Author's Note: We are grateful to Dr. Charlotte Weber for her assistance in the process of writing this article. We are also indebted to the four anonymous IJMES referees whose comments and criticisms helped us significantly improve our analysis. Special thanks to Dr. Judith Tucker and Sylvia Whitman for their patience and professionalism in the review process.

¹ Boğaç A. Ergene, Local Court, Provincial Society and Justice in the Ottoman Empire: Legal Practice and Dispute Resolution in Çankırı and Kastamonu (1652–1744) (Leiden: E. J. Brill, 2003).

² According to Colin Heywood, whose impressions are based on the accounts of 19th-century European travelers, the Christian inhabitants of the town (predominantly Greek Orthodox and some Armenian families) constituted about 7.5 to 15 percent of the population in the 19th century. Heywood, “Kastamonu,” Encyclopedia of Islam, 2nd ed. (Leiden: E. J. Brill, 1960), 4:738. There is no reference to a Jewish community in Kastamonu court records or in the secondary literature.

³ Ibid.

⁴ See, among others, Ömer Lütfi Barkan, “Edirne Askeri Kassamı'na ait Tereke Defterleri (1545–1659),” Belgeler, III (1968): 1–479; André Raymond, Artisans et commerçants au Caire au XVIII^e siècle, 2 vols. (Damascus: Institut français de Damas, 1973–74); Suraiya Faroqhi, Men of Modest Substance: House Owners and House Property in Seventeenth-Century Ankara and Kayseri (Cambridge: Cambridge University Press, 1987); Haim Gerber, Economy and Society in an Ottoman City: Bursa, 1600–1700 (Jerusalem: Hebrew University of Jerusalem, 1988); Jean-Paul Pascual and Colette Establet, Familles et fortunes à Damas: 450 foyers damascains en 1700 (Damascus: Institut français de Damas, 1994); and Rossitsa, Gradeva , “Towards a Portrait of ‘the Rich’ in Ottoman Provincial Society: Sofia in the 1670s,” in Provincial Elite in the Ottoman Empire, ed. Antonis, Anastasopoulos (Rethymno, Crete: Crete University Press, 2005), 149–200Google Scholar.

⁵ Other designations that indicate affiliation with the religious establishment accompanied these titles. Efendi was a title used for the most respected and higher-ranking members of the religious establishment. The names of virtually all kadıs, müderrises, muftis and other important members of the religious establishment were accompanied by this title. Molla was a title officially given to high-ranking judges and medrese teachers. In Kastamonu, however, and especially when it was not coupled with efendi (as in Molla Ahmed Efendi), this title frequently referred to medrese students and dropouts as well as individuals who had low-ranking religious and scribal positions in local vakıfs. Halife is a tarikat (mystical order) designation, yet in Kastamonu this title was frequently used for the lowest-ranked members of the religious establishment, most notably prayer leaders (imams) in neighborhood mosques. Çelebi was a generic title used for individuals who possessed some religious education or were affiliated with influential ulema families; for the other ways in which this title was used in different contexts, see Güçlü Tülüveli, “Honorific Titles in Ottoman Parlance: A Reevaluation,” International Journal of Turkish Studies 11 (2005): 17–28. A şeyh was a head of a tarikat, and the title dede was specifically used for senior dervishes. See Gustav Bayerle, Pashas, Begs, and Effendis: A Historical Dictionary of Titles and Terms in the Ottoman Empire (Istanbul: Isis, 1997).

⁶ Other military designations, when used, always accompanied these three titles. According to Barkan, the titles ağa, beğ, and beşe do not reveal an individual's exact military status. That is, these terms do not indicate whether he belonged to the janissary corps, was a tımar holder, or was identified with any other military subgroup: see his “Edirne Askeri Kassamı,” 15–17. It is difficult to make an intragroup distinction based on military titles. It may be that an ağa had a higher ranking than a beşe, but it is problematic to make the same kind of distinction between a beşe and a beğ, given that these titles were sometimes used interchangeably in sicils: see Tülüveli, “Honorific Titles,” 21–23.

⁷ We have never encountered a military titleholder who is later identified by a religious title or vice versa. To the extent that we could determine, the use of titles for specific individuals remains constant. For example, an Ahmed Ağa bin Hasan is never called Ahmed bin Hasan in a different entry (one exception to this rule is when two religious titles, such as molla and efendi, are used simultaneously for the same individual; for such cases, see note 38).

