prussian industrial revolution (Becker et al. 2011). Our rich dataset allows us to rule out many possible confounding factors such as income, religious fractionalization, and urbanization. To measure income as accurately as possible, we collected new data on the municipal tax statistics of 1883/1884. We find that the negative impact of linguistic polarization on educational expenditure holds for a given level of economic prosperity as measured by municipal tax revenues. Accounting for the subsidiary role of the central state does not affect our main result either. Additional specifications illustrate the effect of linguistic polarization on other municipal obligations, chiefly infrastructure and poor relief. Consistent with the notion that linguistic polarization was detrimental only for education policies, we find no consistent effect of linguistic polarization on expenditure on infrastructure or on policies related to poor relief. The paper is structured as follows. Section 2 sets out the historical background of 19th- century Prussia. Section 3 surveys the literature. Section 4 describes the dataset. Section 5 presents OLS results. Section 6 presents the instrumental variable approach and its results. Section 7 establishes the falsification exercise. Section 8 examines distributional effects, and Section 9 concludes.
2 2. Historical Background
2.1 The Political Economy of Schooling in late 19th-Century Prussia
In 1886, Prussia’s school system was still locally organized. According to (Lindert 2004), this state of affairs was largely responsible for the school system’s success. Indeed, Prussia’s high school- enrollment and literacy rates made it a role model for other European countries (Clark 2007). At the same time, however, the Prussian administration was aware that the school system in the East, and especially in those regions with a high share of Polish speakers, lagged behind the rest of the country in terms of student-teacher ratios and enrollment rates (Königliches Statistisches Bureau in Berlin 1889), largely due to low school-related investment. We argue that this area’s lower investment in the educational system might have stemmed from high linguistic polarization in these parts of Prussia, a situation itself exacerbated by local organization of the schooling system. The historical literature supports such a relationship by pointing out that municipalities with a linguistically mixed population suffered especially when it came to the provision of school buildings and teachers, explicitly mentioning the coexistence of languages as a reason for this situation (cf. (Grzes 1992); (Heinrich 1992b). Local organization of schooling meant that the local municipality was responsible for levying and allocating school funds (cf. (Lindert 2004; Heinrich et al. 1992; Lamberti 1989). In many places, provincial or even district legislation regulated school financing, but in the absence of such regional regulations, the Allgemeines Landrecht, the Prussian subsidiary law, held. However, regardless of which law applied, the municipality was the decision-making authority when it came to school finance (Königliches Statistisches Bureau in Berlin 1889). In rural municipalities, local patronage, at least in some cases, was able to smooth ethno-linguistic tensions. In cities, however, the organizational reforms (Städteordnung) of 1808 and 1831 regulated school affairs as municipal duties, that is, citizens and, therefore, parents were given a voice in school matter. Moreover, schools were separated by denomination, leading to linguistic segregation in cases where denomination and language collided. Interconfessionally mixed schools (Simultanschulen) existed in municipalities that could not afford operating more than one school (Groeben 1992b). School funding came from various sources, including tuition fees, foundations, schooling societies, municipal taxes, and contributions from the nobility. Funding schemes varied across Prussia and are explored in detail in Section 4.1 of this paper. Decisions as to how to allocate local school funds were made by the school board (Schulvorstand), which consisted of the noble landlord in the estates or the mayor in urban and rural municipalities, the local clergyman, and two to four members of the school district (Schulgemeinde) whose election had to be confirmed by
3 the chief administrative officer of the county, the Landrat. In 1887, the likelihood that members of the Polish-speaking population would sit on these school boards decreased when a ministerial decree ruled that the county’s Landrat should confirm members of the school board only if would-be members were inclined to foster the German-language policy imposed by the Prussian authorities. The purpose of this decree was to assure that Polish and other minority languages were effectively eliminated from being taught (Glück 1979). According to (Grzes 1992), linguistic barriers and diverging interests led to particular problems in the organization of primary schools in linguistically mixed regions. A uniform minimum wage for teachers was not introduced until 1897 (Berger 1897) and a coherent framework for the maintenance of public elementary schools was not enacted into Prussian law until 1906 (Anderson 1970; Lamberti 1989). In addition to school funding, the municipality was also widely responsible for appointing teachers (Glück 1979) until 1886, at which time a law in regard to the appointment of school teachers in the provinces of Posen and West Prussia was enacted. This law shifted the authority to recruit and hire teachers from the local to the state level in these two provinces (Lamberti 1989), but before its enactment, in general, all decisions on hiring teachers, teacher pay, and facility upgrades were made by the municipalities. Thus, 1886, the year in which the reforms were instituted, is the last year in which it is possible to observe the effects of linguistic polarization on educational matters at the local level.
