Revolt on the Nile: Economic Shocks, Religion
and Political Power
Eric Chaney
August 15, 2011
Abstract
Can economic shocks increase the political power of religious leaders? I
investigate this question using over 700 years of Nile ‡ood data. I …nd that
deviant Nile ‡oods decreased the probability of change of Egypt’s highestranking religious authority by resignation or dismissal by roughly one-half. To
better understand if this relationship represents an increase in political power,
I examine historical evidence and the e¤ects of Nile shocks on the construction
of religious structures. In addition, I use variation in the popular followings of
religious leaders, data on the tenure of Egypt’s highest-ranking tax collector
and data on solar eclipses and earthquakes. Together the evidence suggests
that deviant Nile ‡oods led to an increase in the political power of religious
authorities that was rooted in their control over popular support.
Harvard University. I thank CREI for hosting me while part of this project was carried out and
numerous individuals and seminar participants for helpful suggestions. The library sta¤ at the Bibliothèque Nationale de France, Manuscrits Orientaux greatly facilitated the Nile data transcription.
Any remaining errors are my own.
1
Recent research has stressed the role politically in‡uential groups play in determining institutional and economic outcomes. This research suggests the importance
of identifying politically powerful groups and the determinants of their in‡uence in
order to understand the development and evolution of a society’s institutions. Although economists have studied the political and institutional impact of a wide range
of interest groups, the role of religious organizations remains poorly understood.1
Adam Smith (2009 [1776], p. 284) was an early proponent of the political importance of religious organizations, claiming that “the authority of religion is superior to
every other authority.” Marx (1982 [1844], p. 42) echoed Smith’s insights, implying
that religious leaders play a role in discouraging institutional change by preaching
obedience.2 More recent investigations have provided systematic evidence suggesting the political relevance of religion both historically and today (Lipset 1959, 1994;
Barro 1999). Despite such advances, both the extent to which religious leaders wield
political power and the determinants of this in‡uence are unclear.
In this paper, I use a historical dataset containing over 700 years of Nile ‡ood
data to investigate how economic downturns a¤ected the political power of premodern Egypt’s highest-ranking religious authority. I …nd evidence that Nile-induced
economic shocks increased this leader’s political power. I also …nd evidence that this
increase in political power was rooted in this leader’s control over popular support.
To examine the e¤ect of deviant Nile ‡oods on the political power of Egypt’s
highest-ranking religious authority, I begin from Acemoglu and Robinson’s (2006, p.
173) de…nition of political power. They de…ne political power as “a measure of how
in‡uential a particular group (or individual) is [...] when there is con‡ict over which
policy should be implemented.”
While it is di¢ cult to measure precisely an individual’s ability to in‡uence policy,
1
For the institutional and economic importance of politically in‡uential groups see, among a
large literature, Grossman and Helpman (1994), Dixit and Londregan (1995), Krusell and RíosRull (1996) and Acemoglu et al. (2005). For the literature on the interface between religion and
politics see section 1 below.
2
See also Montesquieu (1989 [1748], p. 61) and De Toqueville (1983 [1856] p. 152; 2003 [1840],
pp. 111, 347).
2
pre-modern Egypt provides a promising venue to examine the e¤ects of economic
shocks on the political power of religious leaders for three reasons. First, during
the period covered by the data, the head judge (Egypt’s highest-ranking religious
authority) had a policy agenda that often con‡icted with that of the sultan (the
leader of Egypt’s military). The areas of con‡ict between the head judge and the
sultan provide an opportunity to gauge the political power of the head judge. Second,
the pre-modern Egyptian economy was heavily dependent on the annual Nile ‡ood.
Deviant Nile ‡oods provide arguably exogenous variation in economic outcomes.
Third, scholarship suggests that results obtained in pre-modern Egypt are likely to
be relevant for a sizeable subset of the pre-modern world (e.g. North et al. 2009, p.
39).
If I observed policy implementation in each period, measuring the e¤ect of deviant Nile ‡oods on the political power of the head judge would be straightforward.
An increase in the implementation of the judge’s desired policies would constitute
conclusive proof of an increase in his political power. Although systematic data
detailing policy implementation in pre-modern Egypt are not available, historical
evidence suggests that a ceteris paribus increase in the sultan’s implementation of
the judge’s desired policies made the head judge less likely to oppose the sultan.
This decrease in opposition, in turn, made the sultan less likely to dismiss the head
judge or for the head judge to resign. I begin by using the probability of a change
in the head judgeship as a proxy for the head judge’s political power. Results show
that this probability decreased by roughly one-half during deviant Nile ‡oods and
are consistent with a Nile-induced increase in the head judge’s political power.
This empirical result does not conclusively demonstrate that deviant Nile ‡oods
increased the political power of the head judge. One alternative explanation is that
the head judge was an irrelevant bureaucrat and during periods of Nile-induced
famine the sultan was less likely to dismiss any member of the bureaucracy. Another
is that deviant Nile ‡oods led to an increased demand for religious services or for
religiously-provided social services. In other words, deviant Nile ‡oods may have
a¤ected the probability of a change of the head judge through other channels rather
than by increasing the implementation of his preferred policies (and consequently his
3
political power).
To further investigate the extent to which the decrease in the dismissal probability
stemmed from increased political power I investigate the e¤ect of Nile shocks on the
implementation of one of the head judge’s preferred policies. To do this, I gather
data on the construction dates of Egypt’s most important monuments. I …nd that
construction of religious structures relative to secular structures increased during
periods of deviant Nile ‡oods. The head judge bene…tted from increased allocations
to religious structures and thus preferred the allocation of a larger share of resources
to such constructions. Consequently, this result provides evidence for a Nile-induced
increase in at least one of the head judge’s preferred policies.
I then use the historical record to better understand these two results. The
historical evidence suggests that the sultan increased the implementation of the head
judge’s preferred policies during deviant Nile ‡oods in order to prevent the judge from
endorsing a revolt. The cost of preventing a judge-sanctioned revolt was higher during
Nile shocks because these shocks increased the probability that such a revolt would
succeed. Qualitative evidence suggests that both the decrease in the probability of
changes in the judgeship and the increase in resources allocated to religious structures
re‡ect this Nile-induced increase in the head judge’s bargaining power.
To test the empirical relevance of this conceptual framework, I employ three
additional empirical strategies. First, I use variation in the number of followers of a
head judge to test whether the negative relationship between deviant Nile ‡oods and
judge dismissal is stronger for judges who had larger followings. The data show this
is the case, linking the negative relationship between deviant Nile ‡oods and judge
changes to a judge’s popular in‡uence.
Second, I investigate the e¤ects of Nile shocks on the probability of change of
Egypt’s highest-ranking tax collector. Unlike the head judge, Nile shocks made
changes in this post more likely. This result is not consistent with explanations
suggesting that Nile shocks made the sultan less likely to dismiss any member of the
bureaucracy. In addition, this result provides suggestive evidence for a Nile-induced
increase in the threat of popular revolt since it is consistent with historical evidence
that the sultan attempted to appease Nile-induced rioters by dismissing such o¢ cials.
4
Third, I use data on solar eclipses and earthquakes to investigate whether the
results are due to increased demand for religious services or religiously-provided social services rather than because the head judge could alleviate the threat of revolt
by their in‡uence over the populace. Historical evidence suggests that both solar
eclipses and earthquakes increased the religiosity of the populace and their demand
for religious services such as collective prayers and fasts. In addition to increasing
religiosity, earthquakes led to increased demand for religiously-provided social services. There is no evidence, however, that eclipses or earthquakes led to an increase
in the probability that a head-judge sanctioned revolt would succeed.
Results show that neither eclipses nor earthquakes signi…cantly a¤ected the probability of judge change or the construction of religious structures relative to secular structures. In other words, eclipses seemed to have increased religiosity and
earthquakes increased both religiosity and the demand for religiously-provided social
services without a¤ecting the probability of a change in the head judgeship. This
evidence casts doubts on explanations attributing the results to an increase in the
demand for religious services or an increase in the demand for religiously-provided
social services.
By providing evidence that religious organizations historically wielded political
power, the …ndings in this paper complement the research of Greif (1994, 2006) and
Tabellini (2008, 2010) which suggests the importance of culture (here religion) in
determining political and institutional outcomes. In this sense, the paper also adds
to a literature emphasizing the economic importance of religion and culture (Barro
and McCleary 2003, 2005, 2006; Guiso et al. 2003, 2006; Iyigun 2008).
The results also support a literature stressing the political importance of religious
leaders in the Islamic world.3 This literature complements the discussion in North et
al. (2009, pp. 39, 45) that suggests that religious elites historically used their popular
in‡uence to perpetuate rent-extracting institutional equilibria. Although the extent
to which religious leaders have sti‡ed democratic change historically remains a topic
3
Among a large literature see Lewis (1953), Gibb (1955), Lapidus (1975), al-Sayyid-Marsot
(1973), Burke (1987) and Coşgel et al. (2009). For negative correlations between democracy and
the percentage of the population that is Muslim today see Barro (1999) and Fish (2002).
5
for future research, the results suggest that a better understanding of the political role
of religious organizations may both enhance our knowledge of institutional evolution
in the past and shed light on political equilibria today.
The remainder of the paper proceeds as follows. The …rst section brie‡y reviews
the literature on the interplay between religion and politics and examines the political
role of religious leaders in pre-modern Islamic Egypt. The second section provides
historical evidence suggesting that deviant Nile ‡oods increased the political power
of the head judge and that this increase was rooted in the judge’s popular in‡uence.
This section also develops a conceptual framework to better understand the use of
head judge changes as a proxy for the head judge’s political power. The third section
describes the data. The fourth section documents the e¤ect of Nile-induced famines
on the probability of head judge changes and investigates the e¤ect of deviant Nile
‡oods on the construction of religious structures. The …fth section shows that the
negative relationship between deviant Nile ‡oods and head judge changes is related
to a judge’s popular following, documents that deviant Nile ‡oods increased the
probability of change of the head tax collector, and uses data on solar eclipses and
earthquakes to investigate the plausibility of alternative explanations for the results.
The sixth section concludes.
1
Religion, Politics and Islamic Egypt
Barro (1999) notes that despite evidence that religious a¢ liation is a determinant
of democracy “the theory of the interplay between religion and political structure
is even less developed than other aspects of the theory of democracy.” Baruch de
Spinoza (1632-1677) –an in‡uential enlightenment thinker–was an early proponent
of one of the few existing frameworks that explores the political e¤ects of organized
religion. He has been paraphrased as arguing that “organized religion is nothing
other than a political and social device instituted to serve the well-being of men
[...Men use it] to astonish, terrify and elate the people and by this means manipulate
them [...T]he inevitable result [of this] is oppression and despotism”(Israel 2001, pp.
6
701, 708-709).
