Islam and Authoritarianism – Detailed Study Notes (Fish, World Politics, 2002)

Overview and Research Question

• Investigates whether predominantly Muslim (Islamic) countries are systematically less open/democratic than non-Muslim countries.

• Primary aim: test empirically if Islam is linked to authoritarianism after controlling for other determinants of regime type.

• Conceptual anchor: democracy defined via Dahl’s procedural minima (polyarchy) and Freedom House (FH) ratings as a practical measure of political liberties and civil rights; also uses Polity scores as a robustness check.

• Core finding (across analyses): predominantly Muslim countries are underachieving democratically compared to non-Muslim countries, even after accounting for development, culture, history, and resources. A key explanatory mechanism identified is the subordination of women in Muslim societies, which accounts for part of the Islam–democracy link but not all of it.

Democracy, Measurement, and Data

• Democracy definition used:

  • Dahl’s electoral/procedural conception of democracy (polyarchy): elections, elected accountability, broad eligibility, and rights enabling informed participation.

  • Freedom House (FH) score as a practical operationalization of democracy through political liberties and civil rights. FH scores run from 1 (most free) to 7 (least free); the study uses a ten-year average from 1991-1992 to 2000-2001, with the author re-scaling so higher scores indicate more openness.

  • Polity scores from the Polity Project as an alternative dependent variable (range:
    $-10 ext{ (most autocratic)} ext{ to } 10 ext{ (most democratic)}$). The study uses an eight-year average (1991-1998).

• Dependent variables used:

  • FH average score (1991-92 to 2000-01): denoted $FH_i$.

  • Polity average score (1991-1998): denoted $Polity_i$.

• Sample and data coverage:

  • Universe of cases = countries with population > 0.5 million.

  • FH data available for 157 countries; Polity data for 154 countries.

• Primary explanatory variable:

  • Islam dummy: Islami=egin{cases}1,& ext{Islamic religious tradition predominates in country }i\0,& ext{otherwise} \ ext{Eritrea excluded (two major confessions share equal adherents).} \ ext{Islamic countries count} N{Islam}=47.
    \ ext{Non-Islamic countries count} N_{Non-Islam}=109.
    \ ext{Note: a simple percentage Muslim variable was considered but a binary dummy is preferred for reliability and theoretical focus on predominance.}

• Controls (six main determinants of regime type):

