State predation in historical perspective: the case of Ottoman müsadere practice during 1695–1839

Abstract

This paper studies the practice of Müsadere in the Ottoman Empire. Müsadere refers to the expropriation of elites—often tax farmers or administrators—by the Sultan. This practice is interesting from both political economy and economic history perspectives as the Ottoman Empire continued to increase its reliance on it during the eighteenth century, a period when European states were investing in fiscal capacity and building bureaucratic tax systems. The main argument is that Sultans faced a “political Laffer curve:” if revenue is too low, the state collapses; if fiscal extraction is too high there is a rebellion and the Sultan risks losing power. While expropriations (müsadere) allow the Sultan to keep taxes low, they are vulnerable to provoking elites to invest in fugitive rather than (more productive) captive assets. We also show that the Sultan is more prone to target politically strong elites when his fiscal capacity is low.

1 Introduction

Recent studies on fiscal regimes and the political economy of premodern states have shown major differences between the origins of the state in the Western Europe and other parts of the world (Monson and Schidel 2015a, b). From Schumpeter (1918/1991) to the pioneering works of Mathias and O’Brien (1976), O’Brien (1988), Tilly (1985, 1990) to Ormrod et al. (1999), the Western scheme of fiscal regimes has been described in four phases: (1) tribute state, (2) domain state, (3) tax state and (4) fiscal state. The first phase is marked by the extraction of tribute from conquered territories by military leaders.¹ In the Middle Ages, these leaders replaced tributes by the income from their own domain. With some exceptions like late medieval England that precociously introduced taxes, it was only during the period 1500–1800 that most European states started to turn to taxes exacted from their own subjects instead of relying on their own personal property or ‘domain’ (Ormrod et al. 1999; Hoffman 2017, pp. 1558–1561). This typology, however, is Eurocentric. It does not explain the fiscal regimes in other countries such as the Inka Empire, the Aztec Empire, and the ancient Near East and Egypt (Monson and Schidel 2015a, b, pp. 3–27). Similarly, although fiscal states characterized by bureaucratic and centralized tax regimes emerged in Europe in the nineteenth century, new research on comparative fiscal history has shown that there was not a single path to modern state (Yun-Casalilla et al. 2012).

In line with renewed scholarly interest in the rise of modern fiscal states, the evolution of fiscal institutions in the Ottoman Empire has been revisited. In particular, fiscal decentralization in the eighteenth century of which the most notable feature was the introduction of life-term tax farming is now regarded as a viable alternative to modern centralism rather than a symptom of decline (Salzmann 1993). Balla and Johnson (2009) have compared fiscal and political institutions of the Ottoman Empire to France during the early modern period, roughly 16th to 18th century. According to them, while both countries made extensive use of tax farming, the organization of tax farming evolved differently in each country. In France, tax-farming was unified in a single, quasi-private organization known as the Company of General Farms that became one of the prominent stakeholders in the French fiscal system since its establishment in 1681. The French tax-farmers were capable of collective action and could constrain the monarchy. In a sense, the French revolution was caused because of the King’s lack of flexibility due to the constraints imposed by the Company. Johnson and Koyama (2014) define the Company-controlled fiscal regime as cabal tax farming, which they consider as an intermediary institution that facilitated the transition from competitive tax farming to centralized tax collection, though much later than in England. By contrast, tax-farming under the Ottoman Empire moved towards greater decentralization during the eighteenth century. “The great paradox of French and Ottoman institutional history is that Ottoman fiscal institutions were more flexible than their Western counterparts and yet, in the long run, Western institutions proved more conducive to growth” (Balla and Johnson 2009, p. 840). The decentralized fiscal system of the Ottoman Empire was more efficacious politically in maintaining the Sultan’s power and shun a revolution but it was less efficacious economically since local elites retained a large part of the revenues (Pamuk 2014)

Another aspect of this political flexibility might be sought in specific features of the tax system of the Ottoman Empire during its expansion between the fourteenth and sixteenth centuries. Coşgel (2015) particularly explains the discriminatory nature of the Ottoman fiscal system on the basis of transaction costs related to the cost of measuring the tax base. Verifying whether the Laffer curve can be applied to the pre-modern Ottoman Empire, Coşgel (2015, p. 415) explores tax registers and finds that “the sources of revenue allocated to local government officials included a higher proportion of variable taxes than those allocated to the provincial and central treasury, indicating that the variance of the tax base affected the allocation.” The explanatory power of purely economic factors notwithstanding, the study demonstrates that political economy constraints particularly the legitimacy of the ruler and the likelihood of a successful revolt against his regime are the key elements in analyzing the tax allocation. To put it differently, a politically revisited Laffer curve is required to grasp the rationale of tax allocation under the Ottoman Empire (Coşgel 2015, pp. 415–426). In this paper, we coin that as a “political Laffer curve”. Unlike Coşgel et al. (2009) that explore the role of religious officials as a source of legitimacy, we focus on the dynamic relationship between the Sultan and fiscal elites during the long-eighteenth century.

Our focus on the long-eighteenth century (1695–1839) requires taking an additional institution into account.² In addition to decentralized tax-farming and variable types and rates of taxes according to different regions, the Ottoman fiscal system was also characterized by the routine confiscation (müsadere) from elites including office-holders and tax-farmers. This confiscatory system was used either as a punishment or more generally as post mortem seizure of the property of, mainly but not exclusively, fiscal elites. Such confiscations existed from the fifteenth century but, as we discuss below, became much more prevalent in the eighteenth century. This specificity of the Ottoman Empire was tangentially studied until recently (for a detailed study, see Arslantaş 2018). By fusing closely related issues of tax collection and confiscation from fiscal elites, this paper addresses the following questions.³ How was müsadere connected with the fiscal system under the Ottoman Empire? To what extent this predatory system could secure the power of the Sultan with regard to local elites? Was it politically efficacious in maintaining the empire under the Sultan’s rule? If so, was this politically efficacious predation also economically efficacious?

Our contribution is twofold. First, to the best of our knowledge, our study is among the first that examines the Ottoman confiscatory system using an analytic narrative method. The phrase analytic narrative captures our conviction that “theory linked to data is more powerful than either data or theory alone.” (Bates et al. 1998, p. 3). By theory, we refer to game theory particularly in its extensive form that readily accommodates the narrative form. By data, we do not mean raw and descriptive figures but stylized historical facts based on empirical observations of regularities, representing durable practices or tendencies. Analytic narrative is about “problem driven” rather than “theory driven” research (Green and Shapiro 1994). The stylized facts provide an understanding of the controversial issues that need to be modelled. They can be explained as historical context before describing the model. But models generate specific propositions that need to be revisited in light of facts. Thus historical explanation can precede as well as follow the model. Analytic narrative does not impose any sequence. “The construction of analytic narratives is an iterative process… we move back and forth between interpretation and case materials, modifying the explanation in light of the data, which itself is viewed in new ways, given our evolving understanding.” (Bates et al. 1998, p. 16). In this paper, we start by giving our historical narrative based on stylized facts such as the process of expropriation (müsadere) and its different historical phases. Although our model does not capture the richness of the reality, it generates specific propositions that are supported by extensive first-hand historical evidence. Thus, we start by case materials or descriptive historical narrative, then we build our model that provides an explanation, and we end up by facts going through theory. Indeed, one of our contributions is to show that müsadere or confiscatory institutions under the Ottoman Empire were rational though possibly not intentional. This result stems from the confrontation of our model and historical evidence. Second, this paper offers the first application of the distinction between ‘fugitive’ and ‘captive’ assets⁴ originally coined by Vahabi (2016a, b) to explain the behavior of a predatory state.