Although subsequent studies have cast doubt on the basic premise of such frameworks (i.e. that religion is the product of widespread manipulation), scholars continue
to stress that the political in‡uence of religious authorities rests on their control over
the actions of their “constituencies.”Gill (1998, p. 61) argues that “[r]eligious leaders
provide the state with [the] legitimation it needs to lower the cost of rule.”Murphy
and Shleifer (2004) view religious leaders as “brokers creating the networks [...] that
bind members together [...and then] “rent out”their networks.”Barro and McCleary
(2005) note that the government “might want to use religion as a cooperative force
for controlling the citizenry.”4
Qualitative evidence suggests the relevance of this conceptual framework in the
Islamic world. Scholars of this region have long claimed that religious elites worked
to check the absolute power of the sovereign. Al-Sayyid Marsot (1973, pp. 133134) provides a clear exposition of this hypothesized function. She notes that in
Muslim societies religious elites “had it in their power to rouse or placate public
opinion [...] [i]t was through using the threat of rousing the mob that the ‘ulamā’
[religious elites...] could restrain the authorities, who recognized the dangers behind
the threat.” In this framework, religious leaders maintained the allegiance of their
followers through ideology (religious beliefs) and by protecting them from extreme
levels of predation by the military (e.g. high levels of taxation and con…scations).
Religious and military leaders often con‡icted over policy implementation. In
pre-modern Islamic Egypt, military leaders were willing to make concessions and
implement policies that bene…tted religious leaders at the expense of the military in
order to prevent rebellion.5 In return for these concessions, religious leaders preached
obedience and helped channel “all potential popular resistance into actions which
bore no threat to [the military leaders’] hegemony” (Lapidus 1984, p. 183). Poli4
For an overview of the social science literature on religion see Iannaccone (1998) and Stark and
Finke (2000). For the medieval Catholic Church as an economic …rm see Ekelund et al. (1996).
5
Throughout the remaining historical section I focus on the Mamluk period (both before and
after the Ottoman conquest) due to source constraints. For the similarity of the institutional
framework in earlier (Islamic) periods see Lapidus (1975).
7
cies that favored religious leaders at the expense of the military included monetary
payments, the military enforcement of prohibitions of “morally deviant” behavior
(such as wine drinking and prostitution) and the construction of religious structures.
Military leaders generally aimed to minimize such concessions since they reduced
the amount of resources available to the military. For example, resources allocated
to religious structures were held in religious endowments (waqfs) which the military
could not tax. Similarly, taverns, prostitution and hashish consumption all yielded
sizeable tax revenues. These tax revenues sharply declined if such activities were
prohibited (Lapidus 1984, p. 172).
As pre-modern Egypt’s most in‡uential religious authority, the head judge pressed
the military for the implementation of policies that bene…tted religious leaders at the
expense of military elites. This judge stood at the head of his “guild”(madhhab) of
religious leaders and had the authority to appoint and dismiss lower ranking religious
leaders.
The sultan appointed and dismissed the head judge and was in theory subject to
the head judge’s rulings as long as the head judge remained in o¢ ce (consequently
the sultan often dismissed the head judge when he went against the judge’s rulings).
While the sultan had an incentive to appoint a pliable “yes man” to the judiciary
–and the historical record provides examples of such judges– such “yes men” were
not likely to retain popular in‡uence.6 Perhaps for this reason, the sultan appears
to have generally appointed head judges from a list of candidates presented by a
council of religious leaders, or to have sought out leaders with substantial popular
followings.7
Consequently, most head judges enjoyed “vast personal authority among [religious
leaders]” and could provide “massive popular backing” (Lapidus 1984, pp. 134,
136). One prominent scholar suggests that “[i]f the holders of any one o¢ ce can be
considered the voice of the [religious leaders] [...] the chief justices can”(Petry 1981,
p. 231). Since the head judge seems to have represented the interests of a sizeable
6
7
See Maqrizi (1997 [1441], VII, p. 450) and Petry (1981, p. 320).
See Petry (1981, p. 315) and Maqrizi (1997 [1441], II, p. 228; III pp. 238-239; IV, p. 101 ).
See Maqrizi (1997 [1441], III, pp. 383, 384, 353) for the importance of a judge’s popular following.
8
subset of Egypt’s religious leaders, I refer to the head judge and religious leaders
interchangeably throughout the paper. For similar reasons I also refer to the sultan
and military elites interchangeably.
2
Nile Floods and the Political Power of the Head
Judge
2.1
The Nile Flood
Agriculture in the Nile basin was historically heavily dependent on the annual Nile
‡ood. A complex system of dikes and irrigation networks helped harness the ‡ood’s
agricultural potential, making Egyptian agricultural yields some of the highest in the
pre-modern world.8
The Nile’s annual ‡ood usually began in earnest in July and had crested (at the
latest) in October. Once the Nile began to subside in October, peasants across the
basin sowed a wide variety of crops including wheat, barley, broad beans, chick-peas
and lentils. Although other crops were grown and harvested on some land throughout
the year (with the aid of waterwheels and other irrigation devices), the Nile ‡ood
irrigated the vast majority of crops. These crops were harvested in April and May.
Ethiopian monsoon runo¤ determined the size of the annual Nile ‡ood. Recent
research has linked variation in this monsoon runo¤ to sunspots (Ruzmaikin et al.
2006), suggesting the Nile’s summer ‡ood level can be considered exogenous. I
maintain this exogeneity assumption throughout the paper.
When the Nile’s summer ‡ood signi…cantly deviated above or below its optimal
level (these deviant episodes are henceforth referred to as Nile failures, deviant Nile
‡oods or Nile shocks) agricultural output sharply dropped the following harvest.9
8
This section draws on Cooper (1976), Frantz-Murphy (1986), Tsugitaka (1997). See also Borsch
(2005, pp. 34-39) for a detailed discussion.
9
Here (and below) I focus on Nile droughts for expositional clarity. See Raymond (1973, p. 83);
Sabra (2000, pp. 152-153) or Petry (1994, p. 105) for evidence that excessive Nile ‡oods had similar
e¤ects.
9
While some Nile failures caused little more than passing shortages, large Nile deviations induced full-blown famines. More severe famines led to widespread unrest and
armageddon-like conditions. Death rates during these episodes could reach rates of
over 500 people a day in Cairo (Maqrizi 1997 [1441], IV, p. 376).10
The severity of a Nile shock primarily depended on the extent to which the
Nile deviated from its optimal level. The historical record is somewhat ambiguous
regarding the frequency of shocks severe enough to cause crises. One scholar suggests
that such shocks occurred once every 8 years between 949 and 1233 CE (Hassan
2007) while another suggests these occurred every 17 years between 1600 and 1800
CE (Raymond 1973, pp. 81-106). These di¤erences in the frequency of Nile-induced
crises may have been driven by long-term variation in Nile ‡ood cycles, improvements
in agricultural technology or by di¤erences in the precision with which the historical
sources recorded such events.
A slow or excessively rapid Nile rise during the summer months unsettled the
population. Food prices rose as individuals hoarded grain in preparation for shortages
in the following year. Consequently, although the agricultural e¤ects of a one-shot
Nile failure did not obtain until harvest the following year the “treatment e¤ect”
of Nile failure began before the physical absence of foodstu¤s occurred. If the Nile
failed, prices of wheat could double or even quadruple. The price increases began
at the …rst sign of an abnormally slow (rapid) rise of the Nile during the summer
months prior to the Nile’s crest (below I assume, following historical evidence, that
evidence of Nile ‡ood failure began to be revealed in July).
Grain prices decreased the following year if the Nile ‡ood was adequate, although
prices generally remained elevated until the harvest was collected in April/May of
the following year (Maqrizi 1994 [1405], p. 51). In other words, the most severe
e¤ects of a one-shot Nile failure on grain prices appear to have lasted roughly from
July of the solar year of the Nile failure (denote this year by t) through June of the
following solar year (t+1). Grain prices remained abnormally high, however, through
April/May of the solar year t+2 (when the harvest irrigated with the ‡ood in year
10
Estimates of Cairo’s population range between 500,000 and 600,000 in the 14th century (Ray-
mond 2002, p. 136).
10
t+1 was collected).
2.2
Nile Floods, Revolt and the Head Judge
Historical sources suggest that Nile failures increased the probability that a headjudge-sanctioned revolt against the sultan would succeed. The historical evidence
suggests that Nile failures increased the probability that a revolt against the sultan
would succeed by increasing the number of individuals that would follow his call to
revolt. This qualitative evidence is consistent with two strands of literature that
suggest i. that economic shocks can help solve the collective action problem and ii.
that the leadership of religious leaders was necessary for the populace to coordinate
a widespread revolt.11
Deviant Nile ‡oods appear to have helped solve the collective action problem
by decreasing the opportunity cost of rioting. More individuals were willing to riot
during Nile failures since if they did nothing they would starve whereas if they rioted
they would receive a share of the spoils if the riot was successful (Lapidus 1984,
p. 168). These individuals organized localized “bread riots” and demanded that
the sultan provide food and remove “obnoxious o¢ cials” such as tax collectors and
market inspectors (Lapidus 1984, p. 147). Although the sultan often acquiesced to
the demands of rioters by dismissing o¢ cials, distributing food and importing grain,
riots often continued despite these concessions, apparently because the sultan could
not obtain su¢ cient foodstu¤s to feed the entire population. These Nile-induced
bread riots were more likely to remained localized if the head judge did not help
the rioters coordinate.12 One prominent historian has gone as far as to suggest that
11
For the role of economic shocks in aiding to solve the collection action problem see Acemoglu
and Robinson (2006, pp. 123-128, 144-145). For the importance of leadership to avoid coordination
failures see Acemoglu and Robinson (2006, p. 125). For the role of religious leaders in facilitating
collective action and their crucial role in preventing (or encouraging) coordination failures both
generally and in Islamic Egypt see Gill (2001) and Lapidus (1984, p. 153).
12
For an example of a judge aiding coordination across city quarters see Lapidus (1984, p. 166).
For coordination in the Friday sermon (khut.ba) and at prominent mosques and educational centers
see Al-Sayyid Marsot (1972, p. 153; 1973, p. 133) and Raymond (1974, p. 432). For the coordi-
11
without the judge’s aid such widespread coordination would have been “impossible”
(Lapidus 1984, p. 190).13
It is important to stress that the historical sources also suggest that Nile failures
may have increased the probability that a head-judge-sanctioned revolt against the
sultan would succeed through two additional channels. First, some passages suggest
that Nile failures may have increased the religiosity of the populace making them
more likely to follow the head judge’s dictates.14 Second, Nile failures may have
also increased the probability that a head-judge-sanctioned revolt against the sultan
would succeed by increasing the probability of civil war among military factions. One
scholar has noted that in such “civil wars, the support of the [head judge], and with it
massive popular backing, was often of decisive importance”(Lapidus 1984, p. 134).
Thus, it is plausible that Nile failures increased the probability that a popular revolt
would unseat the incumbent sultan because deviant ‡oods exacerbated divisions
within the military (Sanders 1994, p. 3; Lev 1991, pp. 14, 76).
Regardless of the relative importance of each of these channels, the available
evidence suggests that Nile failures increased the probability that a head-judgecoordinated revolt would overthrow the incumbent sultan.15 To mitigate this danger,
the sultan granted concessions to the head judge in order to both dissuade him from
supporting a revolt and to enlist his aid in pacifying the populace. In return, the
head judge used his in‡uence to discourage revolt and organized activities to channel
the Nile-induced potential for popular uprisings into actions which did not threaten
the survival of the incumbent regime (Raymond 1974, p. 420). Such actions included
nating e¤ects of fatwas (legal opinions) see Maqrizi (1997 [1441], VI, pp. 12, 227), Lapidus (1984,
pp. 152-153) and Tyan (1960, pp. 116, 424).