  • Economic development: GDPpc{i,1990}$, logged:$ ext{log extbar GDP per capita extbar}{i,1990}$.</p></li><li><p>Sociocultural division: Ethnolinguistic fractionalization index (0 = complete uniformity, 1 = maximum fractionalization):$ELF_i \in [0,1]$(Ethnolologue data).</p></li><li><p>Economic performance: average annual growth rate of GDP per capita from 1975 to 1998:$gi= ext{avg}ig( rac{GDPpc{t}-GDPpc{t-1}}{GDPpc{t-1}}ig) ext{ for } t ext{ in }(1975-1998)
    \ ext{(units: percent per year)}$.</p></li><li><p>British colonial heritage: dummy$Brit_i$(former British colonies = 1).</p></li><li><p>Communist heritage: dummy$Comm_i$(post-communist heritage = 1).</p></li><li><p>Oil/resource endowment: OPEC membership dummy$OPEC_i$(member = 1).</p></li></ul></li><li><p>Robustness/alternative measures discussed:</p><ul><li><p>Agrarian share: percentage of population employed in agriculture/herding/fishing as alternative proxy for development/social structure.</p></li><li><p>Alternative ethnolinguistic measures: ethnic homogeneity index, and Krain/Taylor-Hudson ethnolinguistic fractionalization variants.</p></li><li><p>Additional robustness checks include substituting Polity scores for FH; using alternative samples; and checking for endogeneity concerns (2SLS attempted).</p></li></ul></li></ul><h3 id="645c24cb-790a-46c3-965c-da0b28dbfbbd" data-toc-id="645c24cb-790a-46c3-965c-da0b28dbfbbd" collapsed="false" seolevelmigrated="true">Hypotheses and Model Specification</h3><ul><li><p>Hypotheses (tractable for quantitative analysis):</p><ul><li><p>H1 (Islamic influence): Predominantly Muslim countries have worse democracy indicators, controlling for other determinants.</p></li><li><p>H2 (Developmental effect): Higher economic development (GDP per capita) is associated with higher openness, consistent with standard democratization theories.</p></li><li><p>H3 (Ethnolinguistic division): Higher ethnolinguistic fractionalization reduces openness/democracy.</p></li><li><p>H4 (Oil/resource effect): Oil-rich economies (OPEC membership) tend to be more autocratic due to rents and reduced incentives for accountability.</p></li><li><p>H5 (Legacy effects): British colonial heritage contributes positively, while post-communist heritage may have negative effects; both tested as binaries.</p></li></ul></li><li><p>Core regression equations (conceptual):</p><ul><li><p>FH model:$FHi = eta0 + eta1 Islami + eta2 ext{log GDPpc}{i,1990} + eta3 ELFi + eta4 gi + eta5 Briti + eta6 Commi + eta7 OPECi +
    u_i$</p></li><li><p>Polity model:$Polityi = heta0 + heta1 Islami + heta2 ext{log GDPpc}{i,1990} + heta3 ELFi + heta4 gi + heta5 Briti + heta6 Commi + heta7 OPECi + au_i$</p></li></ul></li><li><p>Note on specification and limits:</p><ul><li><p>The study acknowledges potential selection bias (non-random historical sampling) and bounded in-time inferences.</p></li><li><p>A dummy for Islam is preferred to a percentage Muslim measure due to reliability concerns in data sources for many countries.</p></li><li><p>The analysis uses robust standard errors to account for heteroskedasticity.</p></li></ul></li></ul><h3 id="a0fcbfd1-6f23-4c1d-a0ad-01e0c5ab8c78" data-toc-id="a0fcbfd1-6f23-4c1d-a0ad-01e0c5ab8c78" collapsed="false" seolevelmigrated="true">Descriptive and Preliminary Evidence</h3><ul><li><p>Table 1 (means by Muslim vs. non-Muslim):</p><ul><li><p>Muslim countries tend to have worse FH scores and worse Polity scores than non-Muslim counterparts.</p></li><li><p>Muslim countries also show some disadvantages in determinants of democracy (e.g., higher ethnolinguistic diversity, lower likelihood of British colonial heritage).</p></li><li><p>OPEC membership more common among Muslim countries.</p></li></ul></li><li><p>Implication: There is a descriptive pattern consistent with the Islam–democracy link, but many potential confounds exist; hence the need for multivariate analysis.</p></li></ul><h3 id="4126c49f-dac2-4560-b169-af309a6ee0fa" data-toc-id="4126c49f-dac2-4560-b169-af309a6ee0fa" collapsed="false" seolevelmigrated="true">Regression Evidence: Freedom House and Polity Scores</h3><ul><li><p>Methods: Ordinary Least Squares (OLS) regressions; both bivariate and multivariate specifications; robustness checks with alternate DV (Polity) and alternative measures of development and ethnicity.</p></li><li><p>Key results (FH as DV, Table 2 and Table 3):</p><ul><li><p>Islam dummy is significantly negative across models, indicating Muslim countries have lower FH scores (less open) than non-Muslim ones.</p></li><li><p>Economic development (log GDP per capita, 1990) is positively associated with FH openness.</p></li><li><p>Ethnolinguistic fractionalization has a negative association with FH, but its significance is unstable across specifications.</p></li><li><p>Economic performance (growth in GDP per capita, 1975-1998) is positively related to FH openness.</p></li><li><p>OPEC membership is associated with lower FH openness.</p></li><li><p>British colonial heritage is not consistently significant; communist heritage is not significant in many specifications.</p></li><li><p>Agrarian share (alternative development measure) remains robustly significant; Islam remains highly significant even when agrarian share substitutes GDP per capita.</p></li><li><p>The magnitude: in the most complete model (Table 3, Model 1), the Islam coefficient is around$eta_1