The article is organized as follows. Section 2 provides historical context of müsadere in the Ottoman Empire. After elucidating the process of müsadere, we identify three phases of its practice, namely formation, maturity and aggressive phases. To understand this process, we construct an original simple model introduced in Sect. 3. We first assume homogenous elites and find that using müsadere provides higher fiscal revenues when returns on captive assets are sufficiently high, but always impoverishes the Empire. This proposition is supported by historical narratives. The assumption of the elites’ homogeneity is relaxed in Sect. 5 to focus on the Sultan’s targeting process. We classify elites in two categories according to the strength of their links with the Sultan (‘high’ or ‘low’) and find a negative link between the level of fiscal capacity⁵ of the Sultan and the expropriation of ‘high-type’ elites, which is consistent with our period of interest. A short conclusion follows.

2 Historical context

The Ottoman Empire lasted for six centuries until World War I, ruling at the zenith of its power in the 16th century on a huge territory from the central Europe in the West to the Arab Peninsula in the East. Founded as a small principality in Western Anatolia at the turn of the fourteenth century, the Ottoman Empire heavily relied on military achievements in the fifteenth and sixteenth centuries, while building a feudalism-like system of tax collection in the sense that those who provided military and administrative services were supported by the rent produced by the peasantry (Volckart 2000). One major difference from feudalism, however, was that the holders of tımar (fiscal-administrative unit) neither owned land as private property nor held hereditary rights to it. Instead, their rights on land were restricted to revenue collection from their assigned land for a given period (Imber 2002, p. 196). This is often presented as one of the reasons why an aristocratic class never emerged in the Ottoman Empire.

It is widely accepted that the other reason was the practice of confiscation (known as müsadere in Ottoman political terminology) that can be defined as the transfer of property from the hands of a select group of people who were typically but not necessarily office-holders and tax farmers to the treasury, mostly after normal death but occasionally as sole method of punishment. Thus post-mortem confiscations were the most common, meaning that it was the deceased’s family (not the deceased himself) that played the game of confiscation with the central government. The presence of such a practice leads us to another issue, which is whether the Ottoman Empire was absolutist. It has sometimes been depicted as so especially in the European historiography (Jones 1981; Finer 1997). Yet the power of the Sultans was far from being unconstrained. We believe that these constraints, which were rather informal than in the form of formal checks and balances, can be best studied in the context of the müsadere practice that remained in effect from 1453 to 1839.

The müsadere practice passed through several phases. The first is the formation phase, i.e. 1453–1520, in which confiscations were met with reaction. It is known that Mehmed II’s (r. 1444–1446 and 1451–1481) confiscations of the fortunes of high officials and religious foundations particularly during the 1470s led to discontent among the affected groups (Özel 1999). This explains why his son Bayezid II (r. 1481–1512) pledged to return these properties during his struggle for the throne. Like many institutions of the Empire, however, müsadere was institutionalized during the sixteenth century. The period 1520–1695 should be considered as the maturity phase. In this period, the müsadere practice primarily targeted the deceased office-holders, resting on the principle that any wealth made through state grants should return to the treasury after wealth-holder’s death. From 1695 to 1839 müsadere took a more arbitrary form. The main change during this final phase, which we call aggressive phase, was not only frequency of confiscations but its targeted group and methods of enforcement. Local elites, who rose to power mostly due to privatization of fiscal resources through introduction of life-term tax farming in 1695, became the prime targets of müsadere. Under the influence of the spread of liberal ideas among Ottoman statesmen and intellectuals, müsadere was abolished in 1839 as part of Gülhane edict that started the period of modernization known as Tanzimat era (1839–1876).

The prime targets of confiscation were thus fiscal and political elites which included tax farmers, local administrators or military officials. What made them the target was the sovereign’s desire to redistribute wealth from one to another to shun the emergence of aristocratic structures. Particularly in the eighteenth century, müsadere was employed as a political tool against local elites who were organized in the form of patrimonial families that had the capacity to act independently from the central government. It is important, however, to emphasize that the Sultan discriminately confiscated. That is, not all people who could easily fall into the category of potential targets were confiscated. The determinants of this selectivity are examined by Arslantaş (2018). His findings reveal that the targets were chosen based on their bargaining power vis-à-vis the central government as well as the location and the qualities of their wealth.⁶

While our theoretical model below does some inevitable simplifications, it is essential to understand how a typical case of confiscation was conducted by the Sultan and his agents. After being informed by a local official, the Sultan decided whether to initiate and implement the confiscation. Other possible alternatives were not to confiscate or ask the family of the deceased or the punished to pay a type of inheritance tax estimated according to the expected amount of wealth. The latter was mostly open to negotiation with the family. If the decision was to initiate the process of confiscation, the next step was to send an agent (confiscator) to the premises to prepare an inventory. An inventory was simply a list of assets to be confiscated. Alongside the inventory, the confiscator kept informing the center about difficulties facing the confiscation task. After the final inventory was sent to the center and studied there, the Sultan (usually those at the office of confiscation on his behalf) had to decide between no confiscation, inheritance tax and full confiscation. In case the final decision was full confiscation, the confiscator was ordered either to transport all assets to Istanbul, or sell them locally (or a combination of these methods). This is a simplified narrative of the confiscation process. One should bear in mind that it could take more complex forms in case of a resisting family or disagreement on the ownership of assets to be confiscated.⁷

The paper focuses on the aggressive phase of the practice of müsadere, namely the long eighteenth century, 1695–1839, because of the richness of archival records concerning this period, and also since it allows us to study fiscal pressure and confiscation concomitantly. It is repeatedly claimed in Ottoman historiography that this was a period of decentralization (see, for example: İnalcık 1977).⁸ Considering fiscal pressures associated with military advancements that occurred in Europe during the sixteenth and seventeenth centuries, the Ottoman Empire delegated some power to local elites who were organized as patrimonial families. Despite their wealth and power, these families turned into the main targets of the müsadere practice. While at first sight it is curious why their power did not translate into collective action, a proper examination of their relationships with the center reveals that those families were dependent on the resources granted by the Sultan. Expectant of future grants and offices, they mostly chose not to resist. The ability of the Sultan to confiscate was also constrained for other reasons. He needed the support of local elites, who possessed military resources, for wars waged abroad, domestic rebellions and banditry activities. This explains a good deal of the multiplicity of outcomes of the confiscation process. Inheritance tax, for example, was a type of partial confiscation in which case the family’s power was only partially curbed. Selective and self-constrained confiscation, which was a method of wealth redistribution, enabled the central government to control the elite power. This led to controlled decentralization as a way of surviving through fiscal and political crisis facing the Empire.

Recent historiography of the Ottoman Empire has been emphasizing the pragmatic and flexible nature of the Ottoman governance (Pamuk 2004a, b; Fleet 2003; Coşgel 2015). These interpretations can be translated into a statement made by Frederick Lane (1958) that maximizing rule was more important than maximizing revenue. The broader literature exploring state predation that manifested itself in history in various forms such as outright confiscations, expropriations and debt repudiations has also drawn attention to factors other than pure greed while emphasizing the role of other constraints in their capacity to confiscate than constitutional constraints (see, for example, Koyama 2010). Econometric analysis of müsadere during the period 1750–1839 also lends credence to this argument by examining strategic constraints on the sovereign’s power to confiscate (Arslantaş 2018). Given these considerations, the practice of müsadere offers a fruitful ground for a theoretical analysis of historical predation under constraints.