13
For peasant assemblies and grain revolts and riots in Egypt see Sabra (2000, pp. 136, 142, 155)
and Shoshan (1980). See Grossman and Medoza (2003) for a theoretical discussion of how economic
downturns can lead to popular unrest and “appropriative competition”.
14
For studies linking economic shocks to increases in religious intensity see McCann (1999) or
Chen (2010).
15
See Maqrizi (1997 [1441], IV, p. 368) ; Shoshan (1993, p. 53) and Taghri Birdi (1976 [1468],
V, p. 120; VI, p. 63).
12
rain prayers and other activities to request divine aid.16
2.3
Revolt and Political Power: a Conceptual Framework
The increased probability that a judge-led revolt could unseat the incumbent sultan
seems to have translated into an increase in the head judge’s political power. The
historical evidence provided above suggests a simple conceptual framework for why
this was the case. Since the probability of success of a judge-sanctioned revolt increased during Nile failures, the sultan had to increase the implementation of the
head judge’s preferred policies (at least proportional to total output) in order to
increase his payo¤ and ensure that he did not support such an uprising.17
Qualitative evidence is consistent with this prediction and suggests that the sultan increased the implementation of the head judge’s preferred policies during Nile
failures. One historian has summed up this policy shift by noting that during periods of Nile failure “the sultan would bow to [...] pressure [from the head judge] and
enforce decrees against [...] prostitution, hashish eating, beer drinking, the wearing
of immodest or over-luxurious dress [or] Christian and Jewish functionaries lording
it over Muslims” (Irwin 1986, p. 50). The historical record also provides evidence
that the sultan increased payments and perquisites to religious leaders during Nile
failures.18
Importantly for the empirical analysis below, the historical record includes examples of the sultan increasing allocations to religious structures during periods of
16
For an example of a rain prayer see Taghri Birdi (1976 [1468], III , pp. 77-78). For its use
in calming bread riots see Taghri Birdi (1990 [1468], p. 234). For an example of a procession
see Shoshan (1993, p. 62). For public readings of Bukhari (a famous religious text) to calm the
su¤ering populace after the Nile had failed see Maqrizi (1997 [1441], IV, p. 366).
17
As long as the judge does not …nd it optimal to alter the existing institutional structure, this
prediction should also hold in a dynamic framework. Since there are only isolated and short-lived
examples of religious authorities establishing their own political entities I focus on the implications
of a static framework. See Acemoglu et al. (2010) for a related theoretical framework.
18
For an example of increased payments to the ulama following a Nile failure see Maqrizi (1997
[1441], IV, p. 269). For an example of increased perquisites see the e¤ect of Nile failure on the
ability of the ulama to ride horses in Taghri Birdi (1990 [1468], pp. 220, 238).
13
Nile failures. One such example occurred in 1420 CE when the Nile rose at an abnormally slow rate in July. In response, the sultan ordered the construction of a
mosque and the repair of another religious structure (Taghri Birdi 1976 [1468], III,
pp. 77-79). Similarly, during the deviant ‡ood spanning the year 1415-1416 CE the
sultan gave “the treasurer a large sum and ordered him to go down to Cairo and
distribute it among the mosques, college mosques, and monasteries” (Taghri Birdi
1976 [1468], III, p. 38). Such allocations particularly bene…tted the head judge and
other religious authorities, since they received stipends in return for overseeing and
sta¢ ng these religious structures.19
2.3.1
Political Power and Changes in the Judgeship
In the empirical section, I use the probability of a change in the head judge as a proxy
for the degree to which the sultan implemented the head judge’s preferred policies.
To better understand the use of this proxy, assume that in each period the sultan
proposed a vector of policies to be implemented and the head judge decided whether
to oppose or support the sultan’s proposals. If the head judge opposed a proposed
policy and the sultan decided to implement the policy nonetheless, the head judge
was dismissed or resigned.20
The historical record suggests that ceteris paribus increases in the implementation
of the judge’s desired policies made the head judge less likely to oppose the sultan’s
proposed policies and to resign or be dismissed. Ibn Hajar (1998 [1449], p. 392)
provides one concise example that illustrates this point. In this example the sultan
asked the head judge “to authorize the sale of some [religious endowments]. He [the
head judge] refused to authorize this sale and was consequently dismissed.” In this
and many similar cases it seems likely that had the sultan not asked the head judge
19
For evidence that religious structures disproportionately bene…tted the head judge and other
religious leaders see Amin (1980, pp. 113, 181). For evidence that they decreased resources available
to the military see Amin (1980, p. 279) or Lambton (1997, pp. 307-308).
20
There is overwhelming historical evidence that the head judge was often dismissed or resigned
in response to the sultan’s refusal to implement his preferred policies. For examples see Ibn Hajar
(1998 [1449], pp. 231, 246).
14
for this level of policy concessions (here the sale of religious endowments), the head
judge would not have been dismissed or resigned.
To illustrate the empirical implications of this discussion more formally, let Zt
denote a dummy variable equal to one if there is a deviant Nile ‡ood in Nile year t
(see below for a de…nition of Nile years). Let Dt (p; z) denote the potential outcome
(see Angrist and Pischke 2009 for an overview of the potential outcome framework
and notation) of the incumbent head judge at the start of Nile year t over the
following year where Dt = 1 if there is a change in the head judgeship and Dt = 0
otherwise. Here, a higher p denotes a mix of policy implementations (normalized
by total output) that gives the head judge a higher payo¤ and Zt = z denotes Nile
failure.
Since I assume Nile failures are exogenous E[Dt jZt = 1]
written as E[Dt (P1t ; 1)
E[Dt jZt = 0] can be
Dt (P0t ; 0)]. Here P1t represents the policy implementations
over Nile year t if the Nile failed and P0t represents such implementations when the
Nile did not fail. If in addition I assume that the “exclusion restriction” Dt (p; 0) =
Dt (p; 1) 8p holds, then E[Dt jZt = 1]
E[Dt jZt = 0] = E[Dt (P1t )
Dt (P0t )]. These
assumptions combined with historical evidence that pe > p ! Dt (e
p; z)
8p; pe allow me to conclude that E[Dt jZt = 1]
for E[P1t
Dt (p; z)
E[Dt jZt = 0] < 0 is su¢ cient
P0t ] > 0. In other words, a Nile-induced decrease in the head judge’s
dismissal probability can be taken as conclusive evidence for an average increase in
policy implementations (normalized by total output) favorable to the head judge
during Nile failures.
For the “exclusion restriction” (Dt (p; 0) = Dt (p; 1) 8p) to hold, it must be the
case that a head judge’s potential outcome for a given level of policy implementation
is unchanged by Nile failures. For now I maintain this assumption and further
investigate its relevance in sections 4.2 and 5.1.
15
3
The Data
To investigate the e¤ect of Nile failures on head judge changes I created a data set
of head judge changes using information provided in Ibn Hajar (1998 [1449]). He
provides the month and year of head judge changes. Although Ibn Hajar’s data are
thought to be generally reliable (Escovitz 1984, p. 5), information regarding how a
head judge left power is not available for the majority of the head judges. Throughout
the text I refer to these change as judge changes regardless of whether the judge died
in o¢ ce, was dismissed or resigned. In practice, the missing information on why a
judge change occurred does not threaten the empirical strategy below since I assume
that the Nile ‡ood level is exogenous.21
Ibn Hijazi (1470) provides the Nile ‡ood data. In the appendix I provide a
detailed description of both data sets and how I address the complications that arise
from the use of lunar years in the original sources. I also show in the appendix that
measurement error in the Nile ‡ood data attenuates the coe¢ cients of interest under
plausible assumptions.
I use the Nile year (which runs from July through June of the solar year) as the
unit of observation since each Nile ‡ood level treated these dates. In addition, I use
the mapping provided by Popper (1951, pp. 104-105 table 1) to convert the original
Nile measurements (in cubits and …ngers) to meters above sea-level.
Figure one presents the recorded yearly Nile ‡ood maximums and minimums as
reported by Hijazi. The horizontal axis marks CE years. The left hand vertical
axis details the ‡ood level in meters above sea level and the right hand vertical axis
measures the ‡ood level in cubits and …ngers. On the right hand axis only the level
of 16 cubits (“plenitude”) is ticked and a dark horizontal line marks this level. This
level was considered the optimal ‡ood level at the start of the sample.
Examination of …gure one reveals that the Nile ‡ood level slowly trended upwards
over time. Sediment accumulation caused both the rise of the Nile bed and the sur21
One might worry that deviant Nile ‡oods increased the probability of a judge dying in o¢ ce.
Since in practice I …nd a negative relationship between deviant Nile ‡oods and judge changes such
an e¤ect would bias the results towards zero.
16
rounding land (Popper 1951, pp. 241-247; Borsch 2000). For this reason, throughout
the empirical section I consider the “real”Nile ‡ood deviation by examining its deviation from a time trend. I prefer the use of such a trend over autoregressions or
moving averages since it can be consistently estimated. The point estimates using
residuals from autoregressions or moving averages, however, generally yield qualitatively similar results to those reported. It is reassuring to note that, at the start
of the sample, the linear trend approximately coincides with the level of 16 cubits.
This suggests that the linear trend provides a reasonable estimate of the optimal Nile
‡ood level in each period.
4
Nile Floods and Political Power
4.1
Head Judge
I begin by examining the relationship between head judge changes and Nile failures
using the Nile year as the unit of observation. Figure two provides a non-parametric
estimate of the relationship between head judge changes and the deviation of the Nile
‡oods from a linear trend (I use a lowess smoother with the bandwidth set to 0.8).
Vertical lines mark the upper and lower cuto¤s de…ning the top 10% of Nile ‡ood
deviations. The negative relationship suggests that Nile ‡ood deviations a¤ected the
probability of judge dismissal.
To investigate this relationship formally, I create a dummy variable F ailuret equal
to one if the Nile ‡ood residual from a linear trend is in the upper 5% or lower 5%
of the ‡ood distribution.22 This “extreme value” speci…cation closely corresponds
with historical and empirical evidence that only deviant Nile ‡oods signi…cantly
empowered the head judge.
The baseline regression is the linear probability model of the form:
22
The 10% cuto¤ was chosen using the historical evidence that deviant Nile ‡oods occurred
approximately every 8 years for a subset of the period covered by the data. Using a 5% cuto¤
yields similar results.
17
Judgetd =
where t indexes Nile years and
d
d
+ F ailuretd + "td
(1)
denote decade dummies.23 The variable Judgetd
is a scaled dummy variable (to facilitate presentation) equal to one hundred if the
head judge at the start of Nile year t is dismissed on the interval (t,t+1). This
type of speci…cation is closely related to duration models and particularly suits this
empirical setting (see Lancaster 1990, pp. 10-13).
OLS estimates of equation (1) are reported in table one. The …rst panel uses the
Nile maximum to calculate the dummy variable F ailuretd . In this panel, the …rst column omits covariates whereas the second column includes decade dummy variables.
