Regional variation in the distribution of property rights over land in sixteenth-century Ottoman rural Manisa*

ter and regression) models in order to examine the regional and local differences in more detail. Indeed, our observation that levels of wage dependence – by the imperfect proxy of proportion of servants in total population – significantly reduced the strength of the observed correlations, suggests that any true understanding of interactions between demographic behaviour and poor relief provisions needs to be situated in and analysed in interaction with the local socio-economic context. Given that local socio-economic variation was in turn tied in with regional patterns associated with soil types (the so-called agro-systems), this suggests that the regional level – or at least clusters of municipalities – is probably the most relevant unit of analysis to study these interactions in depth. The maps presented in this chapter already demonstrated the importance of spatial patterns in all of the variables studied. While the coastal areas were characterized by capitalistic relations of production, high-pressure demographic regimes and well-developed relief provisions, the inland areas displayed lower levels of wage dependence, lower-pressure demographic regimes and less developed relief provisions. This indicates that a contextualised spatial approach that takes into account interactions between geography, ecology, production relations, demography and relief provisions at the local and regional level, is probably the best way forward to gain better insight into the complex causalities that tied these factors together. Pinar Ceylan

The Ottoman state of the fifteenth and sixteenth centuries is often cited as a centralised state with a strong imperial bureaucratic apparatus. The imperial centre monitored the rural economy through periodic tax surveys, shaped the organisation of rural production through the state ownership of land, determined policies and agricultural law codes, and set limits on how much local landlords could exploit peasants. 1 The methodology of many studies based on tax registers 2 -or, more specifically, the designation of very large geographical areas as units of analysispoints to scholars' tacit acceptance that the Ottoman land regime was implemented in a largely uniform manner throughout the empire's core lands amid negligible differences in tax rates, as well as centrally defined and identical roles, rights, obligations, and powers for direct producers and surplus-extracting classes. Because of this, regional variations in property structures and power relations in the classical period have largely escaped attention, while striking regional discrepancies in the productive capacities of rural communities have generally been ascribed to differences in environmental and market access conditions in different geographical areas.
This paper demonstrates that, despite the relatively centralised nature of the Ottoman state and its government's accordingly greater influence on the organisation of rural production, the pre-industrial Ottoman rural economy shared a fundamental feature of other Eurasian rural economies -namely, regional variation in property rights structures and the configuration of power within rural communities that resulted in different inequality regimes across space. * This research is part of the GINI project, "Economic growth and inequality. Explaining divergent regional growth paths in pre-industrial Europe from late Middle Ages to nineteenth century" supported by Gent University and conducted jointly by the Economics and History Departments. Concentrating on the Western Anatolian administrative district of Manisa and employing tax surveys dating from 1575, this study reveals the existence of two agricultural production systems characterised by different property and surplus relations in the southern and northern parts of the district. Accordingly, inequality in these areas reflected region-specific patterns in the distribution of property rights within and between the classes of direct producers and landlords. In terms of both producers' access to land and the concentration of agrarian surplus among landlords, there were higher levels of inequality within the rural society in the densely populated, more developed, and highly commercialised south in comparison to the more egalitarian socioeconomic structure that prevailed in the mountainous northern part, which was inhabited by a high number of semi-nomadic or settled groups bearing clan status.