2.2 The Origins of Linguistic Polarization in East Prussia
Prussia’s core territory consisted of the Margraviate of Brandenburg and the Duchy of Prussia. The Margraviate of Brandenburg was primarily inhabited by Germans, with very small shares of non-German speakers. Berlin, the administrative center of the empire, was situated in this territory. As the Duchy of Prussia had been under the rule of the Teutonic Order, the majority of its population traced back to the German colonists. However, the 300 years under Polish sovereignty resulted in a substantial number of Polish- and Lithuanian-speaking Prussians in this territory. The immigration, over centuries, of Huguenots, Lutheran Austrians (Salzburger), Dutchmen, Scots, Lithuanians, Masurians, and Poles had turned the Duchy into a melting pot. The Prussian census of 1886 reports three language groups in this territory: Germans, Poles, and Lithuanians. The Polish-speaking individuals mainly identified as Masurians, who spoke a Polish dialect and were Protestants. They had arrived in the territory in 1460 as religious refugees. Even though the Masurians distinguished themselves from the Catholic Poles, the Prussian statistic does not indicate the Masurians as being a separate ethno-linguistic group. Over time and
4 especially after the war experience of 1871, the Masurians identified primarily as Prussians and increased their usage of German. The Lithuanians in the Duchy of Prussia considered the usage of their language essential to the practice of their beliefs, which put them in opposition to school reforms abolishing the use of minority languages during religious education classes. The archdiocese of Magdeburg fell to Brandenburg in 1541. Germans arrived in this area during the German eastward expansion and thus the originally Slavic territory had been Germanized and Christianized relatively early. After the Thirty Years’ War, Eastern Pomerania became part of Prussia in 1648. Western Pomerania was acquired 70 years later, that is, in 1720 after the Great Northern War. The Pomeranian population consisted of rural colonists stemming from several Germanic tribes that had mixed with local Slavs and who had Germanized in the past. The forest between Pomerania and the former Poland served as a natural border, inhibiting exchange between Poles and Germans at the time the territories fell to Prussia. In the very northeast of Pomerania were the Kashubians, a numerically negligible minority that, nonetheless, succeeded in protecting itself against German influences by being granted far-reaching minority rights. In a long acquisition process that stretched from 1740 to 1763 and involved the two Silesian Wars and the Seven Years’ War, Prussia finally acquired Silesia in 1763. As parts of the province had been devastated by the wars, Frederick the Great initiated a reestablishment plan that involved the settlement of foreigners in the province and reorganization of the province according to the Prussian model. This consolidation phase continued until the end of Frederick the Great’s reign in 1786. In 1886, three-quarters of the Silesian population indicated German as their principal language, whereas this province’s Poles and other Slavs chiefly resided in the district of Oppeln (Neubach 1992). In 1772, the First Partition of Poland resulted in the Netze District, Pomerelia, Chelmno Land, and Warmia becoming part of Prussia. Sixty percent of the population in these territories was German. As the new province connected the core lands of Prussia, that is, the Margraviate of Brandenburg and the Duchy of Prussia, the Prussian authorities were interested in imposing a structured organization on the new province, which was directly controlled by Berlin. Medium- level administration personnel were recruited from nonresidential officials. The German-speaking nobility and the bourgeoisie mixed through intermarriage; however, the Polish Catholic nobility was excluded from administrative positions. Its market access to both the Margraviate of Brandenburg and the Duchy of Prussia made the newly acquired territory of value which led to an expulsion of the land-holding Polish nobility. The Polish nobility had held 80 to 90 percent of the landholdings in 1772, but by 1806, this number had fallen to around 50 percent. This rollback