The third column includes a lead and a lag of the dummy variable F ailuretd . In
all speci…cations, the estimated relationship is negative and statistically signi…cant
(Newey-West standard errors with 7 lags are reported in parentheses) and shows that
the e¤ect of Nile failures are concentrated in the impact year.24
Columns (4)-(6) of table one report the results using the disaggregated data at
the lunar month level. Here the dependent variable is equal to one hundred if the
incumbent judge at the start of lunar month h is dismissed on the interval (h,h+1).
The point estimates are again statistically signi…cant (Newey-West standard errors
with 17 lags are reported in parentheses).
In the second panel, I use the Nile minimum to construct a placebo F ailuretd .
Rainfall in the Nile’s equatorial catchment area (which includes Western Kenya,
Uganda, Rwanda, Burundi, the Central African Republic, the Republic of Congo
and the Democratic Republic of Congo) determined this minimum (as opposed to the
Ethiopian monsoon runo¤ which determined the maximum). As …gure one makes
clear, the Nile’s yearly minimum and maximum levels are imperfectly correlated
(Hassan 2007).
The placebo F ailuretd is equal to one if the Nile ‡ood minimum was in the top
5% or lower 5% of the minimum ‡ood distribution. Results using this placebo help
23
24
Probit or logit speci…cations yield qualitatively similar results.
1
The truncation parameter was chosen using the rule of thumb m = 0:75T 3 rounded to the
highest integer. Standard errors are virtually identical for alternative m.
18
to address concerns that the results are spurious (e.g. driven by systematic mismeasurement of the optimal ‡ood level). Since the Nile ‡ood maximum determined
agricultural productivity whereas the Nile minimum had no e¤ect on this productivity, I expect the Nile ‡ood minimum to not be related to head judge dismissals.
Results in panel two of table one are consistent with this prediction.25
The estimated magnitudes of the relationship between deviant Nile ‡oods and
head judge dismissal are both statistically signi…cant and economically meaningful.
The smaller (in absolute value) point estimates in panel one suggest that Nile failures
decreased the yearly probability of dismissal by approximately 9 percentage points
and the monthly probability of dismissal by roughly one percentage point (or using
the binomial approximation by approximately 12 percentage points over a lunar
year). Since the monthly probability of judge dismissal is 2.2 percent and the yearly
probability is 21 percent this suggests that deviant Nile ‡oods reduced the probability
of judge dismissal by roughly one-half.
4.2
Construction of Religious Structures
Under the assumptions detailed in section 2.3, the observed decrease in the head
judge’s dismissal probability is su¢ cient to conclude that deviant Nile ‡oods increased the head judge’s political power. In this section, I provide additional empirical evidence that is consistent with a Nile-induced increase in the judge’s political
power. I use data on the construction of religious structures to show that Nile failures
increased the implementation of at least one of the head judge’s preferred policies.
I use Creswell (1919) to compile a data set of the construction dates of major
secular and religious structures. Creswell’s data set includes the lunar dates of construction for 132 secular and religious monuments built in Egypt that overlap with
the ‡ood data.26 The construction dates are drawn from both written sources and
25
The drop in the number of observations is due to missing data in the Nile minimum series.
Limiting both series to years in which both the Nile maximum and minimum are provided yields
similar results.
26
The vast majority of these monuments were constructed in Cairo.
19
archaeological work.
Using Creswell’s classi…cation of structures, I create a variable measuring the total number of secular and religious structures on which construction began in each
lunar year (the majority of these construction projects seem to have lasted less than
one lunar year). These structures include Cairo’s most important religious monuments (e.g. mosques or madrasas) as well as secular structures (e.g. palaces or the
city walls).27 Military elites funded almost all the structures provided by Creswell.
Creswell’s data set likely provides a reasonable approximation to the population of
important structures built over the period covered by the data.28
To investigate the e¤ect of Nile ‡ood deviations on the rate of construction of
religious structures I construct a measure of the number of religious structures built
in each Nile year. Since the construction dates are given by lunar year, I take
the weighted average of the total number of religious buildings begun in the lunar
years that overlapped with at least part of a given Nile year (where the weights
are the percent of the Nile year occupied by the respective lunar year). I do the
same for secular structures. To make the results comparable between religious and
secular structures, I standardize the number of buildings constructed each year by
subtracting the respective mean and dividing by the standard deviation.
As noted above (section 2.3), resources allocated to religious structures such as
mosques and madrasas disproportionately bene…tted the head judge because these
structures were sta¤ed and managed by the head judge and other religious leaders.
Secular structures did not directly bene…t the head judge to the same degree (and
many secular structures did not bene…t him at all). Consequently, the head judge
preferred that a larger share of new constructions be dedicated to religious structures.
Military elites, however, preferred the opposite. The frequent con‡icts that arose over
27
Secular monuments include aqueducts, caravansaries, city walls, bridges, cisterns, citadels,
gates, halls, hospitals, nilometers and palaces. Religious monuments include convents, mosques,
madrasas, khanqas, mashhads, mausolems and zawiyas. Results are qualitatively similar if buildings
associated with death (mausolems, mashhads and zawiyas) are omitted.
28
Such large structures were likely to be noted in the historical sources. In addition, many of
these structures have survived until today (Crecelius 1986, p. 187).
20
the implementation of this policy has led one scholar to view the amount of resources
allocated to religious structures and other religious endowments in each period as
“the most signi…cant indicator of the relative positions of the religious and military
establishments”(Nielsen 1985, p. 120).
I present regression results investigating the e¤ect of Nile failures on the construction of religious structures in table 2. The dependent variable in the …rst column is
100 times the standardized number of religious buildings constructed in each year.
The empirical speci…cation is identical to that in equation (1) substituting this dependent variable for the dummy Judgetd . The point estimate in the …rst column
shows that during Nile failures construction of religious structures increased by 0.12
standard deviations (0.05 buildings).
Although this result is not statistically signi…cant at conventional levels, the
result in column (1) of table 2 probably underestimates the increase in the head
judge’s desired policy since Nile failures decreased the total allocation of resources
to new construction. Results in column (2) are consistent with this hypothesis and
suggest that Nile failures decreased resources allocated to the construction of secular
structures by 0.14 standard deviations (0.03 buildings).
Inasmuch as this decrease in secular structures provides a counterfactual for what
would have happened to the construction of religious buildings during deviant Nile
‡oods had the political power of the head judge not increased, the di¤erence between
results in columns (1) and (2) provided in column (3) gives the causal e¤ect of
Nile-induced increases in the judge’s political power on the construction of religious
structures. This result in column (3) suggests that deviant Nile ‡oods increased the
construction of religious structures by 0.27 standard deviations (0.11 buildings) and
is statistically signi…cant at the 5% level. Results in panel 2 show that the placebo
minimum Nile ‡ood are negative and statistically insigni…cant.
The results in this section provide direct evidence for an increase in the implementation of one of the head judge’s preferred policies during periods of Nile failure.
This increase coincides with the observed decrease in the probability of head judge
changes and is consistent with both the historical evidence and the conceptual framework developed in section 2.
21
5
Popular Support, Judge Changes and Alternative Explanations
The historical record suggests that Nile ‡oods increased the head judge’s political power by increasing the probability that a head-judge sanctioned revolt would
succeed. In this section, I investigate the empirical relevance of this conceptual
framework. I begin by providing evidence linking the Nile-induced decrease in the
probability of change of the head judge to the size of his following. I then investigate
the empirical relevance of alternative explanations for the results by presenting evidence that Nile shocks increased the probability of change of the head tax collector
and by using data on solar eclipses and earthquakes.
To empirically examine whether the head judge’s increase in political power was
related to his control over popular support I use three sources of variation in popular
adherence to the head judge’s “denomination”(school of law). First, I note that the
Egyptian population slowly converted to Islam, giving the head judge an increasingly
large “constituency.” Second, I observe that during the Fatimid dynasty (969-1169
CE) both the ruling dynasty and the head judge were Shia (with a few isolated
exceptions) whereas the overwhelming majority of Egypt’s Muslim population was
Sunni. Third, I use the fact that after 1265 CE the military appointed four head
judges. These four head judges represented the four main Sunni law schools (Sha…i,
Maliki, Hana… and Hanbali), each of which had di¤erent levels of adherence among
Egyptian Muslims.
The empirical strategy can be understood as follows. Suppose the extreme case
in which no one in the populace shares the head judge’s religious a¢ liation. Since
such a leader does not control any popular support, I would not expect Nile failures
to increase his political power. Similarly, if di¤erent head judges have di¤erent
“stocks” of adherents I expect the head judge with the larger stock to experience a
larger decrease in the probability of dismissal than judges with smaller stocks.
Table three presents the results. The …rst column presents a regression identical to equation (1) adding the interaction term F ailuret
22
P ercentM uslimt where
P ercentM uslimt is the percentage of the population that is Muslim in Egypt each
year as given in Bulliet (1979). Column (2) includes the term F ailuret Shiat , where
the variable Shiat is equal to one if the head judge was Shia. The F-statistic and
corresponding p-value testing the hypothesis that both interaction terms are zero is
given in the row labeled F-Stat.
Results are consistent with both existing scholarship and the historical narrative
developed in section 2 and suggest that deviant Nile ‡oods decreased the probability
of change more for judges with larger followings. Results in column (2) suggest that
Nile failure actually increased the head judge dismissal probability when the head
judge was Shia. For all other head judges (who were Sunni like the overwhelming
majority of Egypt’s Muslim population), the results suggest that the e¤ect of Nile
failures on judge dismissal became increasingly negative as the population converted
to Islam.
Column (3) limits the sample to after 1265 CE, and reports results running equation (1) for the Sha…i, Maliki, Hana… and Hanbali head judges using a seemingly
unrelated regression framework. Although statistics of the distribution of Egypt’s
Muslims across the four law schools during this period do not exist, there is broad
scholarly agreement that the Egyptian population was “mainly Shā…’¯¬”(Nielsen 1984,
p. 172) during the period covered by the data after 1265 CE. This popular in‡uence meant that the Sha…i head judge was the “highest ranking” of the four head
judges (Escovitz 1984, p. 25) and led one medieval observer to note that any sultan
who did not adhere to the Sha…i law school was “quickly ousted or killed” (cited in
Jackson 1996, p. 33). Escovitz (1984, p. 25-27) suggests that, after the Sha…i judge,
the Maliki judge was the most in‡uential, followed by the Hana… and the Hanbali
judges.
The coe¢ cient for the Sha…i head judge is given in the row marked F ailuret in
column (3) of table three. Those for the Maliki, Hana… and Hanbali head judges
are provided in the fourth, …fth and sixth rows. The F-statistic testing the equality
of the coe¢ cients with its corresponding p-value is again given in the row labeled
F-Stat. Results show that Nile failures were associated with statistically signi…cant
decreases in the probability of dismissal of the Sha…i and Maliki head judges (who
23
had the greatest popular in‡uence). The data reject the null-hypothesis that Nileinduced famines had similar e¤ects on the probability of dismissal of all the head
judges.
