Taxes, Efficiency, and Redistribution

Department of Economics Working Paper Series
Taxes, Efficiency, and Redistribution: Discriminatory Taxation
of Villages in Ottoman Palestine, Southern Syria and Transjordan
in the Sixteenth Century
Metin Cosgel
University of Connecticut
Working Paper 2002-22
October 2002
341 Mansfield Road, Unit 1063
Storrs, CT 06269–1063
Phone: (860) 486–3022
Fax: (860) 486–4463
http://www.econ.uconn.edu/
Abstract
Governments can tax productive activities with either uniform or discriminatory
rates among taxpayers. Although discriminatory rates can cause productive
inefficiency and require high cost of administration, they can be preferred because
of their advantage in distributional flexibility. This paper studies the discriminatory
taxation of production in the Fertile Crescent. Using information from the
Ottoman tax registers, it examines the basis, distortionary effects, and distributional
consequences of discriminatory rates quantitatively. The results challenge
widely held beliefs about the basis for discriminatory rates in this region and the
Ottoman government’s motivation in adapting systems of taxation in newly conquered
lands.
I wish to thank the participants at the Economic History and Development
Workshop at UMass, Amherst; the 2002 Annual Cliometrics Conference in La
Crosse, WI; the 2002 Economic History Association Meetings in St. Louis, MO,
and the 2003 Iowa Alumni Workshop in Iowa City, IA for helpful comments and
suggestions. Ali Ozdemir, Hesna Taskimir, and Sadik Yildirim provided valuable
research assistance.
1
Taxes, Efficiency, and Redistribution:
Discriminatory Taxation of Villages in Ottoman Palestine, Southern
Syria, and Transjordan in the Sixteenth Century
ABSTRACT: Governments can tax productive activities with either uniform or discriminatory
rates among taxpayers. Although discriminatory rates can cause productive inefficiency and
require high cost of administration, they can be preferred because of their advantage in
distributional flexibility. This paper studies the discriminatory taxation of production in the
Fertile Crescent. Using information from the Ottoman tax registers, it examines the basis,
distortionary effects, and distributional consequences of discriminatory rates quantitatively. The
results challenge widely held beliefs about the basis for discriminatory rates in this region and
the Ottoman government’s motivation in adapting systems of taxation in newly conquered lands.
A fundamental question governments face in taxing productive activities is whether the
rates should be uniform or discriminatory among taxpayers. Whereas under a uniform rate
structure the same rate would apply to all taxpayers, discriminatory rates would vary among
taxpayers based on their abilities, socio-economic status, what they do, or how much they earn.
Although uniform rates are simpler and cheaper to administer, discriminatory rates may also be
observed in a variety of contexts, primarily because of their advantage in distributional
flexibility. By varying rates among taxpayers, governments are able to redistribute income
between the sectors and regions of the economy or between political, demographic, or income
groups. Discriminatory rates induce inefficiencies, however, because they cause the after-tax
cost of production or input provision to be greater for some taxpayers than others. In deciding
2
whether the rates should be uniform or discriminatory, governments have to balance the loss in
efficiency against the social value of redistributing income through the system of taxation.
The cost of administering a system with discriminatory rates can be very high when the
characteristics of taxpayers do not differ systematically or when these differences cannot be
easily observed. It is generally easier to identify differences between the sectors of the economy
than within each sector, making it harder to implement discriminatory rates within a sector.
Although rates have historically varied between the agricultural and manufacturing or between
the market and non-market sectors, they have been more uniform among producers within each
of these sectors. In most medieval and early modern systems of taxation, for example, although
manufacturing and trade activities may have been taxed at different rates than agricultural
activities, all agricultural producers typically paid taxes at the same rate, usually as tithes.
The system of discriminatory taxation that prevailed in parts of the Fertile Crescent until
the nineteenth century stands out as a novel phenomenon in the history of taxation. Although the
origins of the system are unknown, it was widely practiced at the time of the Ottoman conquest,
possibly much earlier in history. When the Ottomans conquered this region in the early sixteenth
century, they inherited a system of agricultural taxation with discriminatory rates among villages.
Whereas in other parts of the Empire taxes on grains were levied at uniform rates (usually onetenth)
among villages, the rates in the newly conquered lands varied between one-seventh to
two-fifths of output. Despite the contrast between the two systems, the Ottomans preserved the
prevailing system of taxation in these lands and simply reassigned the tax revenues to
themselves.
This paper seeks to understand the nature and consequences of discriminatory taxation in
this region, examining it quantitatively within the relevant theoretical and historical context.
3
Three objectives guide the inquiry. The first is to examine the basis for discriminatory rates.
Although historians have generally argued that the rates depended on local factors like the
fertility of the soil and the availability of irrigation facilities, no satisfactory quantitative
evidence has been found for support. I use the Ottoman tax registers as primary sources of data,
which include information about over 1,300 villages in seven Ottoman districts corresponding to
Palestine, southern Syria, and Transjordan in 1595-96 (H"tteroth and Abdulfattah, 1977). I also
use maps of the region to construct measures of some of the physical characteristics, such as the
availability of irrigation water, of each village. Although a regression analyses shows various
variables that affected the tax rates significantly, the effects of some of the variables were not in
expected directions. Framing previous arguments into a coherent whole, I discuss their relative
merits.
The second objective is to determine the distortionary effects of discriminatory taxation.
Whereas the tax rates varied among villages for some products like grains, they were uniform for
other products like fruits and vegetables. By determining the relationship between the tax rates
and the outputs of these two types of products, we can determine the distortion caused by
discriminatory taxation. Economic theory of taxation has shown how taxes raise the after-tax
cost of a product and cause producers to inefficiently shift resources toward other products. All
else being the same, the magnitude of this distortion would rise with the rate of taxation. The
implication for a discriminatory rate structure is that, assuming substitutability between products
to be the same among villages, a village with a higher rate of taxation than another would be
expected to substitute a greater amount of the discriminatorily taxed product for other products.
Taxation theory would thus lead us to expect a negative relationship between the tax rates and
4
the output of a discriminatorily taxed product and a positive relationship with the outputs of
other products. Using data from the tax registers, I find that the results confirm the expectation.
The final objective is to examine the effect of discriminatory rates on income distribution.
To compare income distribution under discriminatory and uniform rates, I first determine the
existing distribution of income among villages at different points of the distribution and calculate
the income shares of the villages in each group. I then use the regression results of the
relationship between the tax rates and output to simulate what the incomes would have been
under a uniform rate structure. Comparing the distribution of income under the two rate
structures, I examine whether discriminatory rates improved or worsened income distribution.
The results challenge widely held beliefs about the Ottoman government’s motivation in
adapting systems of taxation in newly conquered lands.
DISCRIMINATORY RATES IN THEORETICAL AND HISTORICAL CONTEXT
The basic elements of a tax system are the tax base and the rate structure. Governments
have various choices in levying taxes on a productive activity, such as to impose a lump-sum
payment on the firm or to make taxes based on the revenue or profits earned from the activity or
on the returns to one or more of the inputs used in production. Corresponding to each tax base,
there is also the choice of whether the rates should be uniform or discriminatory among
taxpayers. The rates in discriminatory taxation may vary according to various criteria, including
geographic location (rural versus urban, developed versus backward), product category (farming
versus industrial, consumer versus capital goods), or personal characteristics (age, marital
status). Because tax collection is coercive, the choice of tax base and rate structure can have
important implications for various political and economic outcomes in a society, including the
allocation of resources and the distribution of income.
5
Uniform and discriminatory rates have significant implications for equity and efficiency
in taxation.1 By altering the relative costs of economic activities, all (non-lump-sum) taxes affect
behavior and cause inefficiencies as taxpayers naturally adjust to taxes by substituting other
activities for taxed ones. Although both uniform and discriminatory rates thus cause distortions,
economic inefficiencies are likely to be greater with discriminatory rates than with uniform rates.