5 policy generated mistrust between the Polish- and the German-speaking populations, which became evident during county councils and other assemblies at the local level. Even though language itself did not create insurmountable barriers, different traditions in administration and different values resulted in irritation and conflict, leading to the establishment of stereotypes as well as to different understandings of governance (Bömelburg 2013). After the Second and Third Partitions of Poland in 1793 and 1795, Posen became part of Prussia. This meant that Posen, a province with a majority of Polish-speaking individuals, needed to be integrated into the Prussian Empire. Only one-third of the population in this province was of German descent. However, Germans and Poles cohabited relatively peacefully until initiation of Germanization policies in the 1830s (Struve 2013). Germanization peaked with the 1873 and 1876 laws on language usage, with the perhaps not unforeseen effect of fostering Polish nationalism. The laws were passed during the Kulturkampf, the struggle between the Catholic Church and the Prussian authorities between 1871 and 1879, which ultimately destroyed Polish- German relations in the province of Posen (Unruh 1992a). Although Napoleon reestablished a Polish state—the Duchy of Warsaw—between 1807 and 1815, Prussia was restored in its borders before Napoleon’s conquest during the Congress of Vienna in 1815. Moreover, Lower Lusatia and New Western Pomerania were bestowed on Prussia by the Congress of Vienna. Lower Lusatia was populated by the Wendish minority, a Slavic people that had converted to Protestantism during the Reformation (Unruh 1992b). In short, the coexistence of German and non-German speakers observed in Prussia at the end of the 19th century primarily dates back to the conquest of Pomerania and Silesia in the 17th and 18th centuries and to the three partitions of Poland at the end of the 18th century as well as to the so-called fourth partition of Poland during the Congress of Vienna in 1815 (Hansen & Wenning 2003). The partitions of Poland had the consequence that one-third of the Prussian population had formerly belonged to the Polish-Lithuanian Commonwealth (Alexander 2008). These Slavic regions had experienced an in-migration of German settlers since the early Middle- Ages, following the territorial expansion of both the Holy Roman Empire and the Teutonic Order in clear West-to-East lines. The end result was that Poles and Germans lived side by side, even before these regions became Prussian (Zernack 2008). After coming under Prussian rule, Prussia’s active settlement policy, which had as its objective the population of vacant or devastated land, increased the level of polarization. While settlers initially came from German- speaking regions such as the Palatinate or Franconia, periodic wars and epidemics led to a shortage of German colonizers. This is what ultimately led to Lutheran Austrians (Salzburger),
6 Masurian Poles, and Lithuanians, as well as the French Huguenots,1 being invited to settle in the southern and eastern regions of East Prussia. After the three partitions of Poland, the Prussian state authorities reengaged in their population-enhancing policies by inviting and sponsoring German colonizers to settle in the newly acquired Grand Duchy of Posen and, especially, in West Prussia (Zbroschzyk 2014). The West-to-East migration was only reversed after foundation of the German Reich in 1871 when people—of both Polish and German origin—began to move to the industrialized regions of the Rhineland (Wünsch 2008).
