¹ Fish, , “Islam and Authoritarianism,” World Politics 55 (October 2002CrossRefGoogle Scholar). Specific page references to Fish's article are embedded in the text.

² In the political science literature Adam Przeworski, Michael Alvarez, Jose Cheibub, and Fernando Limongi provide perhaps the most persuasive evidence, particularly for the ability of reasonably developed states to sustain democracy once it has been established. See Przeworski, , Alvarez, , Cheibub, , and Limongi, , Democracy and Development: Political Institutions and Weil-Being in the World, 1950–1990 (New York:Cambridge University Press, 2000CrossRefGoogle Scholar).

³ See Collier, Paul and Hoeffler, Anke, Greed and Grievance in Civil War:Justice Seeking and Loot Seeking in Civil War (Washington, D.C.:World Bank, 2001Google Scholar); Fearon, James and Laitin, David, “Eth-nicity, Insurgency, and Civil War,” American Political Science Review 97, no. 2 (2003CrossRefGoogle Scholar). The idea of a “resource curse” is often traced to Jeffrey Sachs and Andrew Warner, “Natural Resource Abundance and Economic Growth,” Development Discussion Paper no. 517a (Cambridge: Harvard Institute for International Development, 1995). Michael L. Ross considers fuel (mostly oil and natural gas) and other mineral exports separately but gets somewhat stronger results for fuel; Ross, , “Does Oil Hinder Democracy?” WorldPolitics 53 (April 2001Google Scholar). That also seems more relevant for our focus on Islamic countries, for which petroleum products are more often the principal export. We use the measure of fuel exports as a percentage of GDP; it is more highly correlated with authoritarian government than is OPEC membership.

⁴ World Bank, World Development Indicators, http://devdata.worldbank.org/dataonline (accessed May 2003). We use the average measure of fuel exports as a percentage of GDP from the period 1991–2000. Data from earlier years were used for Afghanistan, Cambodia, Chad, Democratic Republic of Congo, Cuba, Guinea-Bissau, Guyana, Iraq, Liberia, Myanmar, Rwanda, Sierra Leone, Somalia, and Vietnam.

⁵ Data for about 1990 are mostly from Vanhanen, Tatu, “Domestic Ethnic Conflict and Ethnic Nepotism: A Comparative Analysis,” Journal of Peace Research 36, no. 1 (1999CrossRefGoogle Scholar).

⁶ Huntington, Samuel P., The Third Wave: Democratization in the Late Twentieth Century (Norman:University of Oklahoma Press, 1991), 46, 273Google Scholar–74.

⁷ Maoz, Zeev, Domestic Sources of Global Change (Ann Arbor:University of Michigan Press, 1996CrossRefGoogle Scholar)

⁸ Gleditsch, Kristian Skrede, All International Politics Is Local: The Diffusion of Conflict, Integration, and Democratization (Ann Arbor:University of Michigan Press, 2002CrossRefGoogle Scholar); Pevehouse, Jon, Democracyfrom Above? Regional Organizations and Democratization (New York:Cambridge University Press, 2004Google Scholar).

⁹ The Democracy in the Neighborhood variable is the average Polity score of contiguous states 1991–2000. Marshall, Monte and Jaggers, Keith, Polity IV Dataset, 2000Google Scholar. Polity is used more often in the international relations literature than are Freedom House rankings, which begin only in 1973; and the rankings for the 1970s were rather controversial. The difference is nevertheless not consequential, as one would expect by comparing Fish's Tables 2 and 3 with his Tables 4 and 5. Contiguity is coded as either directly contiguous by land or contiguous by sea within 150 miles; Douglas Stinnett, Jaroslav Tir, Philip Schafer, Paul Diehl, and Charles Gochman, “The Correlates of War Project Direct Contiguity Data, Version 3,” Conflict Management and Peace Science 19, no. 2 (2002).

¹⁰ A classic statement is Harold Lasswell's concept of the garrison-police state; Laswell, , National Security and Individual Freedom (New York:McGraw Hill, 1950Google Scholar). There is little evidence of an “autocratic peace” corresponding to the peace between democracies. See Mark Peceny and Caroline Beer, with Sanchez-Terry, Shannon, “Dictatorial Peace?” American Political Science Review 96, no. 1 (2002Google Scholar); idem, “Peaceful Parties and Puzzling Personalists,” American Political Science Review 97, no. 2 (2003), 339Google Scholar–42; Reiter, Dan and Stam, Allan C., “Identifying the Culprit: Democracy, Dictatorship, and Dispute,” American Political Science Review 97, no. 2 (2003), 333CrossRefGoogle Scholar–37

¹¹ Huntington (fn. 6), 270–71.