With its emphasis on the regional variation in surplus extraction and landholding institutions, as well as on the need for a comparative regional analysis, this paper adopts an approach that is different than that of other studies in the field. It is informed by the "social agrosystems theory," 3 which offers a regional, systemic, and social analysis of pre-industrial rural societies. This theory is systemic in the sense that it defines agrarian production as a system that is shaped by the interplay of mutually influencing factors, including soil and environment; social property relations and power structures; the size of holdings and labour input; labour relations and income strategies; agricultural technology; as well as links with other agrosystemic areas. 4 A fundamental change in any of these factors can effect a change in the agrosystem as a whole, while the overall impact of the change on one factor is determined by the pre-change setting of the system. The social perspective of the method implies that among these factors, social property relations and their connections to power structures play a central role in determining the organisation and efficiency of production. Accordingly, the prime agent of change in the system comes from the social sphere. Finally, the theory is strictly regional, as it champions the region as the relevant geographical unit of analysis in understanding economic and social processes and development. This regional focus is grounded on the premise that prior to the nineteenth century, primary economic activities were mainly regionally organised. Therefore, the determining elements of agrarian systems in pre-industrial societies -particularly social relations -varied from one region to another or even within a relatively small area. 5 The data employed in this study comes from late-sixteenth-century Ottoman tax registers that are available at the Ankara General Directorate of Land Registry 3 E. THOEN, 'Social Agrosystems' as an Economic Concept to Explain Regional Differences, An Essay Taking the Former County of Flanders as an Example  and Cadastre Archives 6 . Data on the population of groups with clan status living in or nearby villages is taken from a secondary source. 7 During the classical period, information on the taxpaying subjects and taxable resources of the empire were collected through periodic surveys and recorded in fiscal registers called tahrir defters. Alongside establishing the expected tax revenues on a village basis, these registers were also used for allocating the foreseen revenues among military and administrative officials. 8 The first register used in this study was the Mufassal Tahrir Defteri No. TKGM T115, which dates from 1575. In preparing Mufassal registers, officials conducted detailed land surveys on a village-by-village basis, determining revenue sources. The information in these registers consisted of three main parts. The introductory section included the village name, the administrative affiliation of the district, and the name of the fief-holders who were entitled to the tax revenues of the village. The second part provided a survey of the adult male population in the village and the amount of land possessed by each. Tax-exempted males, as well as landed and landless peasants who had settled in the village, were recorded here. The third part, meanwhile, listed different tax items (personal taxes, land taxes, tithes, et cetera).
The second register was the İcmal Defteri No. TKGM TD226, which dates from 1572-3. These summary registers showed the distribution of the revenues from tax units allocated as tmars among those entitled to such revenues. For each tmar holder, a separate entry was arranged. Each entry recorded the name, title, and, occasionally, the position of the tmar holder. This was followed by information about the income the holder was entitled to. The income of a tmar holder could be a sum of tax revenues from different tax units (usually a single village, but in some cases, an aggregated tax unit composed of several villages located nearby), or the revenue from a single tax unit could be shared among multiple tmar holders. In this case, both the overall revenue of the tax unit and the share of the fief-holder from this amount were indicated in the entry.
The third register was the Evkaf Defteri No. TKGM TD544 from 1575-6. While the Mufassal and İcmal registers were prepared for state-owned land whose revenues were allocated to tmar holders, Evkaf registers were special lists recording only villages and land owned privately and by waqfs.
This study selected a total of 98 villages from around 170 villages that appear in the tax surveys belonging to the administrative district of Manisa in the late sixteenth century. 9 In constructing the sample, this undertaking included villages whose geographical location could be identified, and for which complete infor-  , 37, 2004, n. 2, pp. 87-102, 3, 88. 9 In this study, only villages under the ownership of the state, waqfs and private individuals were included in the sample. Accordingly, villages inhabited by infantrymen cultivating their own small holdings were excluded due to the lack of detailed information regarding them in the registers. mation was available in the tax surveys. In this, some villages were excluded from the sample because the available information was visibly inconsistent or erratic.
Aegean Anatolia has been a prosperous and densely populated area since ancient times. Over many centuries, the region's economy was shaped by a rich agricultural production and favourable market access conditions. The region not only supplied the imperial capital, Istanbul, but also exported significant quantities of agricultural products to European markets from an early age. In the sixteenth century, the administrative district of Manisa (around 250,000 hectares) was the centre of the Ottoman province of Saruhan and a part of the core lands of the empire in which the classical landholding regime prevailed. 10 Map. 1. Ottoman Empire in the sixteenth century As Thoen reminds us, in studying pre-industrial rural agricultural systems, units of analysis cannot be defined based on political or administrative frontiers, as "the definition of a region and its 'borders' can only be the result of the study itself." 11 An examination of preliminary findings from the Ottoman tax surveys revealed visible regional patterns in Manisa in terms of factors determining the organisation of agricultural production (soil and environment; social property relations and power structures; the size of holdings and labour input; labour relations and income strat-10 K. BARKEY, Networks of Contention: Villages and Regional Structure in the seventeenth-Century Ottoman Empire, in "American Journal of Sociology", 102, 1997Sociology", 102, , n. 5, pp. 1345Sociology", 102, -1382Sociology", 102, , 1355 E. THOEN, Social Agrosystems, a Plea, cit., p. 2. egies; agricultural technology; and links with other agrosystems). As a result, this study identified two different "agrosystems" in the northern and southern parts of Manisa, with a middle zone that could be described as a transition area. Table 1 presents the distribution of the villages in the sample according to sub-regions.