Some inefficiencies are distinct outcomes of discriminatory taxation. For example,
discriminatory rates create opportunities for rent-seeking (as each group seeks to reduce its own
rate) and raise the cost of administration and compliance (given the asymmetric information
about taxpayers and their incentives to avoid taxation).2
Moreover, discriminatory rates among taxpayers cause not only allocative but also
productive inefficiencies.3 One of the requirements of production efficiency is for the marginal
rate of transformation between any two outputs to be the same in all producers, which can be
achieved when all producers face the same prices for outputs. Although a tax on an output that is
uniform across producers would not affect this condition, discriminatory rates would cause
producers to face different prices and cause the economy to be productively inefficient.
To compare the distortionary effects of uniform and discriminatory taxation in a simple
example, consider an economy with two groups of producers producing two goods, A and B.
Suppose that this economy is initially producing efficient quantities of A and B and that the
government decides to raise revenue by taxing A. Taxing this product at a uniform rate between
1 For the classical article on the efficiency of differential commodity taxes, see Ramsey (1927).
Despite some parallels, however, the problem is quite different with rates that differ among
commodities than with those that differ among taxpayers. For general reviews of efficiency and
equity in taxation, see Slemrod (1990) and Stiglitz (1999).
2 For a general analysis of public choice issues in taxation, see Holcombe (1998).
3 Diamond and Mirrlees (1971) analyze conditions for production efficiency in taxation in a
general framework.
6
the two groups would result in a deadweight loss by raising the price of A and causing an
inefficiently large production of B. Despite the loss in welfare, however, the economy can still
be productively efficient (on the production possibilities frontier) because all producers would
face the same prices. Raising the same tax revenue with discriminatory rates, on the other hand,
would result in not only a deadweight loss but also a loss in production efficiency. For the tax
revenue to be the same under the two rate regimes, the discriminatory rate has to be higher than
the uniform rate for one of the groups and lower than the uniform rate for the other group.
Because the tax rates are now different between the two groups, they face different costs and
produce the two goods at different marginal rates of transformation. The economy is no longer
productively efficient. Assuming supply elasticities to be the same between the two groups, the
high rate group would be producing too little A and the low rate group too little B, compared to
productively efficient levels. Aggregate output can be increased by diverting resources from B
to A in one group and from A to B in the other until the marginal rates of transformation are
equalized.
Are the inefficiencies of discriminatory taxation justified? Governments engage in
discriminatory taxation primarily for income redistribution.4 A variety of factors can affect the
choice of a rate structure, including the magnitude of inefficiencies, the nature of pre-tax
inequalities, and the balance of power between various social and political groups. These, in
turn, would be determined by such things as production technology, natural resources and their
4 If taxpayers’ supply elasticities differ systematically among producers, discriminatory taxation
can also help to maximize tax revenue by assigning higher rates to producers with lower
elasticity. See Ramsey (1927) for a parallel argument in differential taxation of commodities.
Because of my focus on income redistribution and the difficulty of determining supply
elasticities, I ignore the way discriminatory rates may also have affected state’s tax revenue.
7
allocation in the population, the availability of skilled staff to administer taxes, and a society’s
history of taxation.
Although discriminatory taxation was observed in the fiscal history of the Fertile
Crescent, the details of the system before the Ottoman period are not known. A basic component
of the tax systems observed in the region was the proportional taxation of some products, known
in Islamic taxation theory as the muq#sama (sharing) method.5 Although the muq#sama method
was not practiced uniformly throughout the region or continually over time, historical records
show that it extended far back into Islamic history, possibly even to the pre-Islamic period.
Discriminatory rates were an integral part of the method in some regions. Based on their
interpretation of the Islamic Law, Muslim jurists generally agreed that the rates of taxation could
vary among villages, and secondary sources indicate that some rulers collected taxes on grains
like wheat and barley at rates that varied among villages between one-twentieth to one-half of
total output. The lack of detailed and reliable data, however, makes it impossible to determine
with certainty how widespread was the system of discriminatory taxation during the time of the
early Islamic empires, what exactly were the rates in each village, and how the system changed
over time until the Ottoman rule.
Thanks to the rich information recorded and preserved by the Ottomans, it is possible to
analyze the details of discriminatory taxation in this region during their reign. Upon conquering
new lands, the Ottomans typically conducted a cadastral survey and recorded information about
tax-paying subjects and taxable resources in tax registers, updated periodically as circumstances
5 That the Abbasid caliph al-Mahdi (CE 775-85) reintroduced the method upon request by the
peasantry indicates that it existed before his reign, possibly much earlier in history. For Islamic
taxation methods, see L$kkegaard (1950). See also Johansen (1988) for Islamic law on land
taxes. For the historical background to agricultural taxation in the region, see also Lambton
8
changed over time, in order to have current information on the empire’s sources of revenue.6 At
the beginning of each tax register was the tax code of a region, which laid down the basic tax
regulations and specified the rates corresponding to different circumstances.7
The tax codes of some of the districts in the Fertile Crescent make it clear that the
Ottomans inherited a discriminatory rate structure from previous rulers and that this differed
from the system of taxation in other regions. The tax code of the district of Quds (Jerusalem),
for example, begins: “In this district, because the tax rate of each village is different, the rate is
recorded separately.”8 The tax registers of this and several neighboring districts indeed recorded
the tax rate for each village separately, in addition to the detailed information about the taxpayers
and the expected revenues from a variety of productive activities in each village. The wealth of
information that these registers provide make it possible to study the nature, causes, and
consequences of discriminatory rates in this region in great detail.
By studying the tax codes of various Ottoman districts, we can determine the general
structure of their fiscal system. There were three general categories of Ottoman taxes: personal
taxes that were based on a taxpayer’s characteristics like land ownership and marital status, trade
taxes based on the market exchange of goods and services, and production taxes that applied to
(1962), Lewis (1979: 112-14), and Poliak (1977: 65-66). Venzke (1997) studies tax rates and
agricultural productivity in Aleppo.
6 These registers were called defter-i h#kan% [imperial register], commonly known as the tahrir
defterleri (s. defter). For details, see Co!gel (2003), &nalc'k (1954), and &nalc'k (1994: Chapter
5). For the registers of the Arab lands, see Lewis (1951) and H"tteroth and Abdulfattah (1977:
1-11). See also Kark (1997) for a history of cadastral surveys of Palestine and Singer (1990a;
1990b) for research possibilities based on the Ottoman registers of this region.
7 The tax code of each province or district was called k#n(nn#me. See &nalc'k (1960) for the
history and types of k#n(nn#mes. For collections of Ottoman k#n(nn#mes, see Barkan (1943)
and Akg"nd"z (1990). For similarities and differences between the tax systems of the Ottomans
and other contemporary Islamic states, see Floor (1998) and Moosvi (1987).
8 See Singer (1994: 48-49) for an English translation of the Quds tax code.
9
various farming and manufacturing activities.9 Production taxes consisted of three subcategories
depending on the tax base: output taxes that were based on the total output of an activity, input
taxes that were based on one of the inputs used in production, and enterprise taxes based on the
activity as a whole. Output taxes applied primarily to grains, legumes, and fibers, and they were
assessed as a share of the total output. Input taxes typically depended on the quantities of the
land, trees, animals, or other inputs used in production, rather than on total output. For example,
taxes on fruits, nuts, and dates depended on the number (sometimes also the age, height, and
type) of trees. Taxes on vineyards similarly depended on the number of vines, taxes on
vegetables depended on the amount of land allocated to them, and taxes on animal products
depended on the numbers of beehives or animals. Enterprise taxes depended not on the amounts
of the total output or one of the inputs used in production but on the activity as a whole.