⁸ There were many beşes, if not ağas or beğs, among the ranks of Kastamonu's artisans during our period—a situation not unique to Kastamonu. In our context, the title beşe probably indicated membership in local janissary divisions: see Yi, Guild Dynamics in Seventeenth-Century Istanbul: Fluidity and Leverage (Leiden/Boston: E. J. Brill, 2004), 132–43; and André Raymond, Cairo (Cambridge, Mass.: Harvard University Press, 2000), 219.

⁹ On social stratification and relative class positions for our period, see Abraham, Marcus , The Middle East on the Eve of Modernity: Aleppo in the Eighteenth Century (New York: Columbia University Press, 1988), 66–76Google Scholar; Linda, Schilcher , Families in Politics: Damascene Factions and Estates of the Eighteenth and Nineteenth Centuries (Wiesbaden, Germany: F. Steiner, 1985), 108–10Google Scholar; Karl, Barbir , “Wealth, Privilege, and Family Structure,” in The Syrian Land in the Eighteenth and Nineteenth Century: The Common and the Specific in the Historical Experience, ed. Thomas, Philipp (Stuttgart, Germany: Franz Steiner Verlag, 1992), 179Google Scholar; and Ruth, Roded , “Social Patterns among the Urban Elite in Syria in the Late Ottoman Period,” in Palestine in the Late Ottoman Period, ed. David, Kushner (Jerusalem: Yad Izhak Ben-Zvi, 1986), 147–52Google Scholar.

¹⁰ We have information about the occupational characteristics of 167 artisans and merchants in our sample of 778 (22 percent). A majority of these individuals (131) were men without titles; the rest (18) belonged to the military establishment. Of these, 12 artisans and 2 merchants were beşes, 2 artisans were beğs, and 2 merchants were ağas. No religious titleholder was found within this group.

¹¹ For this study, we consulted the microfilm copies of Kastamonu court records at the National Library of Turkey in Ankara. This collection begins in 1684 and includes the 18th and 19th centuries. Some of the earliest registers, however, are in poor condition and lack many pages. Hence, we chose to start our research at a later date, when the quality of documentation improved.

¹² We did not include in our sample inventories those who died while visiting Kastamonu because these documents do not contain any information about their assets outside the town. We also excluded a handful of seemingly incomplete and redundant terekes.

¹³ It is arguable that some who died poor or in bankruptcy may have been considered wealthy at specific points in their lifetimes by their contemporaries. Our analysis does not attempt to identify these people or to separate them from others who left behind modest estates.

¹⁴ Süleyman Özmucur and Şevket Pamuk, “Real Wages and the Standards of Living in the Ottoman Empire, 1469–1914,” The Journal of Economic History 62 (2002): 293–321.

¹⁵ The presence of extreme or unexpected values in any data set (and especially historical ones) may result from errors made during data recording and computation or directly inherited from sources. Trimming is a standard statistical procedure to eliminate the possibility of contamination caused by such errors.

¹⁶ Gradeva, “‘The Rich,’” 152–63. Also see Establet and Pascal, Familles et fortunes, 31–32; and Dror Ze'evi, “The Use of Ottoman Shariʿa Court Records as a Source for Middle Eastern Social History: A Reappraisal,” Islamic Law and Society 5 (1998): 43–45.

¹⁷ On the problem of the underrepresentation of estates owned by women, see Jean-Paul, Pascual and Colette, Establet, “Women in Damascene Families around 1700,” Journal of the Economic and Social History of the Orient 45 (2002): 302–303Google Scholar.

¹⁸ We excluded these inventories from our sample due to concerns about their representative value.

¹⁹ See Ergene, Boğaç A. , “Costs of Court Usage in the Seventeenth- and Eighteenth-Century Ottoman Anatolia: Court Fees as Recorded in Estate Inventories,” Journal of Economic and Social History of the Orient 45 (2002): 20–39CrossRefGoogle Scholar.