    oughly -1.2 ext{ to } -1.3$for FH; in some specifications it reaches up to about$-1.68$; an order-of-magnitude-dominant effect relative to the FH scale.</p></li></ul></li><li><p>Robustness with Polity DV (Tables 4 and 5):</p><ul><li><p>Using Polity scores, Islam remains associated with a substantial negative effect (e.g., a predominantly Islamic tradition linked with about a$-7$point reduction in Polity, i.e., roughly one-third of the regulatory range).</p></li><li><p>Development and OPEC maintain their signicant associations; British/communist heritage remain largely insignificant.</p></li><li><p>The results using Polity as the DV are consistent with FH results, supporting the Islam–democracy link from another democratic-eligibility dimension.</p></li></ul></li><li><p>Robustness checks described:</p><ul><li><p>Alternative development measures (agrarian share) yield substantively similar patterns; Islam coefficient remains large and highly significant.</p></li><li><p>Alternative ethnolinguistic measures yield the same qualitative results that ethnicity alone does not explain away the Islam effect.</p></li><li><p>Substituting Polity for FH yields consistent conclusions about the Islam variable.</p></li><li><p>Some discussion of endogeneity (2SLS) is offered; results are substantively similar, though instrumenting such social-psychological variables is methodologically challenging.</p></li></ul></li></ul><h3 id="edeaa989-8a1d-468a-99ba-ab64df1b7376" data-toc-id="edeaa989-8a1d-468a-99ba-ab64df1b7376" collapsed="false" seolevelmigrated="true">What These Results Do Not Support</h3><ul><li><p>Plausible but unsatisfactory ideas investigated and weighed:</p><ul><li><p>Islam–poverty link explanation: Muslim countries are poorer on average, but accounting for development does not erase the Islam effect on FH/Polity.</p></li><li><p>Oil rents explanation: OPEC membership accounts for some variance but does not fully explain the Islam–democracy link; oil rents are not a complete driver of democratic deficit.</p></li><li><p>Ethnic fractionalization: While Muslim countries tend to be more diverse, heterogeneity does not by itself predict openness; Islam remains significant after controlling for ethnic composition.</p></li><li><p>Violence propensity: Contrary to Huntington’s claims, the data do not show Muslims are inherently more violent; after controlling for development, Islam is not significantly related to lower stability or higher violence in multivariate models.</p></li><li><p>Trust: World Values Survey-based trust indicators do not show a robust, unconditional lower trust in Muslim countries when development is controlled; the trust gap observed in simple comparisons vanishes with a proper control for development.</p></li></ul></li><li><p>Overall takeaway: While several popular explanations fail to fully account for the Islam–democracy link, the empirical results show a robust association between predominantly Muslim status and lower political openness, even after accounting for multiple alternative determinants. The link is not reducible to poverty, oil, ethnicity, or violence alone.</p></li></ul><h3 id="76e2de4a-16ae-4703-b60e-523f4fb42592" data-toc-id="76e2de4a-16ae-4703-b60e-523f4fb42592" collapsed="false" seolevelmigrated="true">The Connection Between Islam and Authoritarianism: A Plausible but Empirical Mechanism</h3><ul><li><p>The “problem of female subordination” as a leading mechanism:</p><ul><li><p>Muslim societies exhibit larger gaps between male and female education (literacy), skewed sex ratios (males to females), lower women’s representation in government, and lower Gender Empowerment Measures (GEM).</p></li><li><p>Primary indicators used:</p></li><li><p>Literacy gap: difference between male and female literacy rates (1990 data used as baseline; later tables show multiple checks).</p></li><li><p>Sex ratio: number of males per 100 females (cross-country comparison; issues with outliers in Gulf states due to foreign workers are discussed).</p></li><li><p>Women in government: percentage of high-level positions held by women (e.g., ministers, heads of agencies).</p></li><li><p>GEM: UNDP Gender Empowerment Measure (0 to 1) reflecting income, workplace status, and legislative presence of women.</p></li></ul></li><li><p>Descriptive differences (Table 8):</p><ul><li><p>Literacy gap (Muslim) ≈ 18.7 percentage points vs Catholic ≈ 4.3 points.</p></li><li><p>Sex ratio (Muslim vs Catholic) shows Muslims have higher male-to-female ratios (roughly 102/100 vs about 97/100 for Catholics).</p></li><li><p>Women in government (Muslim ≈ 5.2$ ext{FH change} imes 0.04$per point of gap), so a 20-point literacy gap relates to about a$0.8$FH-point decrease.</p></li><li><p>The sex ratio difference (Muslim vs Catholic) also associates with meaningful differences in democratic openness; higher male dominance correlates with lower FH openness and lower government representation by women.</p></li><li><p>Each additional 1 percentage of women in high government positions is linked to an improvement of roughly$0.08$FH points.</p></li><li><p>Endogeneity concerns acknowledged; exploratory 2SLS analyses yield substantively similar results, but identifying strong instruments for gender-related variables is difficult.</p></li></ul></li><li><p>Synthesis: The status of women emerges as a systematic, significant mechanism linking Islam to lower openness, but it does not fully account for the Islam–democracy relationship. Other factors, such as resource endowments and social-structural characteristics, also contribute.