Table 1 shows examples of tax instruments and rates in representative districts of the
Ottoman Empire during the sixteenth century.10 Differences among districts in tax rates and
bases reflect regional economic conditions and also the way pre-Ottoman customs were adapted
into the Ottomans system of taxation. Because personal taxes often had origins in wellestablished
feudal obligations that prevailed in areas conquered by the Ottomans, the names and
9 Although the Ottoman budgets included other sources of revenue, such as the tributes from
vassal states, profits from government owned enterprises, and revenues from various fees and
fines like the marriage fees and criminal fines, these revenues are excluded from discussion
because of our focus on the discriminatory taxation of productive activities. Extraordinary levies
to the state called avar'z-' divaniyye are also omitted because of their irregular nature during the
sixteenth century. For Ottoman state revenues, see &nalc'k (1994: Vol. 1, pp. 55-76).
10 These districts represent the geographical diversity of the Ottoman Empire. Quds is in eastern
Mediterranean, Budin is in eastern Europe, and Antep and Malatya are in eastern Anatolia.
Items of taxation were similarly chosen to represent regional similarities and differences in rates.
See Akg"nd"z (1990) and Barkan (1943) for the complete tax codes of these and other districts.
10
rates of these taxes could vary among regions.11 Whereas a married peasant in Anatolia who
held farm land workable by a pair ()ift) of oxen paid the )ift tax and a bachelor male paid the
lower bachelor tax, similar taxes in the European districts, such as the “gate-tax” observed in
Budin, were typically levied on the household as a whole. Personal taxes (other than the poll
taxes imposed on non-Muslims) did not even exist in the Quds district (and surrounding districts
as well). Trade and input taxes, on the other hand, were remarkably similar across districts,
reflecting the way the Ottomans were able to standardize the system as much as possible.
A comparison of the output tax rates across districts shows the distinct nature of
discriminatory taxation in Quds and surrounding regions. In other parts of the Ottoman Empire
the output tax rates were uniform across villages in a district. Although the uniform rate itself
could vary from one district to another, such as between the rate of 1/10 in Antep and 1/5 in
Malatya, it was the same rate that nevertheless applied uniformly to all villages within a district.
The rates in Quds and neighboring districts, however, varied significantly among villages.12
The absence of personal taxes in Quds provides a partial explanation for why output tax
rates were generally higher in this district than in others. It may seem to be an unfair burden for
villages to be taxed here at rates as high as forty percent while in other regions the rates were
only ten percent. The higher rates did not necessarily mean, however, that the total tax burden in
Quds was significantly higher, because there were other types of taxes to consider. Table 2
reports the relative shares of different tax categories in representative districts. Although the
higher output tax rates in Quds might have caused the proportion of output taxes to be higher
there than in other districts, the absence of personal taxes (other than the poll taxes imposed on
11 For a detailed account and historical origins of personal taxes, see &nalc'k (1959).
12 Despite the spatial variation, the rates remained remarkably stable over time. See Makovsky
(1984: 104) for the infrequent cases of changing tax rates in the Quds district.
11
non-Muslims) could have compensated for this. The sum of output and personal taxes in Budin,
for example, constitute about the same proportion of total taxes. Moreover, the level and
composition of both taxes and incomes from productive activities were influenced by a variety of
other regional production and market conditions, rendering comparisons of tax burden based
solely on output tax rates misleading.
Rate differences between and within districts continued until the nineteenth century,
when the Ottomans finally fixed these rates uniformly at one-tenth throughout the Empire. The
standardization of the rate was part of the European-inspired reform movement known as the
Tanz%m#t. 13 The reforms included the centralization and universalization of the tax system in the
same way that contemporary European states had recently sought to accomplish. Not everyone
has agreed, however, with the reformers’ contention that discriminatory rates were dysfunctional
and needed to be changed. *mer L"tfi Barkan, one of the leading economic historians of the
Empire in the twentieth century, criticized Tanz%m#t reformers for failing to understand the
redistributive benefits of discriminatory taxation (Barkan, 1964). Barkan and reformers could
both be right, of course, if discriminatory rates once had good basis that ceased to exist by the
nineteenth century. In any case, the question remains whether Barkan and others were right in
asserting that discriminatory rates served useful functions during the sixteenth century.
13 For attempts at making the rates uniform, see Lambton (1962: 1037). See also Barkan (1964)
for the history of muq#sama taxes (*!"r) in the Ottoman Empire and the variation of these rates
among districts.
12
THE BASIS FOR DISCRIMINATORY TAX RATES
Unfortunately, the tax registers did not specify the basis for the tax rates of villages, and
other contemporary sources did not provide direct testimony for what exactly caused the rates to
vary among villages. Writers in the secondary literature have typically presumed the rates to be
determined by factors that could cause the cost or revenue of productive activities to differ
systematically. Combined with information about farming conditions in the Fertile Crescent, this
presumption has led them to argue that the rates depended primarily on differences in
productivity and irrigation possibilities. Lewis (1979: 119), for example, argues that rates were
“determined by the quality and situation of the land, the availability of irrigation, and of course
the existing usage.” L$kkegaard (1950: 109-10) similarly states that in determining the rates,
“due regard [was] being paid to the facilities for irrigation and the cultivation of the soil.”14
It is well-known that the availability of irrigation water has had great importance for
agricultural production in this region because of its Mediterranean-type semi-arid climate, with a
regular pattern of some winter rains and an absolute summer drought.15 Differences in irrigation
possibilities could clearly affect the cost of farming, as villages with poorer access to irrigation
water were likely to spend more effort and incur higher cost than other villages in producing the
same amount of output. Similarly, differences in the productivity of land could affect the
revenue from farming, allowing farmers on more productive lands to generate higher incomes.
A redistributive system of taxation could in turn be based on these differences by assigning
higher rates to villages with more productive resources and/or easier access to irrigation water.
14 For similar arguments about the relationship between tax rates and productivity, see Ashtor
(1976: 40), Bakhit (1982: 148), Barkan (1964), and S(d% (1996: 40). For a detailed discussion of
the effect of irrigation possibilities on tax rates, see L$kkegaard (1950: 120-22.).
15 For the climate and water resources of the region, see Beaumont, Blake, and Wagstaff (1976:
Chapter 2) and FAO (1997). See also Arnon (1992) for agriculture in dry lands.
13
Although historians have generally shared the view of discriminatory rates as an
instrument of redistribution, they have not been able to confirm any of the presumed
relationships by quantitative analysis. For example, finding the argument about the relationship
between tax rates and productivity reasonable, H"tteroth and Abdulfattah (1977: 64-65) checked
it “very carefully in hundreds of cases,” but they found the results disappointing: “there is no
definite correlation between the [tax rate] and the agricultural productivity of the respective
village lands.” Similarly, based on a quantitative analyses of tax rates in Quds, Makovsky (1984:
102) argues: “Despite the likely general accuracy of the … hypotheses concerning the
progressive nature of [discriminatory] taxation, there seems to be no absolute standard by which
we can predict the taxation rate of a given village.”
These studies may have failed to find the determinants of tax rates not because of a
nonexistent relationship between tax rates and other variables but because of insufficient data or
inadequate methods of investigation. Although they do not state it explicitly, H"tteroth and
Abdulfattah (1977) seem to have reached their conclusion from a merely visual inspection of
hundreds of cases rather than systematic quantitative analyses. Because a variety of factors, in
addition to the productivity of land, were likely to influence tax rates at the same time, it is
impossible to control for all these factors and isolate the effect of productivity by mere visual
inspection. Despite being more detailed and quantitatively more sophisticated, Makovsky’s
(1984) analysis of the relationship between tax rates and population is also likely to fail for the
same reason. Because the effect of population, if any, on the tax rates is likely to be mixed with
many other factors, tabulating the population and tax rates of villages is unlikely to show their
relationship in isolation.