Tab. 1. Number of villages in the sample according to regions
Villages (N)

South 35
Middle 26 North 37 Total 98 Here, this study will briefly touch on regional differentiation in physical geography, market access conditions, population density, and crop patterns before discussing region-specific property rights structures. In terms of physical geography, the main channel of the Gediz River and its tributary, the Kumçay, draw a natural boundary. The northern part is a mountainous area, with lowlands confined between hills. Parts of the area, particularly the rocky hillsides, are unsuitable for cultivation, limiting the extent of arable land. In general, the soil is less fertile compared to the alluvial soil of the river basins in the south. In contrast, the southern part embraces large fertile volcanic and alluvial plains along waterways. While the southern part can be considered a more favourable zone in terms of physical environment -with several rivers facilitating the production of a wide range of marketable products (particularly cotton and rice) -the river system has not always provided an advantage for the region's agricultural economy. 12 The region's rivers and rivulets, especially the Gediz, tended to overflow its banks every year, particularly in autumn and spring when precipitation was at its highest, flooding the agricultural lands nearby. sixteenth-century tax surveys reported several instances in which peasants could no longer cultivate the land because of floods, meaning they had to leave their settlements and could not pay their taxes. As a result, only a portion of the large fertile lands along the riverbanks in this area was available for cultivation at a particular moment. In such areas, the arability of agricultural lands depended on the constant efforts of peasants to reclaim flooded lands.
Regional diversity also applied to market potential. The southern portion incorporated the city of Manisa, which was an urban market of medium size absorbing the agricultural surplus -mainly grains, but also fruits and vegetablesfrom the surrounding countryside. On the other hand, the only natural passageway to the coast and the Foçalar port, which was used for the shipment of local agricultural products, was located between Spil Mountain and the Bozdağ Mountains in the south, giving this part of the district easy access to Istanbul and other distant markets. By contrast, mountains separated the northern part of district from the coast, meaning it had no direct access to large markets (although there were two local markets in this area, Osmancal and Palamut).
It is often stated that the organisation of the agricultural production in Ottoman Anatolia prior to the nineteenth century was characterised by a high landlabour ratio. This implied that the limiting factor of production was labour. 13 Throughout the centuries, the central state made systematic efforts to sedentarise nomads to expand the agricultural labour force, and to keep peasants on their land to maintain continued agricultural production. In this context, the geographical distribution of population was a significant factor determining the productive capacities of rural communities. In the main, settled peasants with subject status -the regular labour form associated with the Ottoman landholding regime during the classical period -and settled and semi-nomadic groups with clan status living in or nearby villages constituted the agricultural workforce in late-sixteenth-century Manisa. 14 Table 2, which presents population figures in three sub-regions of the district, reveals a significant geographical population concentration and a high density in the southern part. 15 The total population living in or nearby the 35 villages in the sample is estimated at around 9,000 inhabitants, with an average settled population 13 S. PAMUK, Osmanl-Türkiye iktisadi tarihi, 1500-1914, Istanbul 2005, 43. 14 A considerable share of Turcoman nomads that came to Western Anatolia in the midthirteenth century fleeing the Mongols, became sedentary or adopted a semi-nomadic livelihood over time, formally retaining the clan status. They remained an integral part of the rural society throughout Anatolia over many centuries. 15 Total population is computed by multiplying the number of married male adults by the average household size and adding the number of bachelor males to this figure. The household size is assumed five in accordance with Ö.L. BARKAN, 'Tarihi Demografi' Araştrmalar ve Osmanl Tarihi, in "Türkiyat Mecmuas", 1953, 10, pp. 1-26. of 257 people per village. The settled population in the northern part was considerably lower than in the south, which is reflected in the much smaller average village size (90 inhabitants). Alongside differences in population density, the two areas differed in terms of the composition of the agricultural workforce, too. The groups with clan status constituted the main component of the direct producer class in the north but held a much smaller share in the southern part, where cultivation by regular peasants was the dominant form. This demographic pattern is accompanied by the existence of a large population of nomadic clans wandering between winter quarters and summer pastures in the north, particularly in the northwestern region of Yunddağ. Sources: Manisa tax registers, 1575. * Production as wheat equivalent in tons, gross of seed and tithes. Table 3 presents the estimated total gross agricultural produce as wheat equivalent in tons and crop composition in different areas. 