²⁰ Poorer estates, in particular, may be underrepresented in our sample. There is evidence that legal fees and taxes extracted from poorer estates at the time of their assessment and division in court were disproportionately higher than those taken from wealthier estates in late 17th and early 18th-century Anatolia: see Ergene, “Costs of Court Use,” and idem., Local Court, chap. 5. This situation may have encouraged poorer parties to avoid court involvement in assessment and division of inherited estates. Overrepresentation of the wealthy in estate inventories is not peculiar to Ottoman sources; it also appears in early-modern U.S. and British inventories. See Lindert, Peter H. , “Unequal English Wealth since 1670,” The Journal of Political Economy 94 (1986), 1132CrossRefGoogle Scholar; and Carole, Shammas , “Constructing a Wealth Distribution from Probate Records,” Journal of Interdisciplinary History 9 (1978), 297Google Scholar.

²¹ Establet and Pascual indicate that a woman and a child would have needed 17 to 27 guruş and a man about 50 guruş per annum in early 18th-century Damascus for their most basic needs: Familles et fortunes, 126–27.

²² The proportion of average income (not wealth) levels of the wealthiest 20 percent to the poorest 20 percent ranges from 3.4 (Japan) to 8.4 (United States) in the industrialized world. The same ratio ranges from 2.6 (Azerbaijan) to 8.3 (Georgia) in the former Eastern Bloc countries and from 4 (Rwanda) to 56.1 (Namibia) in the rest of the world. See U.N. Human Development Report 2006 (New York: Palgrave Macmillan, 2006), 335–38.

²³ The coefficient of variation and the Gini coefficient are two different measures of statistical dispersion. Although some studies use the coefficient of variation as an indicator of wealth/income disparities, in this article, in line with the tendency prevalent in the scholarly literature, we use the Gini coefficient as the primary measure of wealth inequality, but the coefficient of variation is presented simply as a unit-free measure of variation. For reasons why the Gini coefficient is a better measure of inequality, see Frank Cowell, Measuring Inequality (London: Prentice–Hall, 1995), 25.

²⁴ Theoretically, the Gini coefficient could be any value between 0 and 1. It would be 0 if wealth were distributed absolutely equally and 1 if a single individual were to possess all wealth in a given society. The Gini coefficient ranges from 0.24 (Japan) to 0.41 (United States) in the industrialized world, from 0.19 (Azerbaijan) to 0.41 (Turkmenistan) in the former Eastern Bloc countries, and from 0.29 (Rwanda) to 0.74 (Namibia) in the rest of the world: see U.N. Report, 335–38.

²⁵ Colette Establet, Jean-Paul Pascual, and André Raymond, “La mesure de l'inégalité dans la société ottomane: utilisation de l'indice de Gini pour Le Caire et Damas vers 1700,” Journal of the Economic and Social History of the Orient 37 (1994), 177, 180.

²⁶ The Gini coefficient for the untrimmed sample is 0.67053.

²⁷ Carole Shammas, “A New Look at Long-Term Trends in Wealth Inequality in the United States,” The American Historical Review 98 (1993), 420. Shammas's findings correct earlier calculations, which tended to underestimate the Gini by as much as five-tenths of a point: see Jeffrey G. Williamson and Peter H. Lindert, American Inequality: A Macroeconomic History (New York: Academic Press, 1980), 38–30. For regional variations, see G. B. Warden, “Inequality and Instability in Eighteenth-Century Boston: A Reappraisal,” Journal of Interdisciplinary History 6 (1976): 585–620; Gloria L. Main, “Inequality in Early America: The Evidence from Probate Records of Massachusetts and Maryland,” Journal of Interdisciplinary History 7 (1977): 559–81; Alice H. Jones, “Wealth Estimates for the New England Colonies about 1770,” The Journal of Economic History 32 (1972): 98–127; and Billy G. Smith, “Inequality in Late Colonial Philadelphia: A Note on Its Nature and Growth,” The William and Mary Quarterly 41 (1984): 629–45. These studies indicate that there may have been as much as four-tenths of a point of difference among the Gini coefficients of the Middle Colonies (which ranges between 0.43 and 0.51 in different studies), New England (ranges between 0.55 and 0.61), and the South (0.83).

²⁸ James, John A. , “Personal Wealth Distribution in Late Eighteenth-Century Britain,” The Economic History Review 41 (1988), 551CrossRefGoogle Scholar. The author does not discuss what, specifically, “insurable wealth” included. According to James's calculations, the Gini coefficients of different regions in Britain ranged from 0.5 (West Midlands) to 0.73 (Scotland); ibid., 556.