3 The model

We consider an empire inhabited by a unitary population of identical elites⁹ ruled by a Sultan. At the beginning of the game, elite i receives one unit of resource and invests a share $c_{i} \in \left[ {0,1} \right]$ in captive assets, and $1 - c_{i}$ in fugitive ones.¹⁰ Fugitive assets possess two properties: “(1) Mobility: capable of escaping from a given state space, because they can be easily hidden or displaced geographically; this also refers to the possibility of altering political allegiance without any geographical displacement. (2) Non-confiscable: hardly subject to confiscation because (1) any attempt to transfer property rights through coercion destroys the asset or reduces its value to almost nil, and (2) the costs of confiscation are greater than the benefits of confiscation.” By contrast, “captive assets are: (1) unmovable in the sense that they are not invisible or capable of escaping from a given state space and (2) readily subject to confiscation.” (Vahabi 2016b, pp. 158–159). This distinction captures the strategy implemented by elites to avoid confiscation through müsadere from the Sultan.¹¹ Returns associated with assets differ: fugitive assets are safer but offer poorer returns than captive assets. Indeed, captive assets are at risk in the sense that they can be affected by two kinds of fiscal measures: their returns are taxed at a rate $\tau \in \left[ {0,1} \right]$, and the Sultan captures an total amount $M > 0$ of captive assets initially detained by $n \in \left[ {0,1} \right]$ elites. If $n = 1$ (resp. $n = 0$), then all (resp. no) elites are affected by müsadere. The practice of müsadere does not consist of a pure confiscation, since there was a partial redistribution of captured assets that typically took the form of public auctions. After one’s wealth was decided to be confiscated, there were three modes of transfer: (1) transportation of all assets to Istanbul in kind, (2) auctioning all assets on premise and transferring revenues in cash and (3) a mix of these modes. In the matter of which mode to choose, the central government followed the principle of cost-minimization, which is evident from the observation that auctions were more often held in more distant regions (transportation costs increased with distance).

Based on a sample of five auctions, Fig. 1 shows that many bidders present at auctions were again office-holders. It displays how much each occupation and title group paid to buy the auctioned assets. Ultimately the administrative officials paid some 44% of the total payment. The categories of ‘no occupation’ and ‘no title’ reflect those not specifying title or occupation. However, the fact that most of the bidders included in the category of ‘no occupation’ are titled as either efendi or ağa¹² means that they likely fall in the occupation categories of either government officials or local elites too. It is striking to see that two purchasers of the assets of Ali Esad Paşa and one of those of Salih Paşa were the officials of the Bureau of Confiscation (located in Istanbul) themselves. The former two purchased some textiles.¹³ Taken together, these lists of purchasers demonstrate that when auctioned, confiscated properties were mostly bought by other elites. This reflects their satisfaction in paying for confiscated assets without much care for the fact that their families were also on the possible path of being denied the inheritance of their wealth when the time came. These kinds of public auctions and all forms of redistribution of captive assets is measured by $\left( {1 - \delta } \right)$, where $\delta \in \left[ {0,1} \right]$ is the share of the captive assets captured using müsadere that is directly appropriated by the Sultan.

Fig. 1

Occupational and Title Distribution of Bidders of Confiscation Auctions in a Chosen Sample (in piasters (kuruş)) Kuruş was the Ottoman monetary unit that emerged in the eighteenth century replacing akçe. One British pound sterling was equal to 8 kuruş in 1800 (Yaycıoğlu 2016). Source: D.BŞM.MHF.d 13289, D.BŞM.MHF.d 13327, D.BŞM.MHF.d 13454, D.BŞM.MHF.d 13341, D.BŞM.MHF.d 13465. Notes: Those cases included in the data presented are Mustafa Mazhar Efendi (Head of Finance Department, 1818), Halil Efendi (Mufti, 1821), İsmail Paşa (Governor of Niğde, 1822), Ali Esad Paşa (Governor of Alanya, 1828), Salih Paşa (Governor of Damascus, 1828)

Definition 1

(Fiscality)

• 1. We call taxation, $\tau$, a puncture of the return on assets. All returns, on fugitive and captive assets, are appropriable.

• 2. We call müsadere, $M$, the confiscation of assets of $n$ elites. Only captive assets may be confiscated.

Elites’ payoffs depend on the probability to be affected by müsadere, ${\mathbb{P}}_{i} \in \left[ {0,1} \right]$. Let $r_{c}$ and $r_{ - c}$ be respectively the return on captive and fugitive assets, elite i receives $U_{i}^{1} = \left( {1 - c_{i} } \right) + \left( {1 - \tau } \right)\left[ {r_{ - c} \left( {1 - c_{i} } \right) + r_{c} c_{i} } \right]$, if he is affected by müsadere. On the other hand, his payoffs equal $U_{i}^{0} = 1 + \left( {1 - \tau } \right)\left[ {r_{ - c} \left( {1 - c_{i} } \right) + r_{c} c_{i} } \right] + \left( {1 - \delta } \right)M/(1 - n)$ when he avoids confiscation. Payoffs of representative elite i are given by:

$$U_{i} = {\mathbb{P}}_{i} U_{i}^{1} + \left( {1 - {\mathbb{P}}_{i} } \right) U_{i}^{0}$$

(1)

The Sultan chooses the level of taxation and müsadere which together defines the fiscal policy in the Empire. It should be noted that the amount of müsadere, M, is such as:

$$M = c_{i} n$$

(2)

Indeed, since elites are identical, they invest the same level of resources in captive assets, $c_{i}$. Fiscal rent is defined as follows:

$$T = \tau \left[ {r_{c} c_{i} + r_{ - c} \left( {1 - c_{i} } \right)} \right] + \delta M - sn^{2}$$

(3)

Expression 3 describes the composition of the fiscal rent. The first term corresponds to the revenues of the sultan collected through tax. The last two parts are related to the use of müsadere: $\delta M = \delta nc_{i}$ is the benefits obtained by the Sultan and $sn^{2}$ are convex costs associated with the expropriation of the n elites with $s \ge 1$. Indeed, enforcing müsadere requires the employment of agents and transportation of assets both incurring high costs. In addition to the afore-mentioned agency costs, there were also transportation costs associated with müsadere. If the assets were not auctioned on the spot, they were typically transported to Istanbul, or occasionally to the warzone. The more wealth seized, the more sophisticated the transportation means had to be supporting convex costs assumption.

The fiscal policy is described by the couple $\left\{ {\tau ,n} \right\}$ that pursued two hierarchically ranked objectives: (1) maximizing the Sultan’s probability to stay in power, $p \in \left[ {0,1} \right]$ and (2) extracting the highest fiscal rent $T \ge 0$ from the elites subject to survival objective.

We also build the following indicator of fiscal pressure:

$$f = 1 - \frac{U}{W}$$

(4)

where $U = \mathop \smallint \nolimits_{i = 0}^{1} U_{i} di = U_{i}$ because elites are identical, and $W = \left( {1 + r_{ - c} } \right)\left( {1 - c_{i} } \right) + r_{c} c_{i} + \left( {1 - n} \right)c_{i} + \left( {1 - \delta } \right)M$ corresponds to the maximal payoffs of the population with all assets available in the economy. By definition, if both levels of taxation and confiscation are high, then fiscal pressure depicted in expression (4) is also high. By contrast, if $U = W$ (i.e. $\tau = n = 0$) then $f = 0$.

We propose a sequential game with complete information in which representative elite i plays first and chooses the share of the resource devoted to captive assets, $c_{i}$, that maximizes $U_{i}$ (stage 1). The Sultan observes the outcome of the first stage and decides the fiscal policy $\left\{ {\tau ,n} \right\}$ that maximizes its probability, p, to stay in power and, his fiscal rent, T (stage 2). The timeline of the game is summarized in Fig. 2.¹⁴

Fig. 2

Timeline of the game

4 Resolution

Since we have a two-stage game with complete information, solutions for the model are determined iteratively, by backward induction.

4.1 Stage 2: Fiscal rules enacted by the Sultan

For tractability purposes, we rely on two assumptions.

Assumption 1

Returns on fugitive assets are nil $\left( {r_{ - c} = 0} \right)$.

Assumption 2

There exists a ‘political Laffer curve’ depicting an inverted U relationship between the rate of fiscal pressure, f, and the Sultan’s probability of staying in power, p.