16 Grains, particularly wheat and barley, predominated in overall agrarian production everywhere in the district. The main cash crops in Manisa were cotton and paddy. Viniculture was also widespread in several villages, while a variety of crops were cultivated on a smaller scale, including sesame, vetch, chickpeas, vegetables, and fruits. A discrepancy across the sub-regions is visible in the productive capacities, crop composition, and variety. The total agricultural output in the south far exceeded the output levels in the other two regions. More than 40% of the total agrarian value produced in the south came from crops other than grains, while the significant share of cotton production (25%), points to the relative importance of market-oriented production in the area. In the north, grain production -mainly for subsistence purposes -held a more significant place (75% of overall yields), while paddy, which was produced for urban markets, was the most important cash crop. Paddy 16 Agricultural yields are computed from tithes, which was an output tax levied proportionally on agricultural yields. Nominal tithe values over different agricultural products were multiplied by the relevant rate and the amount of annual gross agricultural yields is obtained in nominal terms. The yields were then converted to wheat equivalent in tons, using product prices recorded in the registers. cultivation generally took place on the soil along riverbeds, as such areas were unsuitable for the cultivation of other crops. 17 From regional differentiation in physical geography, market access conditions, population levels and production patterns in late-sixteenth-century Manisa, we now turn to the institutions that oversaw property rights. The most important medium of the Ottoman state's control over the agricultural economy was the classical landholding regime, which is often associated with "state ownership of land" and the central state's decisive role in the distribution of agrarian rent among landlords. In the core lands of the empire (Anatolia and the Balkans), a significant share of arable land was registered as "state land," which could not be bought, sold or inherited by private individuals. On such land, surpluses were extracted from direct producers in the form of tax revenues which were centrally assigned as annual revenue grants to members of the ruling class, particularly military groups, in exchange for services. This tax collection and revenue sharing system, which emerged in the fourteenth century, was known as the tmar system.

Tab. 3. Crop composition and yields
In previous scholarship, an oversimplified understanding of the classical Ottoman landholding regime painted a rigid and schematic picture of rural society consisting of a small independent peasantry undifferentiated within itself and a state-dependent landlord class with limited powers over direct producers and vis-àvis the central state. But in a recent study 18 based on empirical evidence from tax surveys, we demonstrated that there was a significant level of institutional diversity in landholding and surplus-extraction, and that this diversity resulted in a rural society composed of many actors with varying degrees of power, rights, authorities, and obligations that ushered in a marked socioeconomic hierarchy at the local level. The present paper advances the argument a step further by demonstrating that the diversity of property rights institutions was not random across space and that even within a small area, such as the administrative district of Manisa, the sub-regions were characterised by different property and surplus relations, which, in turn, created region-specific inequality structures.
In late-sixteenth-century Manisa, the primary sources point to the existence of a multiplicity of actors directly involved in agricultural production under different contractual forms. These included settled or village-based semi-nomadic groups with clan status who were involved in agricultural activity as tenants or sharecroppers; registered peasants on waqf land or private estates; sharecropping peasants on imperial domains; peasants cultivating land under simple tenancy agreements; tax-exempt groups, including infantrymen, who possessed small holdings; owner-occupiers and more. This study limits itself to the two most prevalent forms of landholding, which were associated with two different contracts: tapu (perpetual lease agreement) and mukataa (simple renting contract).
Land freeholder. They possessed hereditary usufruct rights over small holdings, which were cultivated with a pair of oxen and family labour. In contrast, settled and seminomadic groups with clan status cultivated land under a mukataa. Plots not given as family holdings under tapu, particularly in uninhabited villages, were also given under mukataa to landless regular peasants and peasants cultivating small amounts of land under tapu elsewhere. 19 An important difference between tapu and mukataa was the legal guarantee that the former represented in terms of the security of producers' property rights over arable land. Family holdings under tapu could not be taken from the peasant unless the peasant violated the obligations deriving from his status. With appropriate authorisation, the peasant could transfer his usufruct rights over the holding to a third party, and upon the death of the possessor, the holding was bequeathed to the eldest son. 20 Hence, the perpetual lease agreement guaranteeing the irrevocable and hereditary nature of peasant's usufruct rights made the peasant the "hereditary tenant in perpetuity," 21 protecting him against expropriation and ensuring the longevity of the family holding. Unlike tapu, the simple renting contract did not recognise the right to transfer the holding to another farmer or bequeath it to sons, meaning it did not provide the producer with the same secured position against the surplus-extracting class.