¹² Przeworski et al. (fn 2); Maoz (fn. 7).

¹³ Previous experience with democracy is measured as the average Polity IV Political Competition component score 1970–90. Transitional authority codes were deleted from the analysis. Zero is the lowest level of political competition and 10 is the highest. This index focuses on the presence and fairness of elections; Marshall and Jaggers (fn. 9).

¹⁴ On how an unbalanced sex ratio for unmarried young males may disrupt civil order and encourage autocratic government, see Hudson, Valerie M. and Den Boer, Andrea, “A Surplus of Men, a Deficit of Peace: Security and Sex Ratios in Asia's Largest States,” biternational Security 26, no. 4 (2002Google Scholar). Similarly, powerful men in polygamous societies control not only women but also the lower-class men who have no wives; see Betzig, Laura, Despotism and Differential Reproduction (New York:Aldine, 1986Google Scholar). Our analysis below, however, does not find that an imbalanced sex ratio produces authoritarian government generally.

¹⁵ Women in parliament is an imperfect measure, since some states with rubber-stamp parliaments have high female participation. We drop the Gender Empowerment Index, which was available to Fish for only 92 countries, of which just 20 are Islamic, as compared with 153 to 156 (up to 47 of them Islamic) for the other empowerment variables. Moreover, we think it preferable to assess the effect of specific women's rights rather than of such an aggregated measure. It performed less well in Fish's Table 10 and not better than specific measures in his Table 9. Similarly we do not use the Gender Development Index in United Nations Development Programme, Human Development Report 2002 (New York:Oxford University Press, 2002Google Scholar). An indicator of violence against women would be desirable, but we know of no adequate cross-national database. Low fertility rates might be used as an indicator of female reproductive rights, but they could also reflect coercive population policies enforced on women. Our data are primarily from the Human Development Report. See the appendix for precise definitions and information about missing data. We have changed some signs for consistency, so that all measures of the status of women equate higher values with higher status.

¹⁶ "Education shall be free, at least in the elementary and fundamental stages. Elementary education shall be compulsory”; rights “to a standard of living adequate for ... health and well-being [and] medical care”; “to take part in the government”; and “to just and favorable remuneration insuring an existence worth of human dignity.” The declaration, reflecting the usage of 1948, frequently refers to persons with the masculine pronoun.

¹⁷ See Sharabi, Hisham, Neopatriarchy: A Theory of Distorted Change in Arab Society (Oxford:Oxford University Press, 1988Google Scholar) and others cited in Fish's fn. 34 (p. 24). See also Fernea, Elizabeth, In Search of Islamic Feminism: One Woman's GlobalJourney (New York:Doubleday, 1998Google Scholar); and essays in Kurzman, Charles, ed., Liberal Islam: A Sourcebook (Oxford:Oxford University Press, 1998Google Scholar).

¹⁸ Alfred Stepan, with Graeme Robertson, “An ‘Arab’ More Than ‘Muslim’ Electoral Gap?” Journal of Democracy 14 (July 2003).

¹⁹ Lust-Okar, Ellen, “Why the Failure of Democratization? Explaining Middle East Exceptionalism” (Manuscript, New Haven, Yale University Political Science Department, May 2004Google Scholar).

²⁰ Arab League members are Algeria, Bahrain, Comoros, Djibuti, Egypt, Iraq, Jordan, Kuwait, Lebanon, Libya, Mauritania, Morocco, Oman, Qatar, Saudi Arabia, Somalia, Sudan, Syria, Tunisia, UAE, and Yemen.

²¹ See Reuveny, Rafael and Li, Quan, “The Joint Democracy-Dyadic Conflict Nexus: A Simultaneous Equations Model,” International Studies Quarterly Al, no. 3 (2003CrossRefGoogle Scholar). Pevehouse (fn. 8) finds that internalviolence often precedes transitions to democracy. Arab and MID involvement are only moderately correlated (.43). For data source, see the appendix. The period corresponds to that used for democracy in the neighborhood.