2.3 Prussian Language Policy throughout the 19th Century
Historians concur that it was the foundation of the German Empire in 1871 that ultimately initiated German-Polish antagonism (Alexander 2008). The German Empire was understood as a nation state, as opposed to the Prussian Empire, which was built on an abstract and fragmentary common identity of its subjects (Clark 2007). Until the foundation of the German Empire in 1871, the Prussian authorities considered the Polish- or Slavic-dominated regions as “colonies” with their own cultural identity and, as such, supported the use of Slavic languages and local dialects, especially through primary schooling (Clark 2007). The right to freely move across inner-Polish borders and to maintain one’s Polish identity had been legally protected as early as the Congress of Vienna in 1815 and such rights were strengthened by the edict of 1822, which recognized the importance of language and nationality to Polish-speaking citizens. These rights were also enforced in the primary school system. For example, a regulation of 1842 mandated that the language of school instruction was to be that of the majority of the pupils. (Lamberti 1989). After the foundation of the German Reich, Imperial Chancellor Bismarck began to promote Germanization policy. The primary school was to play a key role in this policy by socializing the Polish-speaking students in East Prussia. (Lamberti 1989). Several laws were passed in provinces with Slavic minority populations that gradually established German as the only language of instruction. More than half the curriculum was devoted to German and history lessons aiming at educating loyal Prussian citizens (Lundgreen 1976). Another measure deployed to achieve Germanization was the introduction of the Ostmarkenzulage, a bonus paid to teachers who acted in line accordance with the wishes of the Prussian state (Alexander 2008). However, the Germanization policy did not achieve its objective of a homogenous German-speaking Prussian nationhood. Instead, Poles began to establish their own parallel society by founding banks, organizing themselves in clubs, and passing on Polish-specific human
1 For the effect of the Huguenot immigration into Prussia on productivity, see (Hornung 2014).
7 capital in confirmation classes (Alexander 2008).2 Foundation of Polish cooperatives was a central aspect of the Polish national movement. Moreover, Poles disengaged from the public primary education system and sent their children to private schools (Groeben 1992b). This parallel structure meant that Germans and Poles rarely interacted voluntarily in everyday life, which inhibited the exchange between the two groups that could have resulted in mutual values and beliefs, and, consequently, better cooperation. The social conflict between Germans and Poles peaked with the Polish school strikes in Wreschen in 1901 and in Posen in 1906 (cf. (Lamberti 1989).
3. Literature Review
This paper contributes to two strands of the literature. First, it builds on the vast literature on ethnic fractionalization and the provision of public goods that was initiated by (Alesina et al. 1999) by framing a theoretical model that links the heterogeneity of preferences across ethnic groups to the amount and type of public goods provided. This paper furthermore explores the negative association between ethnic fractionalization and a range of public goods using data from U.S. cities, U.S. metropolitan areas, and U.S. urban counties. Building on neo-institutional economics, the authors argue that the main channel for this negative association is in-group bias, that is, people tending to favor their own kind. The literature provides evidence in support of this argument with its findings that public good provision, redistributive policies, and effective governance are less prevalent in ethnically diverse municipalities or countries (compare the survey by (Alesina & LaFerrara 2005). The early literature restricts itself to showing mere correlations. However, Dahlberg et al. (2012) and Gerdes (2011) exploit natural experiments that randomly placed refugees across Swedish or Danish municipalities, respectively. Dahlberg et al. (2012) find that previous estimates are positively biased and thereby underestimate the true negative effect of fractionalization (for an overview on the recent empirical literature with a focus on the experimental literature also, see the survey by Stichnoth & Van der Straeten 2013). Secondly, our paper contributes to the debate over whether decentralizing provision of public education is beneficial or detrimental. The provision of public education leads to a collective action problem as no party is willing to provide a good that can be used by the other group without exclusion. Overall, the literature examining the role of decentralization in the provision of public schooling is inconclusive (cf. Gallego 2010; Hanushek et al. 2013). For
2 (Alexander 2008) argues that the Prussian policy against linguistic minorities and especially against the Poles eventually led to the country’s decline.