14
For a quantitative analysis of the determinants of tax rates, this paper uses data from the
tax registers of the Ottoman districts of Quds, N#bl(s, !azza, Lajj(n, +Ajl(n, "afad, and Hawran
for the year 1595-96 (1005 H.).16 To construct a complete and reliable data set for the study of
the determinants of tax rates, I omitted fiscal units that made a single lump-sum payment for
taxes (rather than itemized taxes) and those with missing information on inhabitants or taxes.17
Of the 1559 fiscal units (excluding uninhabited fields called mazra’as) reported by H"tteroth and
Abdulfattah (1977), 211 observations were thus dropped, and the remaining 1348 villages
constitute the observations in the data set.
Table 3 shows the distribution of the inhabitants of these villages and their taxes by tax
rates. About half of all villages in the area were taxed at the rate of one-fourth, the majority of
the rest being taxed at one-third of total output. Only 5 villages were taxed at rates below 20
percent. Taxed at the rate of one-seventh was the village of Sukr%r in !azza. Those taxed at the
rate of one-sixth were the villages of Dayr Shayh and Bayt S#h(r an-Na,#r# in Quds and the
villages of Majdal #rkam#s and Tall Kis#n in the subdistrict (n#-iya) of Akk# in the district of
"afad. Nuwayr%, a writer of the Mamluk period, emphasized security concerns, such as being
16 The registers are numbered TK 72, 100, 112, 181, 185, and 192 in the Cadastral Office
archives in Ankara. In addition to reporting the results of their pioneering study of the historical
geography of this region, H"tteroth and Abdulfattah’s (1977) transcribed, categorized, and made
available the data contained in these registers for other researchers. Social, economic, and
demographic studies of the region in the sixteenth century have relied heavily on these registers
as the primary sources of information. See, for example, Makovsky (1984), Rhode (1979), and
Singer (1994). For the registers of the Arab lands, see Lewis (1951) and H"tteroth and
Abdulfattah (1977: 1-11). See also Kark (1997) for a history of cadastral surveys of Palestine.
Bakhit and Hmoud (1989) also published a series of registers of the region corresponding to
today’s Jordan.
15
located on the seacoast or near the enemy border, as the primary reason for why the Mamluks
assigned such low rates of taxation.18 Although one of these villages, Sukr%r, was located near
the coast in the northwestern part of !azza, the other four villages did not have similar locations,
and there were many other villages along the coast with higher rates of taxation.19 There must
have been something distinct about these five villages, such as being in charge of the protection
of (or catering to) the pilgrims during the annual pilgrimage to Mecca or the maintenance of an
important road or bridge, which caused their tax rates to be exceptionally low.
For a regression analysis of the determinants of tax rates, I generated various independent
variables from the tax data. I also used available maps of this region to construct other variables
that represent physical characteristics of a village. The variables of primary interest are those
that have been identified in the literature as the most significant influences on the cost and
revenue of farming operations: irrigation possibilities and agricultural incomes. It is impossible
to find a direct measure of the availability of irrigation facilities in each village in the late
sixteenth century, simply because we have no localized information on irrigation methods (e.g.,
canals or underground water sources) from that period. Rather than a direct measure of irrigation
possibilities, proxy variables have to be used to represent the availability of irrigation water in a
17 In a few number of cases, I filled in the tax rates of surrounding villages for the missing rate of
a village to be able to retain available information on other variables.
18 Makovsky (1984: 102) also discusses the case of Q(fin in the Quds district as being taxed at
the rate of one-tenth. H"tteroth and Abdulfattah (1977) list the same village as being taxed at
one-third in their register. An error may have been made in deciphering, recording, or printing
this record.
19 Nuwayr% (1923-42) also mentioned the possibility of low rates for those villagers leasing
deserted lands. Although their status as being no longer inhabited in the late mandatory period
appears to suggest harsh climatic conditions for three of these villages, the same was true for
over 200 other villages whose past locations were also identified (about 750 villages were still
16
village. One possibility is to use recent maps of the region to approximate a village’s access to
surface sources of water (rivers, streams, lakes), assuming them to be unlikely to have changed
significantly.20 I thus identified each village’s location relative to surface sources of water and
generated a dummy variable for whether a village was within close proximity of water.21 There
are some obvious potential problems of using such a measure, such as omitting other sources of
water that were available to villages in the sixteenth century and including sources that were not
available. Given the current state of our knowledge of irrigation facilities in each of these
villages in the sixteenth century, however, this is the best that we can do.
On the revenue side, the relationship between the tax rates and agricultural incomes could
exist at two different levels: total for the whole village or average per household. Even though
the registers listed the heads of households separately and the tax system made them responsible
for taxes individually, the taxes were recorded as a single sum for the whole village, rather than
broken down or averaged out by households. Although the government could presumably use
this information to calculate simple averages, there is no evidence that this was done. One might
argue that, because the government did not have information about average incomes of
households, tax rates were determined by the total revenue of the whole village, all else being the
same. Discriminatory tax rates applied only to output taxes, so I included as an independent
inhabited in the late mandatory period, and no definite determination could be made for the
status of the remaining 500 or so villages recorded in the tax registers).
20 The assumption is consistent with the evidence on climatic changes and the evolution of land
use in the region. See Butzer(1961) and Whyte (1961) for details.
21 For the location of villages, I relied on the maps produced by H"tteroth and Abdulfattah
(1977). I also relied on their estimation of where a village could be located in cases when the
location could not be identified with certainty in current maps. In locating available sources of
water, I did not include underground sources because no reliable and detailed maps exist for their
location and depth. Similarly, I omitted artificial canals because most canals that are currently
used for irrigation in this region have been built recently.
17
variable the sum of the incomes from products subject to output taxes by multiplying the taxes
listed in the registers for these products by the taxation factor (inverse of the tax rate).
The relationship between income and tax rates could also exist at the level of average
incomes per household. That the government did not calculate taxes or incomes per household
may not necessarily imply that income differences among villages were not well-known or that
these differences did not influence the determination of tax rates. There may have been wellknown
differences in soil productivity among villages, which may have caused incomes per
household to differ systematically. The tax system could have used these differences as the basis
for discriminatory rates, for example by assigning higher rates to villages with higher incomes
per household. To consider this possibility, I calculated each village’s average income per
household from products subject to the output tax.
I included various other variables into the regression analysis to be able to control for
their effects on the tax rates. The religious composition of a village’s inhabitants could be a
factor if, for example, the relative power of the tax recipients over the villagers in influencing the
tax rates depended on the proportion of Muslims in the village. Although the religious minorities
(Jews and Christians) paid a poll tax called jizya in addition to regular agricultural taxes, they
could be paying even higher taxes through higher rates of taxation on their grain production as
well. To consider this possibility, I calculated the proportion of Muslims in a village’s
population (adult males) as another variable.
A village’s distance to the nearest market town could affect its tax rate. This could be the
case if being farther from market towns meant for a village to have a lower tax rate because of
fewer opportunities for alternative employment, higher cost of marketing products, or costlier
access to other urban benefits. To approximate these considerations, I measured the direct
18
distance between a village and the nearest market town as a variable. Although the relevant
distance could vary considerably from this measure because of such factors as the terrain or the
presence and conditions of the roads and waterways, we do not have information about these
factors in the sixteenth century.
The recipient(s) of a village’s tax revenue could also affect its rate of taxation. The
distribution of tax revenue could affect the rates if, for example, the recipients differed
systematically in their incentives to maximize the revenue. 22 To consider this possibility, I
grouped the recipients into three categories based on the way they collected and spent taxes. 23
The first is the central government, which collected taxes through its agents. Because the money
entered the treasury before being spent and thus those who collected the money were different
from those who spent it, the incentives to maximize the tax revenue were different. The
government’s tax collectors, who had the expertise and local knowledge, did not have full
incentives to try to maximize the tax revenue. The second category consists of all members of
22 The causality between the rates and recipients of taxes could also have gone in the other
direction, such as if recipients were determined based on the tax rates, which would be a case of
an endogenous relationship. Despite being an interesting possibility, this interdependence is
beyond the scope of this paper and is not investigated in detail.