²² The Freedom House civil liberties checklist does include the following item (D.4): “Are there personal social freedoms, including gender equality, choice of marriage partners, and size of family?” But it is only one item out of the fourteen that constitute the civil liberties scale, which in turn is only half of the total freedom score, so it does not seriously contaminate use of that score as a dependent variable. We retain Fish's preference for measuring regime as a continuous rather than a dichotomous variable-as Przeworski et al. (fn. 2) do-despite some reservations about the truly interval character of Freedom House and Polity scores; see Vreeland, James, “A Continuous Shumpeterian Conception of Democracy” (Manuscript, New Haven, Yale University, Political Science Department, April 2003Google Scholar).

²³ He surprisingly omits OPEC membership from this table.

²⁴ This was a hard test, and even if Islam proved insignificant here it might well be that Islam had reduced democracy in the earlier period.

²⁵ Another variant of the MID measure—uses of military force over a longer preceding period (1970–90)—leaves Islam statistically insignificant and Arab statistically significant in all equations, but that analysis omits many successor states to Yugoslavia and the Soviet UniomTb lean against our own hypothesis and in favor of Fish's, we do not present it in the main text.

²⁶ A correlation matrix at the end of the appendix includes most variables. Many variables of female empowerment not shown are highly correlated with one another (but only one appears in any equation) and with Islamic population (as Fish hypothesizes). To check for mulicollinearity we entered the independent variables in this sequence: economic development, Islamic population, a measure of female empowerment, fuel exports, neighborhood, previous democracy, Arab League, and MID. Four female-empowerment measures (sex ratio, women in parliament, education ratio, and life expectancy) were insignificant in the presence of only economic development and Islamic population. Economic activity ratio became insignficant when fuel exports were added, as did literacy when Arab League was added. Only women in government stayed significant in the full model. A more systematic check computed the Variance Inflation Factor (VIF) for each model of our tables. It increases our confidence that multicollinearity is not distorting coefficient estimates. The mean VIF for the models of Table 1 did not exceed 2.09, nor 1.79 in Table 2; the highest VIF for an individual variable never exceeded 2.32. In Table 3 only economic development possibly warranted concern, with a VIF ranging from S.52 to 5.56 (but well below 10.0 for a definite problem). To be sure, we reran those equations without economic development and found few changes. The effects of Islamic population and Arab League remained the same; democracy became nonsignificant for the sex ratio after being marginally significant in Table 3.

²⁷ Our variable measures the percentage of women serving in government at the ministerial level in 1996. See the appendix to compare our measure with that of Fish.

²⁸ Huntington (fn. 6), 72–85. See also Mainwaring, Scott, The Catholic Church and Politics in Brazil, 1916–1985 (Stanford, Calif.:Stanford University Press, 1986Google Scholar); idem, “Democratic Survivability in Latin America,” in Handelman, Howard and Tessler, Mark, eds., Democracy and Its Limits: Lessonsfrom Asia, Latin America and the Middle East (Notre Dame, Ind.:University of Notre Dame Press, 1999Google Scholar); Hanson, Eric, The Catholic Church in World Politics (Princeton:Princeton University Press, 1987Google Scholar); Philpott, Daniel, “The Catholic Wave,” Journal' oj'Democracy 15 (April 2004): 32–36CrossRefGoogle Scholar.

²⁹ It is impossible to reproduce the full model from column 4 of Table 1, because not all the necessary data are available and sufficiently reliable. For democracy we use Polity scores (averaged over the decade in question), since Freedom House began its coding only in 1974. Measures of economic development and MID involvement are readily available for the earlier decades. So is previous experience with democracy, except we omit it for the 1960s because there were so many recently independent former colonies. Economic development and previous experience are averaged over each decade to maximize the number of observations. MID involvement is taken as a count over each decade. Lacking data on fuel exports we revert to OPEC membership as a proxy. Islamic and Catholic population percent-ages are relatively constant. Arab League membership changes, but the 2000 membership is still appropriate as a proxy for Arab population. We drop democratic neighborhood because it would require too much new data gathering on changes of borders and sovereign status. Measures of female empowerment in these years suffer from too many problems of data availability and quality-and they show hardly any impact in Table 1 anyway.

³⁰ The relationship remained positive in the 1990s, but only at p < .05 as democracy spread to Asia and elsewhere.