</p></li></ul><h3 id="4d569900-9329-4877-9615-07a18b88f568" data-toc-id="4d569900-9329-4877-9615-07a18b88f568" collapsed="false" seolevelmigrated="true">Interpreting Mechanisms: Patriarchy, Kinship, and Social Structure</h3><ul><li><p>The author proposes a provisional theory about how women’s subordination could influence regime type:</p><ul><li><p>Patriarchal social order (family-level hierarchies) may reproduce in broader society, shaping political culture toward acceptance of authority and reduced political contestation.</p></li><li><p>Isomorphism between primary social relations (family, kinship) and political life may condition how authority is exercised and challenged in the public sphere.</p></li><li><p>Men’s political behavior under kin-based power dynamics may be more tolerable of hierarchy and less conducive to broad-based democratic norms.</p></li></ul></li><li><p>Alternative perspectives discussed:</p><ul><li><p>Kin-based networks could also foster civic associations and citizenship rights under certain conditions, potentially promoting democracy; the evidence is mixed and context-dependent.</p></li><li><p>State formation, colonial legacies, and modernization pathways interact with kin-based structures to shape political outcomes.</p></li></ul></li><li><p>Important caveats:</p><ul><li><p>The causal direction could be bidirectional: regime type can influence women’s status as well (endogeneity concern).</p></li><li><p>A single variable (e.g., Islam or women’s status) is unlikely to capture the whole story; multiple interacting pathways likely operate.</p></li></ul></li><li><p>Implications for democracy:</p><ul><li><p>Variation within the Muslim world matters: different countries exhibit different levels of female empowerment, literacy gaps, and sex ratios, with corresponding implications for democratization prospects.</p></li><li><p>India is highlighted as an exceptional case where high religious diversity and unique political dynamics coexist with relatively open governance, suggesting that patriarchy is not a universal predictor of democratic failure.</p></li><li><p>The analysis calls for attention to within-country variation, policy efforts to improve girls’ education and female representation, and broader structural reforms that may foster democratic deepening.</p></li></ul></li></ul><h3 id="f24a1494-072d-4d51-8b84-b6866a21a137" data-toc-id="f24a1494-072d-4d51-8b84-b6866a21a137" collapsed="false" seolevelmigrated="true">Variation Within the Muslim World and Cross-Regional Comparisons</h3><ul><li><p>Countries with extreme indicators (high male literacy gaps, skewed sex ratios, low female political representation) tend to be less conducive to democracy:</p><ul><li><p>Examples with particularly unfavorable indicators: Afghanistan, Bangladesh, Nigeria, Pakistan, Saudi Arabia, Somalia, Syria, and others.</p></li><li><p>Countries with relatively better indicators on these fronts (and potential for democratization) include Malaysia, Indonesia, and others with smaller literacy gaps and greater women’s representation in government.</p></li></ul></li><li><p>Within non-Muslim countries, the India case illustrates that patriarchy is not a universal predictor of democratic outcomes; other political, social, and economic factors can sustain open governance despite patriarchal patterns.</p></li><li><p>The interplay of economic development, oil dependence, and social structure continues to shape democracy prospects across regions:</p><ul><li><p>Oil dependence (OPEC membership) correlates with more autocratic outcomes, even after accounting for Islam.</p></li><li><p>Economic development has a robust positive association with openness, ensuring that wealthier societies tend to be more democratic, but there are exceptions (e.g., some wealthy regimes with limited political rights).</p></li></ul></li></ul><h3 id="e6beed1f-8e45-4600-82a1-12005be534f1" data-toc-id="e6beed1f-8e45-4600-82a1-12005be534f1" collapsed="false" seolevelmigrated="true">Implications for Policy and Democracy Theory</h3><ul><li><p>The empirical results emphasize within-Muslim variation and caution against monolithic stereotypes about Islam and democracy.</p></li><li><p>A key implication is that improving women’s education, reducing gender gaps, and increasing women’s political participation may meaningfully contribute to democratic deepening—even in Muslim-majority contexts.</p></li><li><p>Caution against overreliance on culturalist explanations: development, resource economics, and governance quality are significant drivers; cultural attributions to Islam are insufficient to explain the democratic deficit.</p></li><li><p>The study calls for nuanced, context-sensitive theory-building that accommodates variation across countries and over time, rather than sweeping generalizations about Islam and political order.</p></li></ul><h3 id="f7c84256-e6a9-41c4-a592-707e950f5aa5" data-toc-id="f7c84256-e6a9-41c4-a592-707e950f5aa5" collapsed="false" seolevelmigrated="true">Relation to Other Works and Critical Perspective</h3><ul><li><p>The article situates its claims vis-à-vis debates about Muslim democracies and the so-called “democracy gap” (Karatnycky; Waterbury) and contrasts with arguments suggesting Islam is inherently incompatible with democracy.</p></li><li><p>It also engages with Huntington’s claims about violence and civilizational conflict, testing them against empirical indicators of political violence and violence-related governance.</p></li><li><p>The analysis aligns with broader democratization literature that emphasizes economic development, governance, and social capital, while highlighting gendered dimensions as a potentially underappreciated channel.</p></li></ul><h3 id="79d07cbe-3927-477a-ac9a-8a0e0b7c4610" data-toc-id="79d07cbe-3927-477a-ac9a-8a0e0b7c4610" collapsed="false" seolevelmigrated="true