23 Under the Classical Ottoman system of government finance, tax revenues were either retained
by the central treasury, awarded to various military, judicial, and administrative personnel as
remuneration for their services, or distributed to other groups or organizations as their rights
recognized by the system of land tenure. The revenue recipients recorded in the tax registers of
this region were the p#d%!#h (the ruler, central government), m%r liw# (provincial governor and
commander of troops), holders of t%m#r and za’#ma (small and large fiefs), waqfs (pious,
charitable foundations), and the holders of such rights as m"lk (private) land and haq ’arab
(share of the Bedouins). In Qa$# .awr/n, some of the revenues were also awarded to the m%r
m%r#n (governor of the wil#yat) of Sh#m (Damascus). The tax revenues of some villages were
divided between two or more recipients, in which case the registers recorded the respective
portions (as a fraction of 24) of all recipients. In such cases, I calculated the amount of tax
revenue to add to each recipient’s grand total by simply multiplying these portions by the tax
revenue of the village. See Co!gel and Miceli (2003) for how risk and transaction costs affected
the allocation of tax revenue among these recipients.
19
the governing and military class in the provinces. There are some obvious differences among the
members of this group, for example between the cavalrymen who typically lived on the same
land as he collected taxes and other members of the governing and military class who lived
farther and thus used agents to collect taxes. More relevant, however, was that all recipients in
this group had the same type of high incentives to maximize the tax revenue, both to increase
their personal living standards and to improve their professional positions.24 In the third
category are the pious foundations (and various other negligible recipients whose combined
revenue was less than two percent of total taxes). Although there were pious foundations of
various types and sizes, as a whole they had the same type of incentives to maximize tax
revenues, which was also distinct from the incentives of those in the two previous categories. As
was the case with the central government, they needed agents to collect tax revenues, and
different people thus collected the taxes and spent the revenue. But because pious foundations
often had a clear focus and local operations and management, they were less likely to suffer from
the same problem of incentives facing the central government.25 By being local and thus clear
and present, the beneficiaries of a foundation were likely to apply strong pressure to increase the
tax revenues. Similarly, although the manager(s) of a foundation did not receive all tax revenues
directly as remunerations for service, they nevertheless had high incentives to maximize tax
revenues in order to fulfill professional obligations and improve service, if not to increase
personal standards of living.
24 There is also a more practical reason for grouping them in the same category. Although these
recipients were clearly identified and separated in other regions, they were not distinguished in
the registers of the qa$# of Hawran.
25 Although some foundations received tax revenues from distant villages, the argument about
differential incentives still applies.
20
To consider the effect of the distribution of tax revenue, I generated three dummy
variables corresponding to the three categories of recipients.26 In the regression equation, I
omitted the dummy variable for the central government to avoid multicollinearity, so the
coefficients of the remaining two dummy variables show the different effect of having a pious
foundation or one of the members of the provincial governing and military class, rather than the
central government, as the primary recipient of the tax revenue.
The distribution of tax revenue could affect the rates differently if there were multiple
recipients of the revenue. For example, one might expect the upward pressure on the tax rate to
be greater in villages where two or more recipients share the tax revenue than in villages with a
single recipient. To account for this possibility, I generated a dummy variable based on whether
a village’s tax revenue was divided (1 if divided between two or more revenue holders).
Finally, I included a set of dummy variables to determine and control for the influences
of possible but unobservable differences among the seven districts comprising the data set. A
variety of unobservable regional factors, such as systematic differences among districts in
climate, topography, and political factors, could have influenced the tax rates. I thus generated
dummy variables corresponding to each of the districts (1 if village is in the stated district). The
dummy variable for Quds is dropped, so the remaining six dummy variables show the way
unobservable factors caused the tax rates of other districts to differ systematically from the tax
rates in Quds.
26 In cases of divided revenue, I coded the holder of the highest share as the primary recipient of
tax revenues.
21
Table 4 shows the results of the regression estimation of influences on the tax rate.27 The
effects of control variables are interesting in their own right. The results show that religious
composition of a village and its distance to the nearest market town did not affect its tax rate
significantly. The distribution of tax revenue and local factors, on the other hand, affected tax
rates significantly and generally in expected directions. The rates were likely to be higher in
villages where multiple parties shared tax revenues. Villages paying taxes to pious foundations
and provincial officials were also likely to pay at higher rates than those paying to the central
government, confirming the expectation about the effect of differential incentives to maximize
the tax revenue. The coefficients of regional dummies are also significant, showing that villages
in other districts except N#bl(s were likely to have lower rates of taxation than those in Quds.
The magnitudes of these influences are also interesting. Because the dependent variable
is in logarithmic form, the coefficients of explanatory variables have a percentage interpretation.
The coefficients of some of the district dummies are considerably high, explaining as much as 32
percent of the difference in tax rates between districts (as can be seen from the difference
between the coefficients of "afad and N#bl(s). For example, given that the dummy variable for
Quds was dropped, the coefficient for "afad indicates that the grain output of a village was likely
to be taxed at a rate about 26 percent lower in "afad than in Quds, holding all other variables
constant. The variables representing the distribution of the tax revenue, on the other hand, had
smaller influences, varying around 3 percent, on the tax rate.
27 The (mean, standard variable) of the non-dummy variables included in the regression
equations are as follows: natural logarithm of tax rate (3.4, 0.19), natural logarithm of the
revenue from output-taxed products (9.4, 0.9), natural logarithm of per-household income from
output-taxed products (6.3, 0.86), percentage of Muslims (0.98, 0.097), distance to nearest
market town (3.0, 2.2).
22
Controlling for other factors makes it possible to identify the individual effects of
incomes and irrigation possibilities. To the extent that a village’s current proximity to sources of
fresh surface water represents the possibilities for irrigation in the sixteenth century, the positive
coefficient of this variable indicates that access to irrigation water did indeed affect the tax rate,
confirming the expectation that the tax rates were likely to be higher in villages with better
access to water than in others.
The coefficients of “Village Income (from Output-taxed Products)” and “Income per
Household” are both significant but with different signs, providing two alternative hypotheses
about the relationship between incomes and tax rates.28 Whereas the significance of the
coefficient of village income supports the view that the whole village’s income served as the tax
base in determining the rates, the significance of the coefficient of income per household
suggests that the tax base was the average income per household. Whereas the positive
coefficient of the former variable indicates a progressive rate-base relationship, the negative
coefficient of the latter indicates that the rate structure was regressive. Unfortunately, we do not
have sufficient information to determine whether it was the total or average incomes that actually
served as the tax base when the rates were first assigned. Further research based on other
sources is required to decide which of these possibilities was the more likely scenario.
Despite being small in magnitude (a one percent increase in average income lowered the
tax rate by about 0.04 percent), the latter result contradicts the argument commonly made in the
literature about the relationship between tax rates and productivity (assuming that higher
28 The simple correlation coefficient between the total income and income per household is 0.44,
suggesting the possibility of a multicollinearity problem between them. The sign and
significance of their coefficients, however, do not change when one of these variables is
dropped. Both variables are thus kept in the reported regression to provide more complete
information about their effects.
23
incomes were generated by higher productivity). One possible explanation for this surprising
result is that the negative relationship between tax rates and average incomes could simply be an
unintended consequence. If, for example, people overcrowded villages with easy access to
irrigation water and the tax rates were based primarily on irrigation possibilities, then (Ricardian)
diminishing returns in agriculture would have caused marginal product and average incomes in
these villages to fall, resulting in a negative relationship between tax rates and average incomes.