A1 enables better tractability by only considering returns on captive assets, $r_{c} > 0$.¹⁵ The existence of a ‘political Laffer curve’¹⁶ (hereafter PLC) assumed in A2 is based on the idea that neither a very low nor a very high level of fiscal pressure allows the Sultan to remain in power. In particular very low level fiscal revenue does not allow the creation of effective military protection forces.¹⁷ Consequently, the regime would be vulnerable to invasions from competing neighbors. In stark contrast, by imposing too high fiscal pressure on his subjects, he would provoke internal protests (see Fig. 3).

Fig. 3

‘Political’ Laffer curve

In order to find an analytic solution, we model the probability to stay in power p as follows:

$$p = {{\upbeta }}f^{\lambda } \left( {1 - f} \right)^{1 - \lambda }$$

(5)

where β is a positive parameter taking into account the legitimacy of the sultan: when β = 0, he will loose with probability one, regardless of the level of fiscal pressure. High values of β may be due to high legitimacy shunning the emergence of internal or external contestants. $\lambda \in \left( {0,1} \right)$ is a weight reflecting Sultan’s concerns regarding external threats relative to internal ones. In particular, a $\lambda$ close to 0 means that external threats are negligible compared to internal threats. Consequently, he would focus on appeasing domestic contestations by choosing a low rate of fiscal pressure. On the contrary, when $\lambda$ tends to 1, the need for military protection of frontiers against invaders is higher than the risk of internal rebellion that would be triggered by a high level of fiscal pressure.

The first objective of the Sultan being to stay in power, he chooses the level of fiscal pressure solving the following problem:

$$\mathop {\hbox{max} }\limits_{{\left\{ {\text{f}} \right\}}} p = {\upbeta}f^{\lambda } \left( {1 - f} \right)^{1 - \lambda }$$

(6)

The solution of the maximization problem (6) is $f^{*} = \lambda$.¹⁸ Therefore, the choice of the level of fiscal pressure fully depends on the relative importance of external threats compared to internal ones faced by the Sultan. Once the Sultan secured his power, he then maximizes his fiscal rent. Consequently, the Sultan’s global maximization problem in the second stage of the game is:

$$\mathop {\hbox{max} }\limits_{{\left\{ {{\uptau},{\text{n}}} \right\}}} T\; s.t. \;f = \lambda$$

(7)

The couple $\left\{ {\tau ,n} \right\}$ that solves the problem in (7) is:

$$\left\{ {\lambda + \frac{{\lambda \left[ {2s - c_{i}^{2} \delta^{2} \left( {1 - \lambda } \right)} \right]}}{{2sr_{c} c_{i} }},\frac{{c_{i} \delta \left( {1 - \lambda } \right)}}{2s}} \right\}$$

(8)

We can deduct from (8) that a high level of confiscation or müsadere,$\delta$, allows the Sultan to reduce fiscal taxes. As recently evidenced, the Ottoman Empire collected the least amount of taxes per capita in early modern Europe together with Poland (Karaman and Pamuk 2010). On the one hand, this was related to the vast size of the empire and the heterogeneity of its population, leading to administrative inefficiencies. On the other hand, the existence of müsadere could refund the treasury through confiscation from highly wealthy individuals.¹⁹ In a way, müsadere provided the central government with immediate, though irregular, revenue (Cezar 1986). This substitutability between tax and müsadere can be understood in light of the following narrative. As fiscal entrepreneurs, tax farmers signed a contract in either primary or secondary markets of tax farming, making them liable to make prefixed payments to be followed by annuities. The revenue they collected above that amount was their profit. The natural outcome of this process was that many entrepreneurs did their best to squeeze tax payers. From the central government’s point of view, oppressing the subjects could permanently impair the tax base by lowering tax compliance through damaging the ruler’s legitimacy. Consequently, in order to avoid over-taxation of the population, the Sultan chose to confiscate the tax farmer’s wealth instead.²⁰

4.2 Stage 1: Elites’ choice of assets

We now address the elites’ choice regarding their investments in captive or fugitive assets. Since elites are assumed to be identical, the Sultan randomly targets them i.e. ${\mathbb{P}}_{i} = n$. Then, by using (8) and (2), we can rewrite the payoffs of the representative elite as follows:

$$U_{i} = \frac{{\left( {1 - \lambda } \right)\left[ {2s\left( {r_{c} c_{i} + 1} \right) - c_{i}^{2} \delta^{2} \left( {1 - \lambda } \right)} \right]}}{2s}$$

(1′)

Elite i solves the following problem:

$$\mathop {\hbox{max} }\limits_{{\left\{ {{\text{c}}_{\text{i}} } \right\}}} U_{i}$$

(9)

Based on (1′), the level of captive assets that solves the problem in (9) is²¹

$$c_{i}^{*} = \frac{{r_{c} s}}{{\delta^{2} \left( {1 - \lambda } \right)}}$$

(10)

It is noteworthy that the level of captive assets positively depends on their returns, $r_{c}$ and on the cost of expropriation, $s$. Finally, (10) exhibits a negative links between the level of appropriation of müsadere, $\delta$: elite i tends to prefer fugitive assets if there is a low redistribution of müsadere.

From (8) and (10) we obtain the following equilibrium analytical values for tax rate and müsadere:

$$\left\{ {\frac{{\left[ {r_{c}^{ 2} s + 2\delta^{2} \left( {1 - \lambda } \right)} \right]\lambda }}{{2r_{c}^{ 2} s}}, \frac{{r_{c} }}{2\delta }} \right\}$$

(11)

One can easily deduct from (2) and (11) that the total amount of captive assets captured through müsadere is $M^{*} = \frac{{r_{c}^{ 2} }}{{2\delta^{3} \left( {1 - \lambda } \right)}}$. This relation emphasizes a positive relation between $M^{*}$ and $\lambda$. In particular, it means that if external threats increase, the Sultan would extract more resources from müsadere. This is consistent with empirical evidence: greater number of wars fought abroad required a heftier treasury, and müsadere was a quick way of extracting that revenue (Cezar 1986). Moreover, high values of the $r_{c}$, encourages the Sultan to confiscate captive assets. It is linked with an abundance effect: if the representative elite i invests intensively in captive assets, it would trigger higher level of müsadere.

The following Lemma deals with the case in which the Sultan cannot use müsadere. For the sake of clarity, we use the subscript r to denote this situation.

Lemma 1

(Without müsadere). If the Sultan cannot use müsadere, i.e.$n = M = 0$, he taxes elite$i$at a rate$\tau^{{*}^{r}} = \lambda \left( {1 + \frac{1}{{r_{c} }}} \right) \ge \tau^{*}$and elite$i$plays$c_{i}^{{*}^{r}} = 1$.

Proof

In the second stage of the game, it can easily be shown that $\tau^{r} = \lambda \left( {1 + \frac{1}{{c_{i} r_{c} }}} \right)$. In the first stage, elite $i$ maximizes (1′), where $\frac{{\partial U_{i} }}{{\partial c_{i} }} = r_{c} \left( {1 - \lambda } \right) > 0.$ Consequently, elite $i$ is always incentivized to devote all his resources to captive assets such that $c_{i}^{{*}^{r}} = 1$. One can easily deduct:

$$\tau^{{*}^{r}} = \lambda \left( {1 + \frac{1}{{r_{c} }}} \right)$$

(12)

Moreover, $\lambda \left( {1 + \frac{1}{{r_{c} }}} \right) - \frac{{\left[ {r_{c}^{ 2} s + \delta^{2} \left( {1 - \lambda } \right)} \right]\lambda }}{{2r_{c}^{ 2} s}} = \frac{{\lambda \delta^{2} \left( {1 - \lambda } \right)}}{{r_{c} s}} \ge 0$. Using (10) and (11) it involves: $\tau^{{*}^{r}} \ge \tau^{*}$.