8 example, Oates (1972), Inman and Rubinfeld (1997), and Lindert (2003) stress the positive role of decentralization, whereas Willis et al. (1999), Bardhan (2002), and Kremer et al. (2003) argue that decentralization leads to inefficient provision of education. In a cross-country study, Hanushek et al. (2013) reconcile these opposing views by finding that decentralization is beneficial in developed countries, but has the opposite effect in the developing world. The authors use principal-agent theory to explain their findings, arguing that decentralization reduces the asymmetric information problem as local information is provided but, at the same time, is detrimental where diverging interests impede decisions. Hanushek et al. (2013) can only hypothesize that the downside of decentralization in lower-income countries stems from diverging interests, but the foundation of the German Empire in 1871 and the subsequent increased antagonism between Germans and Poles in East Prussia allows us to empirically test this hypothesis. As the primary school system was decentralized across the entire Prussian Empire, we are not able to look at the counterfactual. However, previous empirical literature on the negative association between ethnic fractionalization and the provision of public goods has focused on U.S. or cross-country evidence, and thus studying East Prussia, an economically underdeveloped region, provides a good laboratory for testing this channel. Along the same lines as this paper, Miguel and Gugerty (2005) investigate the impact of ethnic diversity on local school funding in Kenya by using historical settlement patterns as an instrument. They find that shifting from complete ethnic homogeneity to average school ethnic diversity decreases school funding by approximately 20 percent of the mean local school funding per pupil. They suggest that the inability of ethnically diverse school boards to enforce social sanctions within the municipality3 is crucial to the allocation of school funds. We are not able to examine the mechanisms behind lower provision of educational funding, but we do explore a unique setting where students are segregated along religious lines and consequently, in many cases, also by linguistic affiliation (Grzes 1992),4 and where funding decisions are taken at the municipal level (where different linguistic groups have to interact in order to agree on funding decisions). In the framework of our historical setting, Bukowski (2015) shows the long-term impact of Prussia’s educational policy in regard to its Polish minority in his investigation of the effect of the Polish Partitions on today’s Polish students, arguing that it was Prussia’s oppressive policy
3 These findings are backed up by the experimental literature in this field (cf. Habyarimana et al. 2007). 4 An exception being the interconfessional schools (Lamberti 1989).
9 toward Poles that has had a persistent and deleterious effect of educational outcomes in formerly Polish regions.
4. Data
The data originate from the first educational census of 1886, which was published by the Royal Statistical Office of Prussia (Königliches Statistisches Bureau in Berlin 1889).5 The census was conducted every five years until 1911 and again after World War I, both in 1921 and 1926. The census provides detailed and comprehensive information on enrollment rates, school facilities, teachers, class size, school organization, teacher income, and student background for Prussian counties, the smallest administrative unit. We restrict our analysis to East Prussia where linguistic polarization was most prevalent and more salient and thus exploit variation across 214 counties.6 We concentrate on the first education census of 1886 since—as discussed in Section 2.1— that was the last year in which the level of school expenditure at the local level was high and the central state played only a subsidiary role. This environment is illustrated in Figure 1, which shows the share of state subventions to primary school funding. Clearly, 1886 is the last year for which state subventions are modest, after which they reach over 40 percent of the total expenditure on public primary education.
[Insert Figure 1 about here]
The education census of 1886 provides data on the language spoken in the pupil’s home. The census distinguishes between German, Polish, Lithuanian, Masurian, Slavic, Danish, and other languages. The census also contains information on whether the reported language is the only one spoken at home or whether both German and the other language are spoken at home. When constructing the index of linguistic polarization we include the latter category of bilingual pupils in the group of non-German speakers. We assume that the interests of this bilingual group are more aligned with the interests of the non-German group. In any event, omitting this group when constructing the polarization index does not affect our main results.
5 For a detailed description of the source, see (Becker et al. 2014). 6 We decided not to include Schleswig-Holstein in the analysis as the instrumental variable approach is not expected to identify linguistic variation in that area. However, our baseline results hold when we include counties in Schleswig-Holstein.
10 We measure the antagonism between Germans and Poles or other Slavic populations by using the polarization index proposed by Esteban and Ray (1994),7 which is also employed by Fearon and Laitin (2003) and Collier and Hoeffler (2004), among others. The polarization index is traditionally used to explain (civil) conflicts. Even though the antagonism between the German-speaking Prussians and the Slavic minorities did not culminate in a violent conflict at the time, German-Slavic antagonism was a severe social conflict in that two groups strongly opposed each other in social interaction, culminating first in two parallel societies and finally in two separate states (Alexander 2008). As the Prussian case is characterized by the dualism between Germans and Slavs rather than by linguistic fractionalization, and as recent work by Esteban et al. (2012) shows that linguistic polarization is related to conflict over public goods, while fractionalization increases conflict over private goods, we use the polarization index as set out below to capture the antagonism between German- and Slavic-speaking populations:8