Yet, alongside paying personal and agricultural taxes to the landlord -as well as providing him with certain services -tapu arrangements also imposed certain limitations in terms of land use and mobility and stipulated some obligations. A tapu-bearer could not convert his holding into a pasture, vineyard, or orchard, and holdings under tapu could be taken from the peasant and given to someone else if they went uncultivated for three consecutive years in the absence of force majeure. Moreover, if the peasant left the holding and moved somewhere else, he was liable to pay a fine and could be summoned back within 10 years of his departure. 22 The simple rental contract, however, did not stipulate any such obligations other than the payment of relevant taxes; in that sense, it was thus more flexible.
Equally importantly, "while the simple rental contract was freely concluded between the state and an individual, the tapu implied a certain status stemming from an original 'subjugation' which entailed, in addition to the tithes, certain personal obligations, such as the payment of [personal taxes]" 23 that were levied at differing rates depending on the peasant's ability to generate income (extent of land in possession and marital status). The simple rental contract conditioned the use of land to the payment of a land tax proportional to the area cultivated, alongside tithes, but did not stipulate any personal taxes or grant a certain status. 19 Table 4 presents the extent of arable land cultivated under tapu and mukataa in different parts of Manisa. 24 The findings clearly show a regional pattern in forms of landholding by producers. In the south, cultivation by regular peasants under perpetual lease agreements, which guaranteed the security of peasants' rights but imposed more rigid conditions on producers, was the dominant form. While half of the arable land was organised into family holdings under tapu, only a third was given according to a simple renting contract. In contrast, in the northern part of the district, two-thirds of the arable land was cultivated under the more flexible but precarious simple tenancy contract.
The predominance of simple rental contracts in the north was linked to the relative scarcity of peasant labour and the existence of nomadic and semi-nomadic populations with clan status in the area. Simple rental contracts offered more favourable provisions than perpetual leases and offered advantages for producers, such as lower tax liabilities or complete tax exemptions. As such, it was a contractual form used particularly to encourage nomadic and semi-nomadic groups to settle and cultivate land in uninhabited or sparsely populated areas. 25 In the south, however, the abundance of peasant labour fostered the regular form of tapu holdings by settled peasant families.
The regional variation is also visible in surplus extraction. Like the direct producers' class, the Ottoman surplus-extracting class was not monolithic or undifferentiated. Three categories can be identified within this class. The first were locally rooted cavalrymen and other small revenue holders who provided services to the state. These were responsible for supervising the use of land by producers; possessed limited powers over producers; and earned modest revenues in directly collecting taxes from the peasantry. 26 The big absentee landlords within the tmar system -that is the sultan himself, members of his household, high imperial bureaucrats, governors, and members of the military elite -comprised the second group. Within this group, alongside the holders of significantly large revenues that were farmed or collected by government employees, there were also holders of revenues of considerable size, such as local governors of officers in the standing army. The third category was the waqfs and private owners, who usually received 24 The extent of arable land cultivated under tapu, mukataa and other contractual forms were traced back from the relevant taxes reported in the tax registers in çifts, the Ottoman surface unit. 25 Table 5 examines the distribution of rural revenues according to landlord groups in different sub-regions. In the south, 70% of rural revenues flowed to big absentee lords, 10% went to waqfs and private owners, while cavalrymen and other servants -the segment perceived as the quintessential component of the Ottoman surplus-extraction mechanism -appropriated just 20% of the revenues. In sharp contrast, this later category received 60% of the agricultural revenues in the northern part of Manisa.
Overall, a closer look at the evidence from tax surveys from the late sixteenth century depicts a meaningful regional pattern in property rights institutions that determined the distribution of income within rural society. The two agricultural systems in late-sixteenth-century Manisa were associated with different social structures. In general, big absentee lords and regular peasants with secure and longstanding property rights were the characteristic actors of the agricultural economy in the south. In contrast, small lords with limited power on the organisation of agricultural production, as well as semi-nomadic and settled groups with clan status who benefited from more favourable conditions in the short term but lacked secure property rights in the medium term, constituted the major components of rural society in the northern part of the district.