Considering the present article’s goal to provide an analytic narrative of the Ottoman müsadere practice during 1695–1839, we only focus on plausible cases, i.e. interior solutions.

Definition 2

(Interior solution) An interior solution is characterized by $n^{*} < 1$, $c_{i}^{*} < 1, \tau^{*} < 1$ and $\tau^{{*}^{r}} < 1$, simultaneously.

Focusing on cases depicted in Definition 2²² offers two advantages. First, it allows to exclude extreme cases such as a confiscation affecting more than the whole population of elites $\left( {n > 1} \right)$, or an elite $i$’s investment in captive assets superior to his initial wealth $\left( {c_{i} > 1} \right)$. Second, it ensures that the analytical resolution of the model is tractable by avoiding the existence of multiple subcases. By restraining the resolution to interior solutions, we are able to formulate the following proposition.

Proposition 1

(Müsadere effect) Considering the definition of interior solution, using müsadere ensures to collect a higher fiscal rent (T) if and only if$r_{c}$is superior to a threshold$\tilde{r}_{c} .$However, müsadere always generates lower maximal payoffs of the elites with all assets available in the economy$\left( W \right)$.

Proof

Proving Proposition 1 results in comparing two situations with and without müsadere. Let $\Delta W = W^{*} - W^{{*}^{r}}$ and $\Delta T = T^{*} - T^{{*}^{r}}$ be respectively the difference of maximal payoffs of the elites with all assets available in the economy, and the difference of fiscal rent between the two situations. First, we have:

$$\Delta W = r_{c} \left( {c_{i}^{*} - c_{i}^{{*}^{r}} } \right) - \delta n^{*} c_{i}^{*}$$

(13)

It can be easily noted that $\Delta W \le 0,$ because $c_{i}^{*} < 1 = c_{i}^{{*}^{r}}$.

Second, using (10), (11) and (12) we can express $\Delta T$ as follows:

$$\Delta T = \frac{{r_{c} \left[ {r_{c} s\left( {3\lambda + 1} \right) - 4\lambda \delta^{2} \left( {1 - \lambda } \right)} \right]}}{{4\delta^{2} \left( {1 - \lambda } \right)}}$$

(14)

It can be shown that $\Delta T \ge 0$ if and only if $r_{c} \ge \tilde{r}_{c} = \frac{{4\lambda \delta^{2} \left( {1 - \lambda } \right)}}{{s\left( {3\lambda + 1} \right)}}$.

Proposition 1 compares a situation in which the Sultan uses müsadere with a situation in which he only has the ability to choose the tax rate, $\tau$. The first implication of Proposition 1 is that the use of müsadere reduces maximal payoffs of the population with all assets available in the economy, implying an impoverishment of the empire on a long run perspective. First, when elites’ productivity is low, we see that $\Delta T \le 0$ meaning that the Sultan would extract higher fiscal rent if $n = 0$. Generally speaking, müsadere provides higher fiscal rent as soon as $r_{c} > \tilde{r}_{c}$, however it generates lower incentives for elites to invest in captive assets. Therefore, in a static framework, müsadere appears more appealing than a fiscal system only based on $\tau$. However, in a more dynamic perspective, it impoverishes the Empire ($\Delta W < 0)$, undermining seriously the survival of the Sultan in the long run.²³

In Fig. 4, we propose a numerical example in which all possible cases (including corner solutions) are explored. Here, the Sultan is exposed to a high level of external threats $\left( {\lambda = 0.8} \right)$ and redistributes 10% of assets seized through müsadere$\left( {\delta = 0.9} \right)$. Few elements are noteworthy. First, we observe that using müsadere always tends to impoverish the empire $\left( {\Delta W < 0} \right)$, which is in line with Proposition 1. Second, the Sultan obtains higher fiscal rent for intermediate values of $r_{c}$. For extreme ones, a fiscal system only relying on a tax over the returns on assets provides higher fiscal rent. In particular, $\Delta T$ is first negative, then positive corroborating Proposition 1. However, Fig. 4 also shows that if $r_{c}$ is high enough, müsadere provides lower fiscal rent. The intuition behind this result is as follows. In the model the Sultan is incentivized to use more müsadere$\left( n \right)$ and less tax $(\tau$) to maximize his probability of staying in power. But, for high returns on captive assets, this strategy is less fruitful than following an alternative strategy of only taxing returns at a higher rate $\left( {\tau^{{*}^{r}} \ge \tau^{*} } \right)$.²⁴

Fig. 4

The rationale of müsadere (numerical example). Note: $\lambda = 0.8$, $s = 1.2$ and $\delta = 0.9.$ Computational details are provided in Annex 2

5 Targeting elites

In the previous section, elites are assumed to be homogenous. Consequently, the Sultan randomly targets elites and the model cannot provide insights regarding the choice of expropriated elites. To circumvent this shortcoming, we consider a new setting only focusing on the decision of the Sultan regarding the müsadere practice. We assume that the Sultan wants to expropriate a given number of elites, $\bar{n} \in \left( {0,1} \right)$ and that elites’ population is formed of two groups: $h \in \left( {0,1} \right)$ elites characterized by a high symbiotic relationship with the Sultan and $l = 1 - h$ elites with low one.²⁵ A high level of symbiotic relationship may be due to three elements: military strength of an elite $i$, the economic reliance of the Sultan with respect to the elite, and political closeness.²⁶ This symbiotic relationship involves two implications with regard to the practice of müsadere. First, expropriating a h-type elite is more costly $(s_{h} \ge s_{l}$). Second, only h-type elites are able to claim for a share of the fiscal rent of the Sultan, $T$ (i.e. they have a bargaining power on the sharing of the fiscal rent). Consequently, by reducing their number, the Sultan secures a larger share of $T$. In this last section we are interested in exploring the rationale of the Sultan to expropriate h-type elites ($n_{h}$) rather than l-type ones ($n_{l}$), with $\overline{n} = n_{h} + n_{l}$. In order to stay tractable, we rely on a few simplifications. Firstly, we normalize to one the l-type elites’ expropriation costs such as $s_{l} = 1$ and $s_{h} \ge 1.$ Secondly, we rewrite the fiscal rent as follows:

$$T = {\varGamma } - s_{h} \left( {n_{h} } \right)^{2} - \left( {n_{l} } \right)^{2}$$

(3′)

where ${\varGamma } = \tau Y + \delta M$ is supposed to be known by the Sultan and not affected by the choice of targeted elites. This last point means that, contrary to the model developed in Sect. 4, we disregard the choice of assets made by elites and we focus on the choice of the Sultan regarding the practice of müsadere.

Fiscal rent being contestable by h-type elites, the share left for the Sultan, $T_{s}$, is given by:

$$T_{S} = \alpha T$$

(15)

where $\alpha \in \left[ {0,1} \right]$ is the level of bargaining power detained by the Sultan over the h-type elites. When it takes the value $1$ he captures the whole fiscal rent, and $\alpha = 0$ refers to a situation in which the h-type elites enjoy $T$. The Sultan maximizes his share of the fiscal rent $T_{S}$, solving the following problem:

$$\mathop {\hbox{max} }\limits_{{\left\{ {n_{h} } \right\}}} T_{S}\; s.t. \; n_{l} = \overline{n} - n_{h}$$

(16)

Proposition 2.a

(Non-contestable fiscal rent). When$\alpha = 1$the Sultan

• equally targets elites if$s_{l} = s_{h}$;

• targets more intensively the l-type elites otherwise.

Proof

For $\alpha = 1$, $T_{S} \equiv T$ for all $n_{h}$. Consequently, the value of $n_{h}$ that solves the problem in (16) is $\frac{\overline{n}}{{s_{h} + 1}}$. Considering that $n_{l} = \overline{n} - n_{h}$, one can easily see that $n_{l} > n_{h}$ when $s_{h} > s_{l}$, and $n_{l} = n_{h} = \frac{{\bar{n}}}{2}$ when $s_{h} = s_{l}$.