5.1
Alternative Explanations
If Nile failures are exogenous, then E[Dt jZt = 1]
E[Dt jZt = 0] = E[Dt (P1t ; 1)
Dt (P0t ; 0)]. This last term can be written as E[Dt (P1t ; 1) Dt (P0t ; 1)]+E[Dt (P0t ; 1)
Dt (P0t ; 0)]. The …rst term (E[Dt (P1t ; 1)
Dt (P0t ; 1)]) gives the causal e¤ect of Nile
failures on judge changes through their e¤ects on concessions to the head judge
during periods of Nile failure whereas the second term (E[Dt (P0t ; 1)
Dt (P0t ; 0)])
gives the “bias” stemming from the fact that Nile ‡oods may have a¤ected head
judge dismissal through channels other than increased concessions. Up until now
I have assumed that E[Dt (P1t ; 1)
E[Dt (P0t ; 1)
Dt (P0t ; 1)] < 0 ! E[P1t
P0t ] > 0 and that
Dt (P0t ; 0)] = 0. In this section I will address both assumptions in
greater detail.
I begin by assuming that E[Dt (P1t ; 1)
and concentrate on the term E[Dt (P0t ; 1)
Dt (P0t ; 1)] < 0 ! E[P1t
P0t ] > 0
Dt (P0t ; 0)]. Here, the concern is that
E[Dt (P0t ; 1)] < E[Dt (P0t ; 0)] since the opposite would “bias” the estimated coe¢ cients upwards. In other words, it is plausible that Nile failures made head judge
changes less likely for reasons unrelated to the implementation of his preferred policies. Theoretical and historical considerations suggest two reasons this might be the
case.
First, the sultan might have been less likely to dismiss all government sta¤ during
Nile failures. To investigate this possibility, I examine the e¤ects of Nile shocks on
the probability of change of Egypt’s highest-ranking tax collector.29 The regression
used is identical to equation (1) where I replace Judgetd with T axCollectortd . The
variable T axCollectortd is a dummy variable equal to one if the head tax collector
at the start of Nile year t changed over the following Nile year. Results presented in
29
The head tax collector was known as the vizier after 1250 (Martel-Thoumian 1991, p. 36).
Data on the tenure of viziers are drawn from ‘Abd al-Rāziq (1980).
24
column (4) of table 3 show that for the years for which data are available the head
tax collector was roughly 31 percentage points more likely to be changed during
Nile failures. Since the yearly probability of change for this tax collector was 44
percent, Nile failures increased this probability by roughly 70 percent. This result
provides suggestive evidence for a Nile-induced increase in the danger of popular
revolt since it is consistent with historical evidence that the sultan attempted to
appease Nile-induced rioters by dismissing tax collectors. In addition, it casts doubt
on explanations attributing the results to a decrease in dismissal of all bureaucrats
during Nile failures, as do the results for the head judge in column (3) of table 3.
Second, Nile failures might have led to a change in the populace’s preferences
resulting from a Nile-induced increase in religiosity. In other words, the populace
may have become more religious during Nile failures increasing demand for religious
services.30 Alternatively, Nile failures might have increased the need for religiouslyprovided social services. This increase in demand for religious services and/or religiously provided social services could have led the sultan to allocate more resources
to religious structures and made the sultan less likely to change the head judge
(because he was needed to oversee construction of religious structures, lead prayers
or provide social services administered by religious leaders) without increasing the
overall implementation of the judge’s preferred policies.
To investigate the plausibility of this scenario, I use data on solar eclipses and
earthquakes.31 Historical evidence suggests that both eclipses and earthquakes led
to signi…cant increases in the religiosity of the populace. Eclipses and earthquakes
30
One might also worry that the sultan became more religious during Nile failures. This expla-
nation, however, is not consistent with the results presented in columns (1)-(3) of table 3 since
the sultan was always Muslim but the head judge’s dismissal was related to the size of his popular
following.
31
Data on solar eclipses are drawn from Espenak and Meeus (2006) and include all solar eclipses
(total, annular, partial and hybrid) that could be seen from Cairo. Earthquake data are drawn
from Ambraseys et al. (1994) and include all earthquakes detailed in this source (i.e. those that
a¤ected Egypt, Arabia and the Red Sea). I experimented using subsets of these data (i.e. only
total eclipses or earthquakes with epicenters in Egypt) and results remained qualitatively similar
to those presented.
25
were portents of the apocalypse (Cook 2002, p. 14; Ambraseys et al. 1994, p. 7).
Earthquakes often led to signi…cant destruction and casualties and were followed by
fasts and calls to desist from the sinful behavior that was thought to have triggered
the earthquake (Suyuti 1973 [1505], p. 40; Little 1999, p. 138). Similarly, eclipses
were viewed as prominent signs of the last day and were accompanied by prayers and
calls for immoral behavior to cease (Akasoy 2007).32
Although both eclipses and earthquakes appear to have increased the religiosity of
the populace, solar eclipses did not a¤ect the material well-being of the population
and there is no evidence of eclipse-induced riots or divisions within the military.
While earthquakes often led to both destruction of the capital stock and causalities,
earthquakes do not seem to have signi…cantly a¤ected food prices or to have led to
popular riots or divisions within the military. Consequently, these events provide an
opportunity to empirically investigate the extent to which increases in religiosity and
demand for religiously-provided social services led to a decrease in the probability of
judge change and an increase in the construction of religious structures when holding
the level of popular and military unrest constant.
Results using eclipses and earthquakes are presented in table 4. Since the treatment e¤ects of solar eclipses and earthquakes did not follow the Nile ‡ood year, I use
the lunar month as the unit of observation. In addition to the variable F ailurems
(here indexed by lunar month m and solar year s) which varies by Nile year, I created
variables Eclipsems and Earthquakems which are equal to one if the lunar month is
in the year following an eclipse or earthquake respectively.33 I then run a regression
similar to equation (1) adding Eclipsems and Earthquakems . All results in table four
present standard errors clustered by lunar year since the standard errors in columns
(3) and (4) are sensitive to the way in which the standard errors are calculated and
those calculated in this manner yield the largest standard errors. Results in the
…rst column omit year and month dummies while those in column (2) include these
32
For further evidence that these events increased religiosity see Poirier and Taher (1980) and
Tucker (1981, 1999).
33
For judge changes I also ran regressions at the month level with dummies for whether an
earthquake or eclipse occurred in that month. These regressions yielded qualitatively similar results.
26
dummy variables. Results in both columns show that the data rejects the hypothesis
that the e¤ects of eclipses, earthquakes and famines on the probability of a judge
change were identical.
In columns (3) and (4) of table 4, I use the di¤erence between religious and
secular buildings constructed in the hijri year corresponding to month m as the dependent variable. While the standard errors on these coe¢ cients are large (which is
not surprising given that the dependent variable varies by lunar year which introduces measurement error into the dependent variable), results in column (4) reject
the hypothesis that Nile failures, eclipses and earthquakes had the same e¤ect on
resources allocated to religious structures.
When taken in unison, these results cast doubt on the importance of increases
in religious belief in driving the Nile-induced decrease in the probability of changes
in the head judgeship. In addition, since earthquakes often led to casualties and
signi…cant destruction of the capital stock (Little 1999) but do not seem to have affected judge dismissal or to have positively a¤ected the (relative) amount of resources
allocated to religious structures, the results are not consistent with explanations attributing the results to increased demand for religiously-provided social services. In
other words, these results are not consistent with Nile-induced increases in religiosity or demand for religious services driving the negative relationship between Nile
failures and judge changes.
The assumption that E[Dt (P1t ; 1)
Dt (P0t ; 1)] < 0 ! E[P1t
P0t ] > 0 is con-
sistent with the historical record and a neoclassical framework in which the judge
compares the sultan’s o¤er with his outside o¤er before deciding whether to resign
or oppose the sultan and be dismissed. One might worry, however, that deviant Nile
‡oods actually made the head judge more compliant leading to both E[P1t
and E[Dt (P1t ; 1)
P0t ]
0
Dt (P0t ; 1)] < 0. In addition to not being consistent with the avail-
able historical evidence, this scenario is not consistent with the empirical evidence in
column (3) of table 3 since one would expect all judges to exhibit similar behavior. In
addition, the observed increase in the probability of change of the head tax collector
during Nile failures also weighs against this possibility since he presumably would
also have had an incentive to become more compliant (i.e. accept a lower wage).
27
In sum, additional empirical evidence presented in this section supports the conceptual framework developed in section 2. Alternative narratives and explanations
for the results that do not invoke an increase in the political power of the head judge
that was rooted in his popular in‡uence are not consistent with the available body
of empirical and historical evidence.
6
Conclusion
A distinguished line of scholars have stressed the political importance of religious
organizations. Results in this paper are consistent with such scholarship. Deviant
Nile ‡oods decreased the probability of change of Egypt’s highest ranking religious
authority by roughly one-half. Nile failures also led to an increase in the construction
of new religious structures relative to secular structures. Historical evidence suggests
that these results are indicative of an increase in the head judge’s political power
that was rooted in his control over popular support.
I used three empirical strategies to better understand the extent to which the
statistical relationship is a product of this historical narrative. First, I used variation
in the number of followers of a head judge to test whether judge changes were less
likely for judges with larger followings. Results using this variation show that the
negative relationship between deviant Nile ‡oods and judge dismissal is stronger for
judges who had larger followings.
Second, I investigated the e¤ects of Nile shocks on the probability of change
of Egypt’s highest-ranking tax collector. Unlike the head judge, Nile shocks made
changes in this post more likely. This result casts doubts on explanations attributing
the Nile-induced decrease in the head judge’s probability of change to a decrease in
the probability of change of all o¢ cials in the Egyptian bureaucracy. In addition,
this result provides suggestive evidence for a Nile-induced increase in the threat of
popular revolt since it is consistent with historical evidence that the sultan attempted
to appease Nile-induced rioters by dismissing such o¢ cials.
Third, I used data on solar eclipses and earthquakes to investigate whether the
28
results are due to increased demand for religious services or religiously-provided
social services rather than because the head judge could alleviate the threat of revolt
through their popular in‡uence. The results show that eclipses and earthquakes did
not lead to a decrease in the probability of head judge change or to an increase
in the construction of new religious structures relative to secular structures. This
casts doubt on the importance of Nile-induced increases in religiosity or demand for
religiously-provided social services in generating the results.
While data limitations do not allow a detailed empirical investigation of the exact
mechanism leading from deviant Nile ‡oods to the increased political power of the
head judge, results suggest this increase was rooted in the head judge’s control over
popular support. This evidence is consistent with recent scholarship suggesting that
the political power of the populace (and their leaders) can increase during periods
of economic downturn, providing a democratic “window of opportunity”(Acemoglu
and Robinson (2000, 2001, 2006), Brückner and Ciccone (2011)).
Egypt did not democratize during periods of economic downturn. This paper
has provided evidence, however, that the political power of the head judge increased
during such periods. The extent to which religious leaders have worked to impede
democratic change historically remains a topic for future research, but results in this
paper suggest that better understanding the political role of these leaders may shed
light on institutional outcomes in many areas of the world.
References
‘Abd al-Rāziq, A. 1980. “Le vizirat et les vizirs d’Égypte au temps des Mamlūks.”
Annales Islamologiques, 16: 183-239.