In what follows, this study will examine how this differentiation in social structures was reflected in inequality regimes. The inequality patterns prevalent in different sub-regions of Manisa are evident through a number of indicators at three levels: the extraction rate, direct producers' access to land, and the concentration of agrarian surpluses in the hands of the landlords.
Agrarian-surplus appropriation -in the form of labour, crops, or money rentwas the main mechanism that generated inequality within pre-industrial rural 27 P. CEYLAN, Land Regime, cit. societies and, as a result, should be central to any insight into socioeconomic hierarchies. The Ottoman surplus extraction mechanism was characterised by the extraction of rent in the form of tax revenues and was regulated by the agricultural codes prepared by the agents of the central government. The rates at which each tax was collected in different circumstances, as well as who was responsible for paying these taxes, was specified by the formal codes of each province. 28 While tax rates were generally standard within a province, they varied from one province to other depending on the status of the province, the pre-Ottoman taxation rates, and the system of taxation. 29 This implied that extraction rates were more or less constant across large geographical areas and that the Ottoman surplus extractors did not have a free hand in determining the rate of taxation. Furthermore, the tax burden of the direct producers did not vary depending on the legal status of the land or of the landlords. As a rule, the rates in the provincial codes applied to all producers, regardless of whether they were cultivating state-owned, waqf or freehold land or whether the surplus they generated was appropriated by small servicedependent landlords, big absentee lords, waqfs, or private owners.
In this system, the agricultural surplus was appropriated mainly in three different tax forms: land and labour-related taxes, production taxes, and trade taxes. Land and labour-related taxes involved personal taxes levied on regular peasants at differing rates depending on their ability to generate income (like land ownership and marital status); and taxes on arable land cultivated under a simple rental agreement. Production taxes in rural areas applied to various productive activities in agriculture, husbandry, manufacturing, and other rural economic activities such as apiculture, milling, fishing, and the like. The most important production tax was the tithe, which was collected in-kind at a predetermined rate based on the agricultural yields. 30 Finally, trade taxes were levied on goods brought to the markets. 31 28 M. COŞGEL, The Economics of Ottoman Taxation, University of Connecticut Department of Economics Working Paper Series, Working Paper 2004-02, 2004 In Saruhan, the rate of tithes was one-tenth of total output, while an additional one-fortieth was imposed on wheat, barley, oat, corn, and rye, which was collected as fodder for the horses of landlords. 31 Ibid., pp. 3-6.

Tab. 6. Extraction rate and tax categories
Sources: Manisa tax registers, 1575. * Wheat equivalent in tons. Table 6 demonstrates the amount of agricultural produce extracted from direct producers in the form of taxes as wheat equivalent, as well as its proportion to overall yields. Despite significantly different population densities and productivity levels (which determined the amount of input and output taxes), no regional variation is observed in the level of agrarian surpluses extracted from producers. 32 Expressed as wheat equivalent, the annual gross agricultural production is estimated at 4,838 tons in the north and 7,438 tons in the south. From this amount, 967 tons and 629 tons, respectively (13% of the gross agricultural yields, and around onefifth of the net production in both regions), were transferred to the landlords. Table  6 also looks at the relative importance of agricultural taxes, particularly tithes, within overall revenue. In both areas, agricultural taxes constituted around 90% of the appropriated rural surplus, reflecting the well-established fact that agriculture 32 These figures do not include extraordinary taxes imposed by the state on grain production, to be paid in-kind in wheat and barley. Besides, the forced government purchases that multiplied over the sixteenth century also functioned as a tax. However, it is not possible to determine the burden of these impositions, since they were not recorded in the fiscal surveys, while separate records are also unavailable. H. İSLAMOĞLU-İNAN, State and Peasant in the Ottoman Empire, cit., p.36. was by far the most dominant economic activity in pre-industrial societies. Tithes appeared to be the most important tax category. While differing extraction rates are often an important source of regional variation in the distribution of agricultural revenues within rural societies, inequality structures might show regional differences, despite constant rates across space. In this respect, another important determinant of rural inequality in the pre-industrial setting was inequality in direct producers' access to land. The Ottoman tax surveys provide a less than ideal source to study land allocation within the class of direct producers. For one, these sources report the amounts of land held by regular peasants in a soil-controlled measurement unit (çift) that masks the real level of inequality in access to land. For another, only the size of regular peasant plots held under perpetual lease agreements were systematically recorded in these surveys, meaning it is only possible to estimate land cultivated under simple rent contracts as a lump sum amount. Nevertheless, the available information can provide us with some clues as to the distribution of land use rights and help us understand whether there was a regional differentiation in the pattern of distribution. Sources: Manisa tax registers, 1575. Table 7 reports the ratio of landless to landed peasants and the estimated average plot size in three sub-regions on the basis of two indicators that can be constructed based on data from tax surveys. The findings suggest that in the densely populated south, opportunities of access to land for producers was significantly more limited than in the northern part. In villages located in the south, the number of landless peasants equaled that of the landed peasantry, and the arable land was divided into small plots estimated at an average of between 2 (lower estimate) and 3.5 hectares (upper estimate). In the north, the average estimated plot size (between 4.6 and 6.4 hectares) was more than double the average plot size in the south, while the ratio of landless to landed peasants was 20% lower.