In Proposition 2.a, the only source of elites’ heterogeneity is related to the cost of expropriation borne by the Sultan using müsadere. As it is costlier to seize the wealth of h-type elites, it appears straightforward to observe that the Sultan targets more severely the l-type elites when the fiscal power is fully centralized. However, when the Sultan is not able to fully secure fiscal rent $\left( {\alpha < 1} \right)$, targeting h-type elites becomes more attractive because it also reduces the number of claimers. In other words, we consider a situation in which $\partial \alpha /\partial n_{h} \ge 0$. More specifically we adopt the following affine formulation:

$$\alpha = n_{h} + \alpha_{0}$$

(17)

where $\alpha_{0}$ is the share of $T$ secured by the Sultan when he does not use müsadere on h-type elites. As a consequence, $\alpha_{0}$ may be seen as a proxy of the level of the Sultan’s fiscal capacity. Indeed, high (resp. low) value of $\alpha_{0}$ means that he is able to secure an important (resp. low) share of fiscal rent without expropriating h-type elites. Considering (17) and the problem in (16), we are now able to formulate our next proposition:

Proposition 2.b

(Fiscal capacity and elites’ targeting). When$\alpha = n_{h} + \alpha_{0} < 1$, The lower the fiscal capacity of the Sultan, the more h-type elites are targeted.

Proof

See Annex 2.

Figure 5 presents partial evidence to the propositions 2.a. and 2.b. Given that the size of wealth was positively correlated with high bargaining power, high average value of assets confiscated in a given 10-year period signifies the targeting of h-type elites. The period 1770–1780, marked by a sharp decline in the average value of confiscated assets per person, was characterized by the beginning of fiscal and political centralization that continued until the end of the period for which data are available. By the last quarter of the eighteenth century, local elites had become too powerful primarily because they exchanged their military power in return for more and more economic opportunities in the form of offices and grants from the government. As a result, they were retaining higher shares of tax revenues before reaching the central treasury. To inject more revenue into the center, Selim III (r. 1789–1807) and Mahmud II (r. 1808–1839), in particular, embarked on a centralization project, which required curbing the power of provincial elites and eventual elimination in the 1820s through confiscation and other methods. Figure 5 proves that during the period when the government started investing in state capacity, which resulted in greater fiscal capacity, confiscations were directed at less wealthy elites. In other words, centralization led to the targeting of l-type elites.

Fig. 5

Average value of confiscated assets per person (in tons of silver). Source: Extensive number of archival sources used behind this graph can be reached at (Arslantaş 2018)

6 Conclusion

Describing the pristine states, Scott (2017, p. 14) defined a state as “a tax collector”, and Usher (1992) explained that taxation is a form of predation. The ideology of state as provider of public goods and services disguises the relationship between taxation and predation. By contrast, confiscations stand as overt forms of predation. The advantage of the specific fiscal system of the Ottoman Empire is that it lends credence to the idea that there is not such a sharp contrast between taxation and confiscation. In fact, müsadere or the routine confiscation from elites including office-holders and tax-farmers was part of the Sultan’s fiscal system. As previously emphasized, müsadere has not been studied sufficiently until recently. One of the originality of this paper is its focus on this specific system and exploring the relationship between fiscal system and confiscations.

We have already mentioned that according to Karaman and Pamuk (2010) the Ottoman Empire collected the second least amount of taxes per capita in early modern Europe. Bang (2015) generalized this finding, claiming that large empires such as Roman Empire, imperial China and the Ottoman Empire imposed low per capita taxes. Their centrally controlled administrations were tiny, and they relied on local elites to support them financially and militarily during critical moments notably in warfare. The main issue was to find methods to eschew siphoning of the revenue by local elites that could be mobilized for the central state. The political economy of fiscal regimes illustrates different trajectories pursued in different empires. These trajectories were principally determined by political factors.

Following analytic narratives methodology, our model demonstrates that müsadere was detrimental in terms of economic performance $\left( W \right)$ compared with a purely fiscal system. Historically speaking, confiscations were positively related to external wars. It can be concluded that müsadere impoverished the Empire in long run. However, müsadere had a specific political function: it allowed a controlled decentralization by the sultan. In other words, its importance was not with respect to revenue-maximizing but rather maximizing predatory rules. Finally, the paper briefly tackled the question of the müsadere targeting process. Using a very simple framework, we found a negative relationship between the ability of the Sultan to secure his fiscal rent and the expropriation of h-type elites (those able to claim for the sharing of the fiscal rent). The results of our case study, which are consistent with qualitative and quantitative evidence from the Ottoman archives, explain the functional aspect of the müsadere practice that had thus far remained ambiguous in the Ottoman historiography.

Further comparative studies between the Ottoman Empire and other types of empires will cast light on the impact of different types of state predation on economic performance. Let us draw some broader insights from our analysis by comparing it to a recent work on the case of China. Ma and Rubin (2019) also explored the relationship between taxation and confiscation though with the aim of explaining why absolutism in early modern China (and in other absolutist regimes) generated much less tax revenue than the leading Western European states at the time. The authors’ findings can be summarized as follows. They mention two paradoxes of power. The first paradox is the difficulty of incentivizing a fiscal agent to collect taxes under the threat of arbitrary confiscation, which is a characteristic of absolutist rulers. The absolutist ruler can credibly commit to refrain from confiscation only when administrative capacity is low and thus monitoring costs are high. The authors then introduce the second paradox, which is that poor monitoring technology (low administrative capacity) would motivate tax agents to collect too much tax from the masses, which in turn may provoke subjects’ revolts. The paper argues that as a solution Chinese rulers opted for extracting less revenue for the center and offered low wages to agents while letting them to keep some extra-legal taxation for their own benefits. The authors claim that although they were motivated by the case of China, their findings could be extended to other absolutist regimes such as the Ottoman Empire. Our approach is admittedly different. We departed from the very existence of many confiscations undertaken against the wealth of tax-collecting families. One can easily see from our narrative that Ottoman rulers did not need to refrain from confiscations. During the period under question, agents were buying the right to collect taxes from a certain tax unit for lifetime. Any surplus they could make after making a lump sum payment belonged to them. Ottoman rulers also wanted to prevent overexploitation of the masses but unlike their Chinese counterparts they did so through the very threat of confiscation presumably because the instrument of low wages was unavailable given the fiscal structure at the time. The upshot was similar though. The Ottoman Empire indeed settled for low taxation but endeavored at its best to compensate this weakness via the tool of selective and flexible confiscation. The present paper’s analysis of müsadere contributes to a better understanding of the specificities of the Ottoman Empire compared to the European as well as Chinese ones.

Appendix

1.1 Annex 1: Game represented in extensive form

See Fig. 6.