Acemoglu, D. and J. A. Robinson. 2000. “Why Did the West Extend the
Franchise? Democracy, Inequality, and Growth in Historical Perspective.”Quarterly
Journal of Economics, 115(4) 1167-1199.
Acemoglu, D. and J. A. Robinson. 2001. “A Theory of Political Transitions.”
American Economic Review, 91(4) 938-963.
29
Acemoglu, D. and J. A. Robinson. 2006. Economic Origins of Dictatorship
and Democracy. New York: Cambridge University Press.
Acemoglu, D., S. Johnson and J. A. Robinson. 2005. “Institutions as a
Fundamental Cause of Long-run Growth.” in Philippe Aghion and Steven Durlauf
eds. Handbook of Economic Growth, Elsevier, 385-473.
Acemoglu, D., D. Ticchi and A. Vindigni.
2010. “A Theory of Military
Dictatorships.”American Economic Journal: Macroeconomics, 2(1), 1-42.
Akasoy, A. 2007. “Islamic Attitudes to Disasters in the Middle Ages: A Comparison
of Earthquakes and Plagues.”The Medieval History Journal, 10 : 387-410.
Ambraseys, N., C. Melville and R. Adams. 1994. The Seismicity of Egypt,
Arabia and the Red Sea. Cambridge: Cambridge University Press.
Amin, M. 1980. Al-Awqāf wa al-H
. ayāt al-Ijtimā’iya f¯¬Mis.r. Cairo: Dār al-Nahda
al-‘Arabiyya.
Angrist, J. and J.-S. Pischke. 2009. Mostly Harmless Econometrics. Princeton:
Princeton University Press.
Barro, R. 1999. “Determinants of Democracy.” Journal of Political Economcy,
107 (S6): 158-183.
Barro, R. and R. McCleary. 2003. “Religion and Economic Growth across
Countries.”American Sociological Review, 68 (5): 760-781.
Barro, R. and R. McCleary. 2005. “Which Countries Have State Religions?.”
Quarterly Journal of Economics, 120 (4): 1331-1370.
Barro, R. and R. McCleary.
2006. “Religion and Political Economy in an
International Panel.”Journal for the Scienti…c Study of Religion, 45 (2): 149-175.
Borsch, S. 2000. “Nile Floods and the Irrigation System in Fifteenth-Century
Egypt.”Mamlūk Studies Review, 4: 131-145.
Borsch, S. 2005. The Black Death in Egypt and England: A Comparative Study.
Austin: University of Texas Press.
Brückner, M. and A. Ciccone. 2011. “Rain and the Democratic Window of
Opportunity.”Econometrica, 79(3): 923-947.
Bulliet, R. 1979. Conversion to Islam in the Medieval Period: an Essay in Quantitative History. Cambridge: Harvard University Press.
30
Burke, E. 1987. “Understanding Arab Protest Movements.” Arab Studies Quarterly, 8(4): 333-345.
Chen, D. 2010. “Club Goods and Group Identity: Evidence from Islamic Resurgence during the Indonesian Financial Crisis.”Journal of Political Economy, 118(2):
300-354.
Cook, D. 2002. Studies in Muslim Apocalyptic. Princeton: The Darwin Press.
Cooper, R. 1976. “The Assessment and Collection of Kharāj Tax in Medieval
Egypt.”Journal of the American Oriental Society, 96(3): 365-382.
Coşgel, M., T. Miceli and R. Ahmed. 2009. “Law, State Power, and Taxation
in Islamic History.”Journal of Economic Behavior & Organization, 71 : 704-717.
Crecelius, D. 1986. “Incidences of Waqf Cases in Three Cairo Courts: 1640-1802”
Journal of the Economic and Social History of the Orient, 29(2): 176-189.
Creswell, K. 1919. “A Brief Chronology of the Muhammadan Monuments of Egypt
to A.D. 1517”Bulletin de l’Institut Français d’Archéologie Orientale, 16 : 39-164.
Dixit, A. and J. Londregan. 1995. “Redistributive Politics and Economic E¢ ciency.”American Political Science Review, 89(4): 856-866.
Ekelund, R.,R. Hébert, R. Tollison, G. Anderson and A. Davidson 1996.
Sacred Trust: The Medieval Church as an Economic Firm. New York: Oxford
University Press.
Escovitz, J. 1984. The O¢ ce of Qād.¯¬ al-Qud.āt in Cairo under the Bah.r¯¬ Mamlūks.
Berlin: Klaus Scharz Verlag.
Espenak, F. and J. Meeus. 2006. Five Millennium Canon of Solar Eclipses: 2000
BCE to 3000 CE. Washington, DC: National Aeronautics and Space Administration.
Fish, M. S. 2002. “Islam and Authoritarianism.”World Politics, 55 : 4-37.
Frantz-Murphy, G. 1986. The Agrarian Administration of Egypt from the Arabs
to the Ottomans. Cairo: Institut Français D’Archéologie Orientale.
Freeman-Grenville, G.S.P. 1995. The Islamic and Christian Calendars AD 6222222 (AH 1-1650). Reading: Garnet Publishing.
Gibb, H.A.R. 1955. “The Evolution of Government in Early Islam.”Studia Islamica, 4 : 5-17.
31
Gill, A. 1998. Rendering unto Caesar: the Catholic Church and the State in Latin
America. Chicago: University of Chicago Press.
Gill, A. 2001. “Religion and Comparative Politics.” Annual Review of Political
Science, 4 : 117-138.
Greif, A. 1994. “Cultural Beliefs and the Organization of Society: a Historical
and Theoretical Re‡ection on Collectivist and Individualist Societies.” Journal of
Political Economy, 102 (5): 912-950.
Greif, A. 2006. Institutions and the Path to the Modern Economy. New York:
Cambridge University Press.
Grossman, G. and Elhanan Helpman. 1994. “Protection for Sale.” American
Economic Review, 84(4): 833-850.
Grossman, H and J. Mendoza. 2003. “Scarcity and Appropriative Competition.”European Journal of Political Economy, 19 : 747-758.
Guiso, L., P. Sapienza and L. Zingales. 2003. “People’s Opium? Religion and
Economic Attitudes.”Journal of Monetary Economics, 50 : 225-282.
Guiso, L., P. Sapienza and L. Zingales. 2006. “Does Culture A¤ect Economic
Outcomes?.”Journal of Economic Perspectives, 20(2): 23-48.
Ibn H
. ajar, S. 1998 [1449]. Raf’u al-Is.r ’an Qudāt Mis.r. Cairo: Maktaba al-Khānj¯¬.
Hassan, F. 2007. “Extreme Nile ‡oods and famines in Medieval Egypt (AD 9301500) and their climatic implications.”Quaternary International 173-174 : 101-112.
Ibn Al-H
. ijāz¯¬, S. 1470. Nail al-Rā’id min al-N¯¬l al-Zā’id. Bibliothèque Nationale
de France, Manuscripts Orientaux: manuscript 2261.
Iannaccone, L. 1998. “Introduction to the Economics of Religion.” Journal of
Economic Literature, 36 : 1465-1496.
Iyigun, M. 2008. “Luther and Suleyman.”Quarterly Journal of Economics, 123(4):
1465-1494.
Irwin, R. (1986). The Middle East in the Middle Ages: the Early Mamluk Sultanate.
Carbondale: Southern Illinois University Press.
Israel, J. 2001. Radical Enlightenment: Philosophy and the Making of Modernity.
New York: Oxford University Press.
32
Jackson, S. 1996. Islamic Law and the State: the Constitutional Jurisprudence of
Shihāb al-D¯¬n al-Qarāf¯¬. New York: Brill.
Krusell, P. and J.-V. Ríos-Rull. 1996. “Vested Interests in a Positive Theory of
Stagnation and Growth.”Review of Economic Studies, 63 : 301-329.
Lambton, A. 1997. “Awqāf in Persia: 6th-8th/12th-14th Centuries.” Islamic Law
and Society, 4(3): 298-318.
Lancaster, T. 1990. The Econometric Analysis of Transition Data. New York:
Cambridge University Press.
Lapidus, I. 1975. “The Separation of State and Religion in the Development of
Early Islamic Society.”International Journal of Middle Eastern Studies, 6 : 363-385.
Lapidus, I. 1984. Muslim Cities in the Later Middle Ages. New York: Cambridge
University Press.
Lev, Y. 1991. State and Society in Fatimid Egypt. Leiden: E.J. Brill.
Lewis, B. 1953. “Some Observations on the Signi…cance of Heresy in the History
Of Islam.”Studia Islamica, 1 : 43-63.
Lipset, S. 1959. “Some Social Requisites of Democracy: Economic Development
and Political Legitmacy.”The American Political Science Review, 53(1): 69-105.
Lipset, S. 1994. “The Social Requisites of Democracy Revisited.” American Sociological Review, 59(1): 1-22.
Little, D. 1999. “Data on Earthquakes Recorded by Mamluk Historians: an Histographical Essay.” in Elizabeth Zachariadou eds. Natural Disasters in the Ottoman
Empire, Rethymnon: Crete University Press, 137-144.
Martel-Thoumian, B. 1991. Les Civils et l’Administration dans l’État Militaire
Mamlūk. Damascus: Institut Français de Damas.
Maqr¯¬z¯¬, T. 1994 [1405]. Ighāthat al-Ummah bi-Kashf al-Ghummah tr. Adel Allouche. Salt Lake City: University of Utah Press.
Maqr¯¬z¯¬, T. 1997 [1441]. Al-Sulūk li-Ma’arifat Duwal al-Mulūk. Beirut: Dār alKutub al-‘Ilmiyya.
Marx, K. 1982 [1844]. “Contribution to the Critique of Hegel’s Philosophy of Right.
Introduction.”In On Religion. Chico: Scholars Press.
33
McCann, S. 1999. “Threatening Times and Fluctuations in American Church
Memberships.”Personality and Social Pyschology Bulletin, 25(3): 325-336.
Montesquieu, C. 1989 [1748]. The Spirit of the Laws. New York: Cambridge
University Press.
Murphy, K. and A. Shleifer 2004. “Persuasion in Politics.”American Economic
Review, 94(2): 435-439.
Nielsen, J. 1984. “Sultan al-Z.āhir Baybars and the Appointment of Four Chief
Qād.is 663/1265.”Studia Islamica 60 : 167-176.
Nielsen, J. 1985. Secular Justice in an Islamic State: Maz.ālim under the Bah.r¯¬
Mamlūks. Leiden: Nederlands Instituut voor het Nabije Oosten.
North, D., J. Wallis and B. Weingast. 2009. Violence and Social Orders. New
York: Cambridge University Press.
Petry, C. 1981. The Civilian Elite of Cairo in the Later Middle Ages. Princeton:
Princeton University Press.
Petry, C. 1994. Protectors or Praetorians?: the Last Mamluk Sultans and Egypt’s
Waning as a Great Power. Albany: SUNY Press.
Poirier, J. P. and M. A. Taher. 1980. “Historical Seismicity in the Near and
Middle East, North Africa, and Spain from Arabic Documents.” Bulletin of the
Seismological Society of America, 70(6): 2185-2201.
Popper, W. 1951. The Cairo Nilometer. Berkeley: University of California Press.