SOUTH
Another aspect of access to land that is often ignored in the relevant literature is common use rights. While tax surveys are silent as to how agricultural production was organised and resources were managed in villages inhabited by semi-nomadic and settled groups with clan status, the collective tax liability and other collective services imposed on the population with clan status, the nature of their livelihood, kin relations, the requirements of the physical environment in the highlands, and the availability of large common lands (woodlands, pastures, and the like) are highly likely to have led to a more collaborative system in the northern part of the district, where these groups constituted the major component of the agricultural labour pool.
The third level of rural inequality this paper investigates is the concentration of agrarian surplus within the landlord class. In pre-industrial societies, the questions of how the agricultural surplus extracted from direct producers was shared among the landlords, as well as whether high amounts of surplus were concentrated in the hands of the few or distributed equally among several small landlords, especially matter with respect to the commercialisation of the agricultural economy. It is widely argued that a certain level of surplus concentration was a prerequisite for the emergence and developments of markets. The size distribution of annual revenues received by landlords in Manisa in the late sixteenth century is presented in Table 8. In terms of the distribution of feudal rent, the findings point to a sharp regional variation in the district. Rural surpluses were highly concentrated in the south, where the mean annual revenue received by landlords equalled 36.4 tons of wheat. Here, the top 30% of landlords had annual revenues equal to or greater than 30 tons. In contrast, the mean revenue equalled 11.2 tons in the north, where only the top 10% received annual revenues equal to or greater than 20 tons of wheat. More than half of the landlords in this part received modest revenues of less than 10 tons.
Ultimately, we need to ask how we can account for the regional differences in property rights structures and inequality patterns. For this, there are two possible explanations that are consistent with the available evidence in sixteenth-century Manisa. The first explanation concerns the physical environment and crop patterns.
Engerman and Sokoloff 33 argue for an association between crop type and inequality. They claim that regions where the soil and the climate were suitable for the cultivation of certain cash crops, such as cotton, sugar, coffee, and rice, were historically marked by highly unequal property rights structures because scale economies exhibited by the cultivation of these crops favoured plantations or largescale farms. In contrast, small production and an egalitarian structure prevailed in places where grain was the dominant crop, as economies of scale were limited for grain cultivation.
Interestingly, despite the general prevalence of small landholdings and typical absence of large-scale farms, the connection between high levels of inequality and cash crop production is also visible in rural Ottoman Manisa. To a large extent, cotton production on the southern plain of Turgutlu took place on imperial lands, with the revenues going to the sultan and members of his household. These domains were managed by state officials appointed by the central government with the prime aim of maximising revenues. It implied that producers on these lands directly interacted with the state, rather than intermediaries, and were subject to higher levels of exploitation. This concerned both the regular peasants who had settled on imperial domains and other populations with servile or semi-servile obligations. 34 Nevertheless, the production of cash crops in Manisa was not limited to imperial domains, and we have seen that tax revenues from land were divided into larger fiscal units in the south, even when cultivation occurred on fragmented peasant holdings under tapu. This suggests that at least part of the explanation should be linked to trade and markets, rather than in economies of scale in the cultivation of cash crops. Since a great deal of the surplus appropriated by landlords was in-kind, it follows that the concentration of agricultural surpluses in the hands of a few big lords was rational only if the amounts appropriated could be transferred to large markets and converted to cash. Thus, it would not be misleading to argue that higher degrees of surplus concentration could only be observed in regions with easy access to big urban centres or distant markets. Alongside proximity to large towns, proximity to ports was also an important determinant of market access given the high costs of land transport in the preindustrial world. From this perspective, the distribution of the agricultural surplus among small landlords in the north of Manisa, in contrast to the prevalence of big absentee lords in the south (which was directly linked to the Aegean coast), does not seem surprising.