Fig. 6

Extensive form of the müsadere game

1.2 Annex 2: Computational details

1.2.1 Positive returns on fugitive assets $\varvec{r}_{{ - \varvec{c}}} > 0$

We adopt the exact same methodology that we present in the paper. Our results are:

$$c_{i}^{*} = \frac{{\left( {r_{c} - r_{ - c} } \right)s}}{{\delta^{2} \left( {1 - \lambda } \right)}}$$

(18)

The analytical values for tax rate and müsadere that are solutions to the Sultan’s program in (7)

$$\left\{ {\frac{{\left[ {s\left( {r_{c} - r_{ - c} } \right)^{2} + 2\delta^{2} \left( {1 - \lambda } \right)\left( {1 + r_{ - c} } \right) } \right]\lambda }}{{r_{ - c} \delta^{2} \left( {1 - \lambda } \right) + s\left( {r_{c} - r_{ - c} } \right)^{2} }}, \frac{{\left( {r_{c} - r_{ - c} } \right)}}{2\delta }} \right\}$$

(19)

1.2.2 On the numerical application

All numerical values are approximated at $10^{ - 2}$. According to (10), (11) and (12), we have: $c_{i}^{*} = 7.41r_{c}$, $n^{*} = 0.56r_{c}$, $\tau^{*} = \frac{{0.33\left( {1.2r_{c}^{2} + 0.32} \right)}}{{r_{c}^{2} }}$ and $\tau^{{*}^{r}} = 0.8 + 0.8r_{c}$. When values of parameters are superior or equal to 1, we systematically normalized them to 1 in order to avoid inconsistent situations e.g. the Sultan taxes more than 100% of the revenues of elite $\left( {\tau > 1} \right)$. We can distinguish three cases for $r_{c} \in \left[ {0,0.7} \right]$.²⁷ First, when $r_{c} \in \left[ {0, 0.13} \right)$, $\tau^{*} = \tau^{{*}^{r}} = 1, c_{i}^{*} < 1$ and $n^{*} < 1$. We have $\Delta W = 3.68r_{c} \left( {r_{c} - 0.27} \right)$ and $\Delta T = 10.77r_{c} \left( {r_{c} - 0.09} \right).$ Second, when $r_{c} \in \left[ {0.13,0.42} \right)$, $\tau^{*} = \tau^{{*}^{r}} = c_{i}^{*} = 1$ and $n^{*} < 1$. One can show that $\Delta W = - 0.5r_{c}$ and $\Delta T = - 0.38r_{c} \left( {r_{c} - 1.34} \right)$. Last, when $r_{c} \in \left[ {0.42,0.7} \right]$, $\tau^{{*}^{r}} = c_{i}^{*} = 1, \tau^{*} < 1$ and $n^{*} < 1$. Therefore, $\Delta W = - 0.5r_{c}$ and $\Delta T = \frac{{ - 0.38r_{c} \left( {r_{c} - 0.58} \right)\left( {rc^{2} + 0.84r_{c} + 0.49} \right)}}{{r_{c} }}$.

Proof of Proposition 2.b.

FOC of the problem in (16) is given by:

$$\frac{{\partial {\text{T}}_{\text{s}} }}{{\partial n_{h} }} = {\varGamma } - s_{h} \left( {n_{h} } \right)^{2} - \left( {\overline{n} - n_{h} } \right)^{2} + \left( {n_{h} + \alpha_{0} } \right)\left( {2\overline{n} - 2n_{h} \left( {1 + s_{h} } \right)} \right) = 0$$

(20)

Let $n_{h}^{*}$ be the value of $n_{h}$ verifying (A3), we have²⁸:

$$n_{h}^{*} = \frac{1}{{3\left( {s_{h} + 1} \right)}}\left[ {2\overline{n} - \alpha_{0} \left( {1 + s_{h} } \right) + \sqrt {\left( {\alpha_{0} (1 + s_{h} ) + \overline{n} } \right)^{2} + 3\left( {{\varGamma }\left( {s_{h} + 1} \right) - s_{h} \overline{n}^{2} } \right)} } \right]$$

(21)

We need to ensure that $\left( {\alpha_{0} (1 + s_{h} ) + \overline{n} } \right)^{2} + 3\left( {{\varGamma }\left( {s_{h} + 1} \right) - s_{h} \overline{n}^{2} } \right) \ge 0$. It is always verified if $s_{h} \left( {{\varGamma } - \overline{n}^{2} } \right) \ge 0$ which holds by definition because we only consider the case $T \ge 0$.

Considering $\alpha_{h}$ as a proxy of the fiscal capacity of the Sultan, we prove the Proposition 2.b. by showing that $\partial n_{h}^{*} /\partial \alpha_{0} \ge 0$. To ease tractability, let $\lambda = \left( {\alpha_{0} (1 + s_{h} ) + \overline{n} } \right)^{2} + 3\left( {{\varGamma }\left( {s_{h} + 1} \right) - s_{h} \overline{n}^{2} } \right) \ge 0$, then

$$\frac{{\partial n_{h}^{*} }}{{\partial \alpha_{0} }} = \frac{1}{3\sqrt \lambda }\left( {\alpha_{0} s_{h} + \alpha_{0} + \overline{n} - \sqrt \lambda } \right)$$

We have $\lambda - \left( {\alpha_{0} s_{h} + \alpha_{0} + \overline{n} } \right)^{2} = 3\left( {s_{h} \left( {{\varGamma } - \overline{n}^{2} } \right) + {\varGamma }} \right) \ge 0$, therefore $\frac{{\partial n_{h}^{*} }}{{\partial \alpha_{0} }} \le 0$. In other words, the lower the fiscal capacity ($\alpha_{0}$) of the Sultan is, the more h-type elites are targeted.

1.3 Annex 3: Archival Sources

Note: All archival documents used in this paper were purchased from the Prime Ministry Ottoman Archives in Istanbul. The following abbreviations refer to the name of the catalogue they are located, while the number part shows the exact location of the document in a given catalogue.

Abbreviations:

AE.III.Mustafa: Ali Emiri 3. Mustafa

AE.SABH.I: Ali Emiri 1. Abdülhamid

C.DH: Cevdet Dahiliye

C.ML: Cevdet Maliye

D.BŞM.MHF.d: Başmuhasebe Muhallefat Defterleri

Sources of Quantitative Analysis:

Ali Emiri Sultan Abdulhamid I

304/20407

Ali Emiri Sultan Mustafa III

26/1797, 193/15240

Ali Emiri Sultan Selim III

5/244, 70/4188, 144/8720, 174/10429, 189/11380, 221/12949, 269/15545, 357/20462, 361/20660, 410/23604

Başmuhasebe Kalemi Muhallefat Halifeliği Defterleri (D.BŞM.MHF.d...)