If villagers and Ottoman officials were unable to systematically compare average incomes and
tax rates across villages, this unintended outcome may have gone unnoticed. As another
possibility, the rates may have been intentionally assigned to achieve a negative relationship
between tax rates and average incomes. If, for example, the fiscal authorities were concerned
solely with minimizing the distortionary effects of taxation (rather than redistribution) and perhousehold
incomes and supply elasticities of output-taxed products were positively correlated,
then (similar to Ramsey taxes in commodity taxation) optimal taxation would require to assign
high rates to villages with low incomes (inelastic supply). Once again, lack of sufficient
evidence forces us to leave it to further researchers to determine which of these possibilities was
the more likely explanation.
DISCRIMINATORY RATES AND OUTPUT DISTORTION
Taxes are mandatory, but taxpayers are free to adjust their behavior in order to avoid
being taxed as much as possible. The Ottoman taxpayers too had opportunities to shelter some
of their income from taxation by allocating their resources among activities. The difference
between the rate structures of output and input taxes provided such an opportunity. Because the
rates differed among producers for output-taxed activities but were uniform for input-taxed ones,
taxpayers could adjust to their own rates by changing the composition of output between these
24
two types of activities. For example, a village could have adjusted to a high output tax rate by
shifting resources from items subject to the output tax to either non-taxed activities or to those
that are subject to the input tax, such as by converting a grain field to vegetable garden.
Although regulations may have prevented producers from altering their product mix significantly
in any one year, they could have nevertheless achieved desired changes in the long run as
cumulative outcomes of small yearly adjustments. We would thus expect taxpayers with higher
output tax rates than others to produce less of the output-taxed products, all else being the same.
As a corollary, the amounts of input-taxed products (as substitutes for output-taxed products)
would be expected to be higher for high-rate producers than others.
These expectations about the distortionary effects of discriminatory taxation can be tested
by a regression analysis of how tax rates affected output-taxed and input-taxed products in two
separate equations. The dependent variable of the first equation is the revenue from output-taxed
products, which we can easily calculate (as we have already done for the previous regression
analysis) by simply multiplying the taxes listed in the registers for these products by the taxation
factor (inverse of the tax rate). It is much harder to calculate the dependent variable of the
second equation: the revenue from input-taxed products. Although the tax codes stated the tax
rates for inputs clearly, we have no direct information on how to relate these rates to the output
or revenue of these products.29 We can nevertheless estimate input-output rates indirectly by
combining the tax data with more recent information about the productivities and production
processes of some of the items listed in the registers. Information about olives and grapes is
29 Faced with the same problem, H"tteroth and Abdulfattah (1977: 79-85) assumed that the taxes
collected from these products were proportional to their output according to the local taxation
factor, so they calculated the revenue from the production of fruits, vegetables, animal products
and various other products by multiplying the tax amount by the taxation factor of the respective
villages.
25
particularly useful because they could be taxed either as raw products or as finished products like
olive oil and grape syrup, and the registers listed these taxes either individually or as combined
with other taxes, and as based on either inputs or outputs. The registers thus provide information
about the quantities of inputs and the prices of outputs, which we can combine with external
sources on current yields of olive trees and grape vines, the oil or juice contents of olives and
grapes, and the conversion rates of weights and measures between the two time periods to
estimate revenue from taxes. Based on the assumption that yields did not change significantly
since the late sixteenth century, I calculated the ratio of revenue to taxes to be remarkable similar
between these products: 8.1 for grapes and 8.2 for olives.30 Assuming yields to be similar within
the region and tax arbitrage to be possible among input-taxed products, I used the rate of oneeights
for all input-taxed products to convert taxes to revenue in all villages.
I used the same set of independent variables in both equations because the same set of
factors could have potentially, if differently, affected output-taxed and input-taxed products. The
independent variable of primary interest is the tax rate. Among the other variables that could
have also affected production, perhaps the most important were the inputs used in production.
Although the tax registers did not directly record the quantities of inputs, they did record the
numbers of adult males in each village, which we can use to generate a general proxy for all
inputs. Assuming the number of adult males to be proportional to the agricultural labor force
and input proportions to be similar among villages, this measure would represent the units of the
input bundle used in production.
30 For estimates of current average yields, I relied on the Isaac, et al’s (1994) team study of
dryland farming in parts of this region. For conversion rates, see Makovsky (1984: 109-110). To
determine whether the results are sensitive to these calculations, I ran the same regressions with
different rates, as low as one-third (proportion of tax to revenue). The results did not change
significantly.
26
Among the variables created earlier for examining influences on the tax rate, the
availability of irrigation water in a village, its distance to the nearest town, and the recipients of
its tax revenue could have also affected production. So I included these variables in this
regression analysis. I used the information from tax registers to create three new dummy
variables that represent other economic activities in a village. To consider the effects of
commercial activities, I created a dummy variable based on whether the village hosted the
periodic regional market or pursued any urban activities (1 if the village paid b#j b#z#r or other
urban taxes). I created two other dummy variables to consider the effects of making investments
in manufacturing activities. One is whether the village had a water mill (1 if village paid taxes
for 0#-(1), whose presence would indicate a lower cost of converting grain to final products and
could thus have had a complementary, positive effect on grain production. The other dummy
variable is whether the village had a press for grape syrup or olive oil (1 if village paid taxes on
ma’sara), which would similarly indicate a lower cost in the processing of fruits and olives into
final products.31
Just as unobservable differences among districts were found to affect the tax rates
significantly, more localized differences among subdistricts could have affected the productions
of output-taxed and input-taxed products. To control for unobservable differences among
subdistricts in, for example, rainfall, climate, prices, and soil quality, I generated dummy
variables, similar to the district dummies of the first equation, for the forty-two subdistricts
represented in the data set.
31 Some types of olive trees called zayt(n R(m#n% were subject to the output tax, but the registers
did not always record the distinction carefully. Sometimes, taxes from olive trees were lumped
together with other items, such as fruit trees, that were subject to the input tax.
27
Table 5 reports the OLS estimates of influences on the revenues of output-taxed and
input-taxed products separately. One might argue that a method of simultaneous estimation
should have been used to estimate these equations. Recall that the revenue from output-taxed
products was one of the independent variables in the regression analysis of the determinants of
tax rates reported in Table 4. That the tax rate is now an independent variable in Table 5 thus
indicates an interdependent relationship between the tax rate and output-taxed products. Testing
for the endogeneity of the value of output in the equation in Table 4 and of the tax rate in the first
equation in Table 5 weakly confirms that these two variables were indeed jointly determined.32 I
therefore used the two stage least squares (2SLS) method to estimate the two equations
simultaneously. The signs and significance of the coefficients, however, were generally
consistent between the OLS and 2SLS estimates. The coefficients were also surprisingly close.33
Table 5 thus reports only the OLS results in order to avoid redundancy and to maintain
consistency between the two equations.
The results show interesting relationships between the control variables and the revenues
from output-taxed and input-taxed products. The number of adult males affected the quantities
of the two types of products positively, as one would expect. Being close to irrigation water had
an insignificant effect on these products, probably because it affected more the cost of
production than its revenue. Villages with a water mill produced more of both types of products
than other villages. Although the positive effect of mills on output-taxed products confirms
expectations about their complementarity, it is difficult to explain why the same type of positive
32 See Wooldridge (2000: 483) for a description of the test performed.
33 The only coefficient that changed significantly was that of the tax rate, possibly because the
rates in the original data were discrete values that ranged between 1/7 and 2/5 whereas the values
in the instrumental variable used in the 2SLS estimation could be continuous beyond this range.
The results of the 2SLS estimation are available upon request.