²⁹ Shammas, “A New Look,” 422.

³⁰ Lindert, “Unequal English Wealth,” 1145.

³¹ See Smith, Daniel S. , “Underregistration and Bias in Probate Records: An Analysis of Data from Eighteenth-Century Hingham, Massachussetts,” The William and Mary Quarterly 32 (1975): 100–10CrossRefGoogle Scholar; Shammas, “Constructing a Wealth Distribution,” 297–307; and Lindert, Peter H. , “An Algorithm for Probate Sampling,” Journal of Interdisciplinary History 11 (1981): 649–68CrossRefGoogle Scholar.

³² To assess the strength of our comparative claims based on differences among mean estate values owned by specific groups, we performed t tests, which measure whether the differences between two averages are statistically meaningful. The t test is performed by dividing the difference between two averages by the standard error of the difference. Unless stated otherwise, all comparisons in our discussion based on the differences between specific averages are significant, at least at the 10-percent level.

³³ Wealth concentration among socially and politically privileged groups is not unique to the Ottoman context. Evidence for a strong relationship between wealth and social prominence in the 18th-century United States and Britain appears in James, “Personal Wealth Distribution”; Shammas, “Constructing a Wealth Distribution”; and Jones, “Wealth Estimates.”

³⁴ Gradeva claims that in late 17th-century Sofia members of the religious establishment were less prosperous than members of the military: see Gradeva, “‘The Rich,’” 173–74.

³⁵ According to Establet and Pascual's calculations, the mean value of estates owned by women in early 18th-century Damascus was about one-fifth that of estates owned by men: Familles et fortunes, 119. Legal restrictions on women's inheritance as well as the practice of transferring property ownership to male relatives before death are two factors that likely kept the value of women's estates at modest levels. On women transferring property to their kin, see Annalies Moors, Women, Property and Islam: Palestinian Experiences, 1920–1990 (Cambridge: Cambridge University Press 1995), 48–76. Lindert's calculations indicate that the average net worth of women in mid 18th-century England was about half that of men's, a finding confirmed by James for late 18th-century Britain. Lindert, “Unequal English Wealth,” 1135; and James, “Personal Wealth Distribution,” 551. Jones claims that in Maryland and Massachusetts during the late 18th century “women . . . fell much more preponderantly than men in the lower wealth classes”: see “Wealth Estimates,” 126.

³⁶ The Gini coefficient is more sensitive to variations in the middle of the wealth distribution than the coefficient of variation, which attributes greater significance to variations in the lower and higher tails of the same distribution. As indicated above, we regard the Gini as the primary measure of inequality.

³⁷ According to Gradeva, ağas were among the wealthiest people in late 17th-century Sofia. Beğs, she claims, were significantly less wealthy: Gradeva, “‘The Rich’,” 170–73.

³⁸ We have observed in a small number of cases that the title efendi could be coupled with molla and şeyh. When molla and efendi were used together in a legator's name (as in the case of Molla Mehmed Efendi bin Abdülcelil), we included his estate into the efendi sample. When şeyh and efendi were used together (as in the case of Şeyh Hüseyin Efendi bin Şeyh Hasan), the estate in question was included in the şeyh sample. We did this because occasionally those identified in one entry simultaneously as molla and efendi are also identified as efendi in different entries (but never as molla and never without a title). Similarly, some who are identified as şeyh and efendi in some entries are merely identified as şeyh in others.

³⁹ See Bayerle, Pashas, 78.

⁴⁰ However, the t test for this comparison could not confirm that the disparity was statistically significant (t = 1.02).

⁴¹ Those artisans whose specialization is explicitly identified in our sources include tanners, dyers, tailors, potters, bakers, shoemakers, jewelers, tinsmiths, coppersmiths, blacksmiths, candle makers, builders, and saddle makers. The occupations of some individuals not specifically identified in the terekes as such were revealed by the contents of their estates; we include seventeen such individuals in our sample of artisans. Merchants are identified simply as “tüccardan” in terekes.