Raymond, A. 1973. Artisans et Commerçants au Caire au XVIII Siècle (I). Damascus: Institut Français de Damas.
Raymond, A. 1974. Artisans et Commerçants au Caire au XVIII Siècle (II).
Damascus: Institut Français de Damas.
Raymond, A. 2002. Cairo: City of History. New York: The American University
in Cairo Press.
Ruzmaikin, A., J. Feynman and Y. Yung. 2006. “Does the Nile Re‡ect Solar
Variability?.”Proceedings IAU Symposium 233 : 1-8.
Sabra, A. 2000. Poverty and Charity in Medieval Islam. Cambridge: Cambridge
University Press.
Sām¯¬, A. 1916. Taqw¯¬m al-N¯¬l. Cairo: Mat.b’a al-‘Amiriya.
34
Sanders, P. 1994. Ritual, Politics, and the City in Fatimid Cairo. Albany: State
University of New York Press.
Al-Sayyid Marsot, A. 1972. “The Ulama of Cairo in the Eighteenth and Nineteenth Centuries” in Scholars, Saints, and Su…s: Muslim Religious Instiutions since
1500, Nikki R. Keddie (ed.) pp. 149-165.
Al-Sayyid Marsot, A. 1973. “The Political and Economic Functions of the ‘Ulamā’
in the 18th Century.” Journal of the Economic and Social History of the Orient,
16(2/3): 130-154.
Shoshan, B. 1980. “Grain Riots and the “Moral Economy”: Cairo, 1350-1517.”
Journal of Interdisciplinary History, 10(3): 459-478.
Shoshan, B. 1993. Popular Culture in Medieval Cairo. New York: Cambridge
University Press.
Smith, A. 2009 [1776]. The Wealth of Nations: Books IV-V. New York: Classic
House Books.
Stark, R. and R. Finke 2000. Acts of Faith: Explaining the Human Side of
Religion. Berkeley: University of California Press.
al-Suyūt¯¬, J. 1973 [1505]. Kashf al-S.als.ala ‘an Was.f al-Zalzala. tr. Said Nejjar.
Rabat: Cahiers du Centre Universitaire de la Recherche Scienti…que.
Tabellini, G. 2008. “Institutions and Culture.”Journal of the European Economic
Association, 6(2-3): 255-294.
Tabellini, G. 2010. “Culture and Institutions: Economic Development in the Regions of Europe.”Journal of the European Economic Association, 8(4): 677-716.
Taghr¯¬Bird¯¬, A. 1976 [1468]. Al-Nujūm al-Zāhira f¯¬ Mulūk Mis.r wa al-Qāhira vol.
1-7, tr. William Popper. New York: AMS Press.
Taghr¯¬Bird¯¬, A. 1990 [1468]. Hawādith al-Duhūr f¯¬ Mada al-Ayyām wa al-Shuhūr.
Cairo: Wizāra al-Awqāf.
Tucker, W. 1981. “Natural Disasters and the Peasantry in Mamlūk Egypt.”Journal
of the Economic and Social History of the Orient, 24(2): 215-224.
Tucker, W. 1999. “Environmental Hazards, Natural Disasters, Economic Loss, and
Mortality in Mamluk Syria.”Mamlūk Studies Review, 3: 109-128.
De Toqueville, A. 1983 [1856]. The Old Regime and the French Revolution. New
35
York: Anchor Books.
De Toqueville, A. 2003 [1840]. Democracy in America. New York: Penguin Books.
Toussoun, O. 1925. Memoire sur l’histoire du Nil, vol. 9. Memoires a l’Institut
d’Egypte.
Tsugitaka, S. 1997. State and Rural Society in Medieval Islam. Leiden: E.J. Brill.
Tyan, E. 1960. Histoire de l’Organisation Judiciaire en Pays d’Islam. Leiden: E.J.
Brill.
7
Appendix
7.1
7.1.1
Data
Lunar and Solar Years
The data sources use the hijri calendar. Hijri years are lunar years consisting of 354
days (355 days in a lunar leap year) and have twelve months which alternate between
29 and 30 days. The …rst day of the …rst hijri year corresponds to the solar date July
16th 622 CE (solar year is henceforth used to denote CE years), the day the Prophet
Muhammad made his Hijra or migration from Mecca to Medina. Since the lunar
year or AH year (both notations are used to denote hijri years) is approximately
11 days shorter than the solar year, the …rst day of the lunar year slowly “cycles
backwards”through the solar year.
The relationship between lunar and solar years can been seen in …gure three.
The brackets denote the solar years 1000-1015 CE. The vertical lines denote the
…rst day of each lunar year. The solar date corresponding to the start of each lunar
year is given above the highest horizontal line. Lunar years run from vertical line to
vertical line and are numbered in the row marked lunar year. Inspection of …gure
one shows that the vertical lines occur earlier in the solar year as time goes on. This
is a graphical representation of the backwards cycling of the lunar year through the
solar year. Consequently, the lunar months have no calendar regularity. This has
important implications for the correct assignment of the Nile ‡ood data as discussed
36
below.
7.1.2
Head Judge Data
Ibn Hajar (1998 [1449]) provides the month and year of head judge changes. Of the
245 judge changes reported by Ibn Hajar on the interval [20, 10th month of 842 AH]
the year of dismissal was available (or could be imputed) for 239 changes (98%) and
the year and month of dismissal was available for 209 changes (85%).
Missing dismissal dates were imputed as follows. Ibn Hajar (1998 [1449], pp. 421) provides a poem that lists Egypt’s judges in chronological order.34 When a judge’s
dismissal date was missing, I replaced this missing date with the appointment date
of the judge that chronologically followed him in the poem. When the appointment
month/year of the following judge was also missing, the dismissal month/year was
left blank. Judge changes missing the month but containing the year of change were
assigned the month 6.
When there was more than one head judge after 1265 CE I included the dismissal
date of the Sha…i judge in the main series (the Sha…i school was the most in‡uential
in Egypt during the Mamluk era).35 After 1265 CE I created 3 additional series for
the head judges of the three other law schools in an identical manner to those for
the Sha…i school.
Since the head judge data do not completely cover the hijri year 842, I discard
it from the sample. The data set consequently contains 239 judge dismissals on the
interval [20,841 AH].
Although the head judge change data is given by lunar months and years, Nile
‡oods followed the solar calendar (see section 2.1 for a discussion of the Nile’s seasonal
‡ood pattern). To correctly assign each Nile ‡ood to lunar months it treated I
developed a mapping from lunar months/years to solar months/years using the tables
34
35
A spreadsheet documenting the construction of this data set is available upon request.
There was also more than one head judge brie‡y under the Fatimids. During the Fatimid
period I included the Ismaili (Shia) head judge since the head judges were generally Ismaili under
the Fatimids.
37
provided in Freeman-Grenville (1995).36
To create the mapping between lunar and solar months I …rst calculated the
percentage of each lunar month occupied by a given solar month. I then assigned a
lunar month to a solar month if the solar month occupied 50% or more (15 days or
more) of the lunar month. When the lunar month was evenly divided between two
solar months I used the earlier of the two solar months. Sometimes 2 di¤erent lunar
months are assigned to one solar month (that is, one solar month occupied 50% or
more of two lunar months).
7.1.3
Nile Flood Data
In the empirical section, I use the maximum Nile ‡ood level each year to construct a
metric of Nile-induced economic shocks. Egyptian political leaders were interested in
gathering this maximum level and other ‡ood statistics to both estimate agricultural
tax receipts (Cooper 1976) and forecast future ‡ood levels (Petry 1994, p. 105). To
this end, they constructed Nilometers.
The Nilometer on the island of Rauda (Cairo) was among Egypt’s most elaborate.37 This Nilometer was sta¤ed by a guardian who provided daily measurements
to government o¢ cials. Although the guardian’s original records are lost, two historians in the 15th century compiled copies of the Nile’s historical minimum and
maximum levels.
The historians Ibn al-Hijazi and Taghri Birdi provide two separate sets of Nile
‡ood data. Hijazi’s (1470) statistics cover the interval [1,873] AH, whereas Taghri
Birdi’s statistics span the years [20,855] AH.38 Both data sets appear to be copies of
the original records kept by the guardian of the Nilometer on the island of Rauda.
The authors give the yearly maximum and minimum level of the Nile ‡ood by lunar
year.
36
I use the Julian Calendar as a proxy for the solar (tropical) calendar for simplicity, while
recognizing that this calendar slowly diverged from the solar calendar. By 1582 this calendar had
diverged from the tropical calendar by 10 days.
37
This section draws on Popper (1951).
38
Taghri Birdi’s data are drawn from Sami (1916, introduction, pp. 4-9).
38
Scholars agree that these data provide credible estimates of the true Nile ‡ood
levels and Hijazi’s data have been extensively used in the climatology literature. The
data, however, contain two sources of measurement error. First, both data sets have
transcription errors. I show below that this source of error attenuates the coe¢ cients
of interest under plausible assumptions. Second, both authors assigned the yearly
Nile ‡ood maxima and minima to lunar years. The assignment of the ‡ood levels by
lunar year introduces an additional source of measurement error if left uncorrected.
Figure three highlights the problems introduced by the assignment of Nile ‡oods
to lunar years. To better understand these problems, consider the ‡ood that occurred
in the year 1008 CE. Hijazi recorded this ‡ood level as having occurred in the year
398 AH (the solar year ‡ood recorded by Hijazi in each lunar year is given in the
…nal row of …gure one). The year 398 AH spanned the interval [17, September
1007; 4, September 1008]. Recall, however, that information regarding the maximum
Nile ‡ood level began to be revealed in July (see section 2.1). Consequently, if I
assigned the ‡ood recorded by Hijazi in 398 AH to that lunar year, only the lunar
dates corresponding to the solar interval [1, July 1008; 4, September 1008] would be
correctly assigned. I would spuriously assign to the remainder of dates in the lunar
year 398 the ‡ood that occurred in 1008 CE when in reality these dates were treated
by the Nile ‡ood that occurred in 1007 CE.
To further complicate matters, neither Hijazi nor Taghri Birdi appear to have
necessarily recorded Nile ‡oods in the lunar year in which the maximum ‡ood level
occurred. Hijazi seems to have assigned solar year ‡oods to lunar years by convention.
The mapping between the ‡ood levels recorded by lunar years and the CE year in
which the ‡ood actually occurred is given in Toussoun (1925).
Although there is enough information to construct such a mapping for Hijazi’s
data, it is not possible to determine which solar year ‡ood Taghri Birdi recorded in
each lunar year.39 For this reason I use Hijazi’s data throughout the paper (although
39
The mapping from lunar to solar years is made possible in Hijazi’s data by the fact that he
skips every 34 lunar year. Taghri Birdi’s data does not follow such a pattern, making the exact
mapping between lunar and solar years unclear. See Popper (1951, pp. 123-149) for a detailed
discussion.
39
results using Taghri Birdi’s data are usually qualitatively similar to those reported).
Equipped with Hijazi’s ‡ood data assigned to CE years, I merged it to the head
judge data set as follows. I de…ned a Nile year to run from July through June of
the solar year (since these were approximately the dates treated by each Nile ‡ood
level). I then assigned the Nile ‡ood from CE year t to the lunar months in the
interval [July, t; June, t+1].