³¹ See Przeworski, Adam, Cheibub, Jose, and Limongi, Fernando, “Culture and Democracy,” World Culture Report: Culture, Creativity and Market (Paris:UNESCO Publishing, 1998Google Scholar). The authors identify three sets of views on whether democracy requires certain cultural preconditions: nonculturalist, weakly culturalist, and strongly culturalist.

³² Level of education is measured by the combined primary, secondary, and tertiary gross enrollment rates as percentage of population for 1999. The source is United Nations Development Programme (fn. 15); missing observations rilled in with 1996 data from World Bank (fn. 4).

³³ Percentage of population living in urban areas in 1997. The source is United Nations Development Programme (fn. 15); missing data filled in from World Bank (fn. 4). Education and urbanization are highly correlated with development and each other (see appendix table), though not with democracy. We include allforthe sake of identifying any remaining contribution of Islam, even though it may be hard to identify from the separate contributions.

³⁴ While we suggest this possibility, its correctness is not obvious. Agrarian production often requires women's full participation in the fields with men, getting them out of the house in ways that may be less available in urban society.

³⁵ Ghobarah, Hazem, Huth, Paul, and Russett, Bruce, “The Comparative Political Economy of Human Misery and Weil-Being,” International Studies Quarterly 48, no. 1 (2004CrossRefGoogle Scholar).

³⁶ As military conflict seems to be bad news for democracy, it may also be bad news for women's rights, as much feminist writing suggests. See Enloe, Cynthia, The Morning After: Sexual Politics at the End of the Cold War (Berkeley and Los Angeles:University of California Press, 1993Google Scholar); Tickner, Ann, Gendering World Politics: Issues and Approaches in the Post-Cold War Era (New York:Columbia sity Press, 2001Google Scholar); Peterson, V. Spike and Runyon, Anne Sisson, Global Gender Issues, 2nd ed. (Boulder, Colo.:Westview Press, 1999Google Scholar); Ember, Melvin, Ember, Carol, and Russett, Bruce, “Inequality and Democracy in the Anthropological Record,” in Midlarsky, Manus, ed., Inequality, Democracy, and Economic Development (Cambridge:Cambridge University Press, 1997Google Scholar); Goldstein, Joshua, War and Gender (Cambridge:Cambridge University Press, 2001Google Scholar); and Caprioli, Mary, “Gendered Conflict,” Journal of Peace Research 37, no. 1 (2000CrossRefGoogle Scholar). But when we added MIDs to the equations for women's rights, it was never significant (equations not shown here).

³⁷ We used a Durbin-Wu-Hausman test to check for possible endogeneity of democracy to women's status and vice versa. We found indications that democracy is endogenous to three of our measures of women's status at most: women in government, education ratio, and economic activity ratio. No measures of women's status were endogenous to democracy. This weak evidence of endogeneity indicates little need to create simultaneous equation models.

³⁸ The sex imbalance may result from Arab countries' relative denial of education, health care, nutrition, and economic independence to women. It is unlikely, however, that it results from practices of infanticide or sex-selective abortion. The former is uncommon in Islamic societies, and the latter is more likely to be practiced by nontraditional secular individuals.

³⁹ Nor are the results different when Catholic is substituted for Islamic, except that democracy falls just short (p < .06) of the p < .05 level, which was barely reached by Islamic.

⁴⁰ The matter of women's participation in high levels of the political system is nonetheless complex. We created an interactive variable, GDP x regime score. For women in parliament and our measure of women in government, the individual terms were significantly negative and the interaction significantly positive, meaning women did worse in wealthy autocracies and low-income democracies. Nepal, Central African Republic, and Papua New Guinea are examples of low-income democracies with poor representation of women in parliament and government, and Singapore and most oil-rich Arab states are high-income autocracies with almost no government posts occupied by women.

⁴¹ For example, Przeworski et al. (fn. 2). It is worth noting that no relationship between female empowerment and political freedom in Arab states is found in the Arab Human Development Report 2002 (New York:United Nations Development Programme, 2002Google Scholar), chap. 2.

⁴² For example, Iraq in 2003 was a very unpromising field in which to plant democracy. All the negative influences in Table 1, in addition to Islamic and Arab population, were present: fuel exports, low income, bad neighborhood, no history of democracy, and many MIDs.