28
relationship existed between mills and input-taxed products. Consistent with grape syrup and
olive oil being output-taxed products, having a press for grape syrup or olive oil affected the
production of output-taxed products (but not input-taxed products) positively. The presence of
urban and commercial activities in a village, on the other hand, affected the input-taxed products
(but not output-taxed products) positively, indicating a complementary relationship between
them. Although the distribution of tax revenue also affected production in interesting ways, it is
beyond the scope of this paper to examine them in detail.34
The coefficients of the tax rate in the two equations show how taxes distorted output.
The negative coefficient of the tax rate in the first equation confirms the expectation that, all else
being the same, taxpayers adjusted to higher rates by producing less of the output-taxed products
(and substituting by others). The positive coefficient of the tax rate in the second equation shows
the other side of the substitution effect: taxpayers substituted output-taxed products with inputtaxed
ones as those with higher rates produced more of the latter. The magnitudes of these
effects are also interesting. Because both variables are in logs, the coefficients of the tax rate
reflect tax elasticities of the two types of products. Tax elasticity of supply is low in both cases,
possibly caused by the immobility of resources and restrictions on changing the composition of
products.
DISTRIBUTIONAL CONSEQUENCES
One of the features of the Ottoman government frequently emphasized in the literature is
their concern about fairness in taxation. The rulers appear to have taken various measures to
34 Because of space limitations, the results of the dummy variables that control for the
differences among the forty-two subdistricts are not reported in the Table. A majority of these
variables affected output significantly, confirming the importance of local factors in production
decisions.
29
prevent inequitable practices in collecting taxes. They announced great pride in changing some
of the taxes in the newly conquered lands, such as by commuting labor services in the European
provinces, in attempts to introduce fair and just taxation. They similarly established clear
regulations about how and when taxes should be collected and how tax collectors should not
abuse their power. When one type of tax was replaced by another, for example, the tax codes
often explicitly prohibited tax collectors from taking advantage of the situation by collecting
both types of taxes. Observing these measures against unjust taxation, leading historians of the
Empire have typically portrayed the Ottoman legal and financial system as fundamentally fair.
&nalc'k (1973: 73-4), for example, writes that the Ottoman laws “introduced a system of taxation
which was in general simpler and less liable to abuse than the earlier systems of feudal services.
[Various] regulations aimed to prevent the military class oppressing the peasantry and, therefore,
assessment according to means and collection according to law were the governing principles of
the tax system.”
The question then becomes whether preserving discriminatory rates in the Fertile
Crescent was consistent with the Ottoman concern for fairness. One of the standards of fairness
is vertical equity, which suggests that individuals who are in a position to pay higher taxes than
others should do so. If there were significant, systematic differences among villages in their
ability to generate income, vertical equity thus requires a discriminatory rate structure to assign
higher rates to those who are in a position to pay more than others.
Although the negative relationship between the tax rates and the per-household revenue
from output-taxed products (Table 4) suggests a regressive system of taxation, this does not by
itself mean that the tax system worsened the distribution of incomes as a whole. As the results in
Table 5 show, the taxpayers were able to adjust their total net incomes by responding to higher
30
rates by decreasing the production of output-taxed products and increasing input-taxed products.
To determine how tax rates affected incomes, we thus need to consider not just the revenue from
output-taxed products but the net incomes from all activities.
I used the tax data and previous regression results to calculate and compare the total
after-tax incomes under discriminatory and uniform rate structures. I used the same procedure
outlined above to estimate a village’s income from output-taxed and input-taxed products with
discriminatory rates. Subtracting the amount of total taxes paid by the village and dividing the
result by the number of households gives the net per-household incomes in a village. To
simulate what the incomes would have been under a uniform rate structure, I used the regression
results reported in Table 5. Simple calculations show that the uniform tax rate that would be
required to generate the same amount of tax revenue (as with discriminatory rates) is 29.25
percent. Substituting this for the tax rate in the data and using the regression equations to
calculate the predicted value of the dependent variable for each village gives the simulated
revenue from output-taxed and input-taxed products corresponding to this tax rate, all else being
the same. Similarly recalculating the taxes corresponding to this rate and subtracting them from
revenues, I estimated the net incomes under a uniform rate structure.
Table 6 shows the distribution of income under the two types of rate structures at
different points of the distribution. The distribution is shown separately for villages that were
near and distant from sources of irrigation water to determine whether any redistribution took
place between them. Contrary to expectations raised by the literature, the distribution is less
equitable under discriminatory rates than under uniform rates. Discriminatory rates decrease the
shares of total income received by the poorest 80 percent of the households, while increasing the
share of the richest 20 percent. It is also interesting that redistribution was more pronounced
31
among villages distant from sources of irrigation. Even though the magnitude of redistribution
was small, the direction of redistribution nevertheless shows that the discriminatory rate structure
was regressive.
This does not mean, of course, that the system of discriminatory taxation was originally
intended to be regressive or that the Ottomans intended to use the tax system to redistribute
income from high-income to low-income villages. Such interpretations of the results have to
wait until more direct evidence can be found about the origins of the system or about the
Ottoman’s intentions in preserving it.
The results nevertheless suggest an alternative to the standard reason offered in the
literature for why the Ottomans preserved some types of taxes after conquest and changed others.
It may have been simple pragmatism and concern with political stability, rather than an
overriding concern for justice and equity, that determined Ottoman policy. They may have thus
chosen to preserve the discriminatory rate structure in the Fertile Crescent not because they knew
about and agreed with its redistributive consequences but because they found it politically more
feasible to preserve than abolish a system that had been in existence for centuries. Unlike the
commutation of labor services in the European provinces that was typically welcomed by all
peasants, the switch from discriminatory to uniform rates would have faced significant resistance
because the Ottomans would have had to lower the rates for some villages but raise them for
others in order to preserve the same tax revenue.
CONCLUSION
Combining information from the tax registers of the Ottoman Empire with other
information about the physical characteristics of villages, this paper has examined the
determinants and consequences of discriminatory taxation in the Palestine, southern Syria, and
32
Transjordan region in the late sixteenth century. The results support the widespread belief that
the availability of irrigation water was one of the determinants of discriminatory rates in this
region. The widespread belief about the positive relationship between the tax rates and
productivity, however, receives no support.
As an adverse consequence, discriminatory rates caused distortion in output. Consistent
with behavior expected from rational decision-makers, the producers in this region responded to
taxes by reallocating their output between discriminatorily taxed and other products. All else
being the same, villages with higher tax rates produced less of the discriminatorily taxed
products and more of other products, and vice versa. By altering the cost of production among
producers and causing them to alter the product mix, discriminatory rates thus caused
inefficiencies in production.
Whereas inefficiencies in taxation can sometimes be justified by more equitable
distribution of income, discriminatory taxation in this region actually made the distribution even
less equitable. Contrary to arguments raised in the literature, a comparison of the distribution of
income under uniform and discriminatory rate structures shows that discriminatory rates
worsened the distribution of income by redistributing it from all other villages to those in the
richest 20 percent group. By preserving, if not initiating, such a regressive system of taxation,
the Ottoman administration must have, if unintentionally, contributed to the persistence of
income inequalities in this region.
33
WORKS CITED
• Akg"nd"z, Ahmed. 1990. Osmanl' Kanunn#meleri ve Hukuk% Tahlilleri. &stanbul: Osmanl'
Ara!t'rmalar' Vakf'.
• Arnon, Itzhak. 1992. Agriculture in Dry Lands: Principles and Practice. NY: Elsevier.
• Ashtor, Eliyahu. 1976. A Social and Economic History af the Near East in The Middle Ages.
Berkeley: University of California Press.
• Bakhit, Muhammad Adnan. 1982. The Ottoman Province of Damascus in the Sixteenth
Century. Beirut: Librairie du Liban.
• Bakhit, Muhammad Adnan and Nufan Raja Hmuod. 1989. The Mufassal Defter of Marj Bani
‘Amir and Its dependents Entrusted to Amir Tarabey 945 A.H./1538 A.D. Amman: The
University of Jordan.
• Barkan, *mer L"tfi. 1943. XV ve XVI. As'rlarda Osmanl' &mparatorlu2unda Zirai
Ekonominin Hukuki ve Mali Esaslar' . Cilt 1: Kanunlar Istanbul 3niversitesi yay'nlar'ndan.
Edebiyat Fak"ltesi T"rkiyat Enstit"s" Ne!riyat'. Istanbul, Burhaneddin Matbaasi.
• Barkan, *mer L"tfi. 1964. “*!"r,” &sl#m Ansiklopedisi IX: 485-88.
• Beaumont, Peter, Gerald H. Blake, and J. Malcolm Wagstaff. 1976. The Middle East: A
Geographical Study. NY: John Wiley & Sons.
• Butzer, K. W. 1961. “Climatic Change in Arid Regions since the Pliocone,” in L. Dudley
Stamp, ed. A History of Land Use in Arid Regions. France: UNESCO, pp. 31-56.
• Co!gel, Metin M. 2003. “Ottoman Tax Registers (Tahrir Defterleri),” Historical Methods,
forthcoming.
• Co!gel, Metin M. and Thomas J. Miceli. 2003. “Risk, Transaction Costs, and Government
Finance: The Distribution of Tax Revenue in the Ottoman Empire,” Working paper.
• Diamond, Peter A. and James A. Mirrlees. 1971. “Optimal Taxation and Public Production
I: Production Efficiency,” and “Part II: Tax Rules,” American Economic Review 61: 8-27 and
261-78.
• Floor, Willem. 1998. A Fiscal Study of Iran in the Safavid and Qajar Periods, 1500-1925.
New York: Bibliotheca Persica Press.
• Food And Agriculture Organization of the United Nations. 1997. Irrigation in the Near East
in Figures. Rome: FAO.
• Holcombe, Randall G. 1998. “Tax Policy from a Public Choice Perspective,” National Tax
Journal 51(2): 359-371.
• H"tteroth, Wolf-Dieter and Kamal Abdulfattah. 1977. Historical Geography of Palestine,
transjordan and Southern Syria in the Late 16th Century. Erlanger Geographische Arbeiten,
Vol. 5, Erlangen.
• Isaac, Jad, et al. 1994. Dryland Farming in Palestine. Applied Research Institute-
Jerusalem.
• &nalc'k, Halil. 1954. “Ottoman Methods of Conquest,” Studia Islamica II: 103-130.
• &nalc'k, Halil. 1959. “Osmanl'larda Raiyyet R"sumu,“ Belleten Cilt 23, Say' 92, pp. 575-
610.
• &nalc'k, Halil. 1960. “K#n(nn#me,” Encyclopedia of Islam, Second Edition.
34
• &nalc'k, Halil. 1973. The Ottoman Empire: The Classical Age, 1300-1600. London: Phoenix
Press.
• &nalc'k, Halil. 1994. [with Donald Quataert]. An Economic and Social History of the
Ottoman Empire, 1300-1914. New York: Cambridge University Press.
• Johansen, Baber. 1988. The Islamic Law on Land Tax and Rent. New York: Croom Helm.
• Kark, Ruth. 1997. “Maml(k and Ottoman Cadastral Surveys and Early Mapping of Landed
Properties in Palestine,” Agricultural History 71(1): 46-70.
• Lambton, A.K.S. 1962. “Khar#dj,” Encyclopaedia of Islam. Second Edition. Leiden.
• Lewis, Bernard. 1951. “The Ottoman Archives as a Source for the History of the Arab
Lands,” Journal of the Royal Asiatic Society 3-4: 139-55.
• Lewis, Bernard. 1979. “Ottoman Land Tenure and Taxation in Syria,” Studia Islamica 50:
109-24.
• L$kkegaard, Frede. 1950. Islamic Taxation in the Classic Period, with Special Reference to
Circumstances in Iraq. Copenhagen: Branner and Korch.
• Makovsky, A. 1984. “Sixteenth-century Agricultural Production in the Liw/ of Jerusalem:
Insights from the Tapu Defters and an Attempt at Quantification,” Archivum Ottomanicum 9:
91-127.
• Moosvi, Shireen. 1987. The Economy of the Mughal Empire c. 1595, A Statistical Study.
Delhi: Oxford University Press.
• Nuwayr%, A-mad ibn ‘Abd al-Wahh#b. 1923-42. Ni-#yat al-‘Arab f% Fun(n al-Adab. Cairo.
14 volumes.
• Poliak, Abraham N. 1977. Feudalism in Egypt, Syria, Palestine, and the Lebanon, 1250-
1900. Philadelphia: Porcupine Press.
• Ramsey, Frank. 1927. “A Contribution to the Theory of Taxation,” Economic Journal 37:
47-61.
• Rhode, Harold. 1979. The Administration and Population of the Sancak of Safad in the
Sixteenth Century. Ph.D. Dissertation, Columbia University.
• Singer, Amy. 1990a. “Tapu Tahrir Defterleri and Kad' Sicilleri: A Happy Marriage of
Sources,” Tarih: a Volume of Occasional Papers in Near Eastern Studies 1: 95-125.
• Singer, Amy. 1990b. “The Countryside of Ramle in the Sixteenth Century: A Study of
Villages with Computer Assistance,” JESHO 33: 51-79.
• Singer, Amy. 1994. Palestinian Peasants and Ottoman Officials: Rural Administration
around Sixteenth-Century Jerusalem. New York: Cambridge University Press.
• Slemrod, Joel. 1990. “Optimal Taxation and Optimal Tax Systems,” Journal of Economic
Perspectives 4(1): 157-78.
• Stiglitz, Joseph E. 1999. The Economics of the Public Sector. Third Edition. New York: W.
W. Norton &Company.
• S(d%, S"leyman. 1996. Osmanl' Vergi D"zeni [Defter-i Muktesid] Isparta.
• Venzke Margaret L. 1997. “The Ottoman Tahrir Defterleri and Agricultural Productivity”,
Osmanlı Araştırmaları [The Journal of Ottoman Studies] 17: 1-61.
• Whyte, R. O. 1961. “Evolution of Land Use in South-western Asia,” in L. Dudley Stamp, ed.
A History of Land Use in Arid Regions. France: UNESCO, pp. 57-118.
35
• Wooldridge, Jeffrey M. 2000. Introductory Econometrics: A Modern Approach. South-
Western College Publishing.
36
TABLE 1
EXAMPLES OF TAX RATES IN OTTOMAN DISTRICTS
DISTRICT
(YEAR)
PERSONAL TAXES TRADE
TAXES
PRODUCTION TAXES
Input taxes
Output
taxes
4ift-tax
Bachelo
r Tax
Gate
Tax
Examples of
Goods Brought
to Market Beehives
Animal
Products Vineyards Tax Rate
Antep (1574) 40 6 1 / camel-load
of
miscellaneous
goods
2 / beehive 0.5 /
animal
0.02 /
vine
Uniform
1 / 10
Budin (1562) 50 4 / wagon-load
of pots and cups
1/10 or
2 / beehive
0.5 /
animal
4 / d5n"m Uniform
1 / 10
Quds
(Jerusalem)
(1562)
20 / camel-load
of linen
1 / beehive 0.5 /
animal
0.1 / vine Discrimi
natory
between
1/7 and
2/5
Malatya (1560) 50 6 [none specified] 1 / beehive 0.5 /
animal
0.03 /
vine
uniform
1 / 5
Notes: All monetary values are in the Ottoman currency of Ak)e. D5n"m is a measure of land.
Because of their customized nature, the rates for enterprise taxes are not reported.
Sources: Ottoman provincial k#n(nn#mes. Akg"nd"z (1990), Barkan (1943).
37