5618, 12442, 12456, 12571, 12585, 12586, 12587, 12590, 12591, 12592, 12593, 12594, 12597, 12603, 12606, 12609, 12612, 12617, 12618, 12619, 12620, 12629, 12634, 12635, 12636, 12637, 12639, 12640, 12652, 12653, 12654, 12656, 12657, 12659, 12660, 12662, 12663, 12664, 12666, 12668, 12669, 12670, 12671, 12673, 12674, 12675, 12676, 12677, 12678, 12679, 12680, 12681, 12682, 12683, 12684, 12685, 12686, 12690, 12691, 12694, 12695, 12696, 12697, 12698, 12699, 12701, 12702, 12703, 12704, 12705, 12706, 12708, 12710, 12711, 12712, 12713, 12714, 12717, 12719, 12720, 12721, 12722, 12723, 12725, 12726, 12727, 12728, 12730, 12731, 12732, 12733, 12734, 12735, 12736, 12739, 12746, 12748, 12749, 12750, 12752, 12754, 12756, 12757, 12758, 12759, 12761, 12763, 12766, 12769, 12771, 12772, 12774, 12775, 12777, 12778, 12779, 12781, 12782, 12784, 12785, 12786, 12788, 12789, 12790, 12793, 12794, 12796, 12797, 12799, 12800, 12801, 12802, 12803, 12805, 12806, 12807, 12808, 12809, 12810, 12812, 12813, 12814, 12815, 12816, 12817, 12818, 12819, 12822, 12823, 12826, 12828, 12829, 12830, 12831, 12832, 12836, 12837, 12838, 12839, 12840, 12841, 12842, 12843, 12844, 12846, 12847, 12848, 12849, 12852, 12853, 12854, 12855, 12856, 12858, 12859, 12860, 12861, 12862, 12863, 12864, 12865, 12866, 12867, 12868, 12869, 12870, 12872, 12873, 12876, 12878, 12882, 12883, 12884, 12885, 12886, 12887, 12888, 12889, 12890, 12892, 12894, 12895, 12896, 12898, 12899, 12900, 12901, 12902, 12906, 12907, 12908, 12909, 12910, 12912, 12913, 12914, 12915, 12917, 12918, 12919, 12920, 12921, 12922, 12923, 12924, 12925, 12926, 12928, 12929, 12930, 12931, 12932, 12934, 12936, 12937, 12938, 12939, 12940, 12941, 12942, 12943, 12945, 12946, 12948, 12949, 12950, 12952, 12955, 12956, 12957, 12958, 12960, 12961, 12962, 12963, 12964, 12966, 12967, 12968, 12969, 12970, 12972, 12975, 12979, 12980, 12981, 12982, 12983, 12984, 12985, 12986, 12988, 12989, 12990, 12992, 12993, 12995, 12996, 12997, 12998, 12999, 13001, 13002, 13003, 13004, 13005, 13006, 13007, 13008, 13009, 13010, 13011, 13012, 13013, 13014, 13015, 13016, 13017, 13018, 13020, 13022, 13023, 13024, 13025, 13027, 13028, 13029, 13030, 13031, 13034, 13035, 13036, 13037, 13038, 13039, 13040, 13044, 13046, 13047, 13048, 13049, 13050, 13051, 13052, 13053, 13055, 13056, 13057, 13059, 13060, 13061, 13062, 13063, 13064, 13068, 13069, 13070, 13071, 13072, 13074, 13075, 13076, 13077, 13079, 13081, 13082, 13084, 13085, 13086, 13087, 13088, 13089, 13090, 13091, 13092, 13093, 13094, 13095, 13097, 13098, 13099, 13100, 13101, 13102, 13103, 13107, 13110, 13111, 13112, 13114, 13115, 13116, 13117, 13119, 13120, 13121, 13122, 13123, 13125, 13126, 13129, 13130, 13131, 13132, 13133, 13134, 13135, 13136, 13137, 13140, 13141, 13142, 13143, 13144, 13145, 13151, 13152, 13153, 13155, 13156, 13158, 13159, 13160, 13161, 13162, 13163, 13164, 13165, 13166, 13167, 13168, 13170, 13171, 13172, 13173, 13174, 13175, 13176, 13177, 13179, 13180, 13181, 13182, 13183, 13185, 13186, 13187, 13188, 13190, 13191, 13192, 13193, 13194, 13195, 13196, 13197, 13198, 13199, 13200, 13202, 13203, 13204, 13205, 13206, 13207, 13208, 13209, 13210, 13211, 13215, 13216, 13217, 13218, 13220, 13221, 13222, 13223, 13224, 13225, 13226, 13227, 13228,13229, 13231, 13232, 13234, 13235, 13236, 13237, 13238, 13239, 13240, 13241, 13244, 13246, 13247, 13248, 13251, 13252, 13253, 13254, 13255, 13256, 13257, 13258, 13259, 13260, 13261, 13262, 13263, 13264, 13265, 13267, 13268, 13269, 13271, 13273, 13275, 13277, 13278, 13279, 13281, 13282, 13283, 13284, 13285, 13286, 13289, 13290, 13291, 13292, 13293, 13294, 13295, 13296, 13297, 13298, 13299, 13300, 13301, 13302, 13303, 13304, 13305, 13306, 13307, 13308, 13309, 13310, 13311, 13312, 13313, 13314, 13317, 13318, 13319, 13320, 13321, 13323, 13324,13325, 13326, 13327, 13329, 13330, 13331, 13332, 13336, 13337, 13339, 13340, 13341, 13342, 13343, 13345, 13347, 13348, 13349, 13350, 13351, 13359, 13360, 13361, 13362, 13363, 13365, 13366, 13367, 13368, 13369, 13370, 13371, 13372, 13373, 13374, 13375, 13379, 13380,13381, 13382, 13383, 13384, 13385, 13386, 13387, 13389, 13390, 13392, 13393, 13395, 13398, 13399, 13403, 13405, 13412, 13423, 13429, 13432, 13436, 13437, 13440, 13443, 13444, 13445, 13446, 13448, 13451, 13452, 13454, 13455, 13463, 13464, 13465, 13466, 13467, 13469, 13470, 13472, 13473, 13474, 13476, 13477, 13490, 13492, 13493, 13496, 13497, 13500, 13511, 13512, 13515, 13518, 13525, 13526, 13527, 13528, 13529, 13530, 13532, 13533, 13536, 13537, 13538, 13540, 13541, 13543, 13544, 13545, 13546, 13547, 13548, 13552, 13553, 13555, 13557, 13561, 13562, 13563, 13564, 13565, 13566, 13569, 13572, 13575, 13576, 13577, 13579, 13580, 13582, 13583, 13586, 13587, 13594, 13599, 13601, 13604, 13605, 13607,13608,13610, 13611, 13612, 13613, 13614,13616,13617,13619,13620, 13621, 13622, 13623, 13626, 13627, 13628, 13629, 13631, 13633, 13634, 13635, 13636, 13637, 13639, 13664, 13665, 13666, 13667, 13668, 13669, 13670, 13671, 13672,13673, 13674, 13675, 13676, 13677, 13678, 13680, 13681,13682, 13683, 13684, 13685, 13686, 13687, 13688, 13689, 13690, 13692, 13694, 13695, 13696, 13697, 13701, 13702, 13703, 13704, 13708, 13710, 13711, 13712, 13713, 13714, 13715, 13716, 13717, 13718, 13719, 13720, 13721,13722,13723,13724, 13725,13726, 41354

Başmuhasebe Kalemi Zimmet Halifeliği

13937

Büyük Ruznamçe Kalemi Sergi Halifeliği

21300, 21303

Cevdet Adliye

9/566, 61/3687, 84/5073, 101/6095, 104/6253, 105/6277, 25/1512, 45/2751, 54/3288, 55/3327, 60/3628, 61/3673, 62/3729, 62/3730, 63/3754, 93/5603, 93/5604, 95/5687, 703/28713

Cevdet Askeriye

1041/45716

Cevdet Dahiliye

2/83, 142/7090, 156/7795, 159/7901, 175/8713, 177/8815, 182/9098, 188/9381, 217/10808

Cevdet Maliye

42/1942, 69/4129, 70/3229, 82/4979, 95/4268, 95/4269, 95/4270, 105/4660, 144/6093, 167/7001, 189/7845, 227/9489, 243/10117, 243/10130, 255/10503, 269/11038, 317/13048, 326/13431, 336/13768, 336/13774, 339/13926, 357/14626, 358/14690, 369/15158, 373/15355, 374/15389, 375/15406, 376/15449, 389/15918, 395/16165, 427/17283, 428/17301, 445/17993, 448/18138, 451/18257, 481/19623, 481/19624, 489/19929, 492/20005, 504/20468, 504/20496, 505/20550, 558/22906, 576/23646, 578/23731, 583/23998, 613/25271, 624/25689, 650/26610, 662/27099, 703/28713, 714/29165, 719/29421, 735/30007, 748/30460, 749/30520, 759/30930, 30930, 762/31051, 769/31398, 774/31578, 776/31697, 777/31744, 780/31846, 784/31951

Darphane-i Amire

692/1245, 9/427, 693, 1246

Hattı Humayunlar

105/4117C

Kamil Kepeci Defterleri

2397, 2448, 2450, 2451, 2452, 2455, 2456, 2459, 2460, 2462, 2463, 3294, 3295, 3296, 3297, 753

Sources of Qualitative Analysis:

AE.III.Mustafa 25565

AE.SABH.I 65/475-3

AE. SABH.I 65/475-6

AE.SABH.I 66/4604-3

C.DH 152/7576

C.DH 329/16413

C.ML 460/18682

C.ML 22924 (1766)

D.BŞM.MHF.d 13289

D.BŞM.MHF.d 13327

D.BŞM.MHF.d 13341

D.BŞM.MHF.d 13454

D.BŞM.MHF 13465.

D.BŞM.MHF.d 13465.