CONCLUSION
Existing scholarship has argued that the Ottoman central government of the fifteenth and sixteenth centuries played a decisive role in distributing property rights over arable land and organising agricultural production at the local level. thermore, it has argued that it did so by implementing a landholding regime in Anatolia and the Balkans that it centrally designated, monitored, and enforced. Employing tax surveys from the late sixteenth century in the Western Anatolian district of Manisa, this study has attempted to show that while the influence of the central state on the local agricultural economy was not negligible, it did not create a social structure that was homogeneous across space. The findings suggest that, as in other pre-industrial Eurasian settings, a multiplicity of agricultural production systems associated with different property and surplus relations existed even within a small area in the Ottoman realm as well. The social structures resulting from regionspecific combinations of property rights institutions, in turn, created different inequality regimes. Based on the findings from tax surveys, this undertaking identified two different agricultural production systems associated with different inequality structures in late-sixteenth-century Manisa. The fertile Turgutlu plain in the Gediz River Valley was linked to the Aegean coast via a natural passageway, allowing it to benefit from favourable market access conditions. The area produced considerable amounts of cash crops, particularly cotton, and grains constituted a lower share of the overall produce compared to the north. The high degree of commercialisation and the resultant market-oriented production seem to have characterised the agricultural production system in the area. On the production side, a high population density, and, thus, an abundance of agricultural labour, was the main characteristic. The agricultural workforce in the area mainly consisted of regular peasants cultivating their family holdings, while a dominant share of the agrarian surplus that these peasants generated was appropriated by big absentee lords, waqfs and private owners. Perpetual lease agreements, which recognised peasants' longstanding property rights over land and which stipulated several obligations on the producers to maintain agricultural production, was the prevalent land use contract. In the southern part of Manisa, the high ratio of landless to landed peasants and considerably small plot sizes reflect direct producers' limited access to land. While land was extremely divided in physical terms, the distribution of income from land within the surplusextracting class showed a different picture. Agrarian revenues were transferred to a relatively small number of landlords in the form of large revenue units, and hence, agrarian surpluses were concentrated in the hands of a few. The agricultural production system in the north of Manisa (Yunddağ and Palamut) differed from that in the south in several respects. This mountainous area, large parts of which were unsuitable for cultivation, was separated from the coast. Unfavourable market access conditions were accompanied by a crop composition dominated by grains. The limited amounts of marketable surplus were mainly sold at the two local markets in the area. The density of the region's settled population was also low, and there were many nomads who wandered between winter quarters and summer pastures. The settled population, in turn, mainly consisted of groups with clan status -rather than regular peasants -who lived in or near villages. Groups with clan status cultivated the land under simple rent contracts that did not guarantee producers' longstanding and secure property rights over land, although it did provide them with more favourable conditions in accessing land and often lower tax liability that encouraged them to engage in agricultural production. The feu-dal rent in the region was transferred mainly to cavalrymen and other lower servants of the state who earned modest incomes and exercised only limited power over the producers. In the villages located in the northern part of the district, larger plot sizes, fewer landless producers, and possible access to common use rights point to more ample and equal opportunities in land access for direct producers. At the same time, the agrarian surplus extracted from producers was distributed relatively equally within the landlords' class at a low level of concentration.
Overall, higher levels of inequality were observed in the densely populated, more developed and highly commercialised area in southern Manisa, whereas a more egalitarian socioeconomic structure prevailed in the mountainous northern part that was inhabited by a high number of tribal groups. While stressing the necessity of a comparative regional approach in studying rural inequality in preindustrial societies, these results also lend support to arguments that inequality levels in these societies were positively associated with the level of market development and population.