⁴² According to Gerber, the mean value of estates owned by merchants in late 17th-century Bursa was about four times that of estates owned by artisans: see Gerber, “Social and Economic Position of Women,” 239–40. Establet and Pascual's calculations confirm the validity of this finding for 18th-century Damascus: Familles et fortunes, 128. In both contexts, the mean value of estates owned by artisans was significantly less (35 to 40 percent) than average-sized estates. In both Britain and the United States, merchants appear to have been among the wealthiest occupational groups during the 18th century. Artisans and craftsmen, in contrast, with a significant degree of wealth variations among specific trades, were largely clustered in middle to lower-middle wealth brackets. See especially James, “Personal Wealth Distribution,” 551–52; Lindert, “An Algorithm,” 663; Shammas, “Constructing a Wealth Distribution,” 301; and Jones, “Wealth Estimates,” 126.

⁴³ See Hülya Canbakal, “On the ‘Nobility’ of Provincial Notables,” in Provincial Elite, 39–50.

⁴⁴ Indeed, according to Establet and Pascual, descendants of Muhammad were about three times as wealthy as the rest of the population in Damascus at the turn of the 17th century: see Familles et fortunes, 129.

⁴⁵ Suraiya Faroqhi, “Anatolian Townsmen as Pilgrims to Mecca: Some Evidence from the XVIth – XVIIth Centuries,” in Soliman le Magnifique et son temps, Recontres de l'École du Louvre, ed. Gilles Veinstein (Paris: La Documentation Française, 1992), 309–25. Yet, according to Pascual and Establet, pilgrims in Damascus were not necessarily wealthier than their nonpilgrim counterparts: Familles et fortunes, 170–76.

⁴⁶ It must be emphasized that we do not assume, for the most part, a causal relationship between our “dependent” and “independent” variables in our regressions. Given the lack of exogenous instrumental variables in our sources, it is impossible to determine, for example, if one was wealthy because he was a pilgrim or if he was a pilgrim because he was wealthy. Instead, the regression analyses are conducted solely to estimate the magnitude and statistical significance of associations between estate values and specific variables. The only exception to this rule appears when we comment on the relationship between estate values and what we call “generational variables.”

⁴⁷ We took men with no military titles as the reference base for military titleholders and men with no religious titles as the reference base for religious titleholders. This was done in order to avoid the problem of perfect collinearity.

⁴⁸ We also prepared a regression model in which we kept military and religious titles as two undifferentiated variables. In this analysis, comparing to those estates owned by men without titles, we estimate that estates belonging to military and religious titleholders were 41 percent and 34 percent more valuable, respectively. These estimations are statistically significant at the 1-percent significance level.

⁴⁹ Our sources rarely provide information on father's occupations; thus, these variables have not been included in the regressions analysis.

⁵⁰ Absent variables in Table 9 are those that are either unavailable or statistically insignificant in all four regressions. For example, artisanal status was not included in the regression model for women and was statistically insignificant in the other three regressions. Coefficients for individual military and religious titles are also excluded because comparable and consistent information exists in Table 8. Note that our regression models for military and religious titleholders confirmed the statistical significance of the differences between coefficients for ağa and two other military titles (beğ, beşe), on the one hand, and between coefficients for efendi and four other religious ones (çelebi, halife, şeyh, dede), on the other hand. The difference between coefficients for efendi and molla, however, was statistically insignificant.

⁵¹ As suggested by one of our referees, we also estimated a regression model in which pilgrimage and father's pilgrimage are interacted with dummy variables for titleholders, in order to find out whether the magnitude and significance of their relationship with wealth varied across these subgroups. Our estimations did not confirm the existence of such variations.

⁵² That military and religious status were frequently inherited from fathers must have some bearing on coefficients for the father's title in regressions for military and religious titleholders. We assume that a degree of the relationship between wealth and the father's military/religious status is captured by coefficients for the legator's military/religious title. Our estimating a statistically significant coefficient for fathers with religious status in the regression for religious titleholders, despite this possible interaction, is noteworthy. This is also why it may not be a coincidence that the coefficient for fathers with military titles is quite sizable (1.57) in the regression for men without titles, given that no title variables were included in this model. It is a fact that the coefficient for fathers with religious status is insignificant in the same regression. No meaning should be attributed to this finding, however, as there are only two titleless men in our sample who had fathers with religious titles.

⁵³ Using a similar procedure to that outlined in note 51, we also explored whether the magnitude and statistical significance of the relationship between father's title and wealth varied across title-based subgroups. Our estimations did not provide any evidence to that effect.