For most of the empirical section, I use the Nile year as the unit of observation.
Collapsing the monthly data down to the Nile year level was straightforward. The
merged data sets span the interval [20, 841 AH]. Since the year 841 AH ended on
June 23, 1438 CE, the data set contains 797 Nile years [641,1437]. In other words,
the data cover the interval [July, 641; June, 1438 CE].
When regressions are run at the lunar level there are 9857 observations. This
number can be understood as follows. If I ran the regressions at the solar month
level I would have 797*12=9564 observations. There are 293 solar years containing
13 lunar months and the total number of monthly observations at the lunar month
level is 9564+293=9857.
7.2
Bias Introduced by Measurement Error
Without loss of generality consider the regression at the Nile year level without
covariates:
Judget =
0
+
1 F aminet
+ "t
where Judget is a dummy variable equal to one if the incumbent judge at time t
is dismissed on the interval (t,t+1) and F aminet is a dummy variable equal to one
if the correctly measured Nile ‡ood is in the top 5% or the lower 5% of the ‡ood
distribution. I assume that all necessary moments exist.
I observe DevN ilet = DevN ilet +
1(DevN ilet
t
and F aminet = 1(DevN ilet
c5 ) +
c95 ) where c5 is the lower 5% cuto¤ of the observed Nile ‡ood distri-
bution and c95 denotes the cuto¤ for the upper 5% of this distribution. Assuming
40
that E["t jF aminet ] = E["t ] (a su¢ cient condition for this is the joint independence
of DevN ile ; t and "t ) standard arguments imply that the coe¢ cient ^ 1 converges in
t
probability to
1 [Pr(F aminet
= 1jF aminet = 1)
Pr(F aminet = 1jF aminet = 0)].
Using Bayes’rule this can be written as 1 [ 10
Pr(F aminet = 1jF aminet = 1) 91 ].
9
Consequently, ^ will converge to an attenuated constant that has the same sign as
1
1
as long as Pr(F aminet = 1jF aminet = 1)
41
0:10.
600
800
Hijazi(Min)
MinTrend
Hijazi(Crest)
CrestTrend
1000
CE Year
1200
1400
16 Cubits
8
10
Meters Above Sea Level
12
14
16
18
Figure 1: Maximum and Minimum Nile Flood Levels
Sixteen Cubits (16.47 meters) marks “plenitude”
2
Lowess Fit
Judge Change
1
0
Nile Deviation
−1
−2
0
.2
Probability of Change
.4
.6
.8
1
Figure 2: Nile Variation and Probability of Change
Graph details (raw) non-parametric relationship between changes in the
judgeship and the Nile maximum’s deviation from a linear trend. Vertical
lines mark the top 5% and lower 5% of the maximum flood distribution.
Solar Date of First
Day of Lunar Year
Solar
Year
Lunar
Year
Solar Year
Flood Recorded
in Lunar Year
N
o
v
1
0
N
o
v
2
0
D
e
c
1
O
c
t
3
0
O
c
t
1
8
1000
1001
1002
1003
1004
1005
390
391
392
393
394 395
S
e
p
1
7
S
e
p
5
A
u
g
2
5
A
u
g
1
5
A
u
g
4
J
u
l
2
3
J
u
l
1
3
J
u
l
2
1007
1008
1009
1010
1011
1012
1013
1014
S
e
p
2
7
O
c
t
8
1006
396
397
398
399
400
401
402 403
404
J
u
n
2
1
1015
405
1000 1001 1002 1003 1004 1005 1006 1007 1008 1009 1010 1011 1012 1013 1014 1015
Figure 3: Lunar and Solar Years
Brackets denote solar years. Vertical lines mark the first day of the lunar
year denoted to the right. The solar year of the Nile flood Hijazi assigned
to each lunar year is provided in the final row.
Table 1: Flood Deviations and Judge Changes
=100 if judge at t was dismissed on (t,t+1)
Lunar Month
Nile Flood Year
(1)
(2)
(3)
(4)
(5)
(6)
Panel 1: Nile Max.
F ailuret+1
-3.33
-0.96
(5.55)
F ailuret
(0.87)
-8.93∗∗
-15.11∗∗∗
-14.03∗∗
-1.11∗∗∗
-1.71∗∗
-2.04∗∗
(4.38)
(5.75)
(5.83)
(0.43)
(0.70)
(0.92)
F ailuret−1
797
2.35
0.19
(5.07)
(0.73)
N
Panel 2: Nile Min.
F ailuret+1
797
795
9857
9857
F ailuret
7.16
8.04
8.39
0.53
0.63
-0.39
(5.37)
(5.38)
(5.47)
(0.56)
(0.76)
(0.89)
2.87
-1.09
(5.21)
F ailuret−1
N
Decade Dummies?
Solar Year Dummies?
Solar Month Dummies?
750
No
No
No
750
Yes
No
No
9833
(1.01)
-6.18
-0.94
(4.22)
(0.83)
737
Yes
No
No
9275
No
No
No
9275
No
Yes
Yes
Notes: estimates of equation (1) reported in panel 1. The dependent variable is 100* a
dummy variable equal to one if the incumbent judge at time t left office on the interval
(t,t+1). t denotes Nile flood years ([July,June] of the solar calendar) in columns (1)(3) and lunar months in columns (4)-(6). F ailuret is a dummy variable equal to one if
the Nile flood in Nile year t was in the upper 5% or lower 5% of the flood distribution.
Panel 2 constructs the variable F ailuret using the minimum Nile level. This variable
is equal to one if the minimum level in Nile year t was in the upper 5% or lower 5% of
the minimum flood distribution. Coefficients are reported with Newey-West standard
errors with 7 lags (Nile year) and 17 lags (lunar month) in parentheses in panel 1.
Robust standard errors are presented in panel 2 due to gaps in the minimum flood
data. ∗∗∗ ,∗∗ and ∗ indicate significance at the 1%, 5% and 10% levels.
9128
No
Yes
Yes
Table 2: Nile Floods and Monument Construction
Religious Secular
Diff.
(1)
(2)
(3)
Panel 1: Nile Max.
F ailuret
12.49
-14.02∗∗ 26.51∗∗
N
Panel 2: Nile Min.
F ailuret
N
Decade Dummies?
(9.65)
(6.34)
(13.19)
797
797
797
-6.46
-1.64
-4.82
(6.88)
(6.53)
(9.91)
750
Yes
750
Yes
750
Yes
Notes: in column (1) the dependent variable is 100* the standardized value of the
weighted average of religious constructions commenced during the lunar years that
overlapped with a given Nile year. In column (2) the dependent variable is identical to that in column (1) where secular constructions are substituted for religious
constructions. In column (3) the dependent variable is the difference between the dependent variables is column (1) and column (2). F ailuret is a dummy variable equal
to one if the Nile flood in Nile year t was in the upper 5% or lower 5% of the flood
distribution. Panel 2 constructs the variable F ailuret using the minimum Nile level.
This variable is equal to one if the minimum level was in the upper 5% or lower 5% of
the minimum flood distribution. Coefficients are reported with Newey-West standard
errors with 7 lags in parentheses in panel 1. Robust standard errors are presented in
panel 2 due to gaps in the minimum flood data. ∗∗∗ ,∗∗ and ∗ indicate significance at
the 1%, 5% and 10% levels.
(14.34)
(14.21)
Yes
All
797
OLS
Decade Dummies?
Sample
N
Estimation
Yes
Post 1265
172
SUR
[0.05]
2.63∗∗
(5.61)
-3.23
(10.57)
9.98
(2.95)
-15.69∗∗∗
(6.27)
Yes
Post 1250
187
OLS
(10.54)
Notes: the dependent variable in columns (1)-(3) the dependent variable is 100* a dummy variable equal to one if the incumbent
judge at time t left office on the interval (t,t+1). In column (4) this variable is 100* a dummy variable equal to one if the incumbent
tax collector at time t left office on the interval (t,t+1). P ercentM uslimt is the percent of the Egyptian population in year t that
was Muslim. Shiat is a dummy variable equal to one if the head judge was Shia. Results in column (4) are from a seemingly
unrelated regression estimation of equation (1) run separately for judges of the four Sunni law schools. The coefficient for the
Shafii head judge is reported in the row labeled F ailuret . The entries in the row F − Stat provide the test statistic and p-value
testing the hypothesis that both interaction terms are equal to zero in column (2) and testing the hypothesis that the four head
judge coefficients are equal in (3). Coefficients are reported with Newey-West standard errors with 7 lags in parentheses in columns
(1),(2) and (4) and SUR-robust standard errors in column (3). ∗∗∗ ,∗∗ and ∗ indicate significance at the 1%, 5% and 10% levels.
[0.01]
[p − value]
(15.17)
4.86∗∗∗
Yes
All
797
OLS
-36.67∗∗
-24.82∗
37.32∗∗
(10.68)
(10.69)
F − Stat
HanbaliF ailuret
Hanaf iF ailuret
M alikiF ailuret
F ailuret ∗ Shiat
F ailuret ∗ P ercentM uslimt
F ailuret
Table 3: Judge Changes, Popular Support and Tax Collectors
100*Judge 100*Judge 100*Judge 100*TaxCollector
(1)
(2)
(3)
(4)
∗∗∗
-1.79
-0.17
-21.53
30.52∗∗∗
7.69∗∗∗
[0.01]
No
No
9857
F − Stat
[p − value]
Solar Year Dummies?
Solar Month Dummies?
N
No
No
9857
[0.25]
1.38
(21.97)
Yes
Yes
9857
[0.02]
3.89∗∗
(14.12)
-24.04∗
(6.00)
-7.12
(5.29)
Notes: The dependent variable in columns (1) and (2) is 100* a dummy variable equal to one if the incumbent judge in lunar month
m left office on the interval (m,m+1). In columns (3) and (4) the dependent variable is 100* the difference between the standardized
religious and secular structures constructed in the lunar year corresponding to m. F ailurems is a dummy variable equal to one if
the Nile flood in Nile year corresponding to lunar month m was in the upper 5% or lower 5% of the flood distribution. Eclipsems
is a dummy variable equal to one if the lunar month m falls in a twelve month interval after a solar eclipse. Earthquakems is a
dummy variable equal to one if the lunar month m falls in a twelve month interval after an earthquake. The entries in the row
F −Stat provide the test statistic and p-value testing the hypothesis that the coefficients on the variables F ailurems , Eclipsems and
Earthquakems are equal. Coefficients are reported with robust standard errors clustered by lunar year in parentheses. ∗∗∗ ,∗∗ and
∗
indicate significance at the 1%, 5% and 10% levels.
Yes
Yes
9857
[0.00]
(1.07)
4.63∗∗∗
17.37
(9.19)
(0.48)
(0.39)
0.31
-2.38
0.32
(10.82)
(0.67)
1.36∗∗∗
0.51
(0.41)
(0.69
Earthquakems
Eclipsems
Table 4: Alternative Explanations: Eclipses and Earthquakes
100*Judge 100*Judge
Diff.
Diff.
(1)
(2)
(3)
(4)
F ailurems
-1.12∗∗∗
-1.71∗∗
22.22∗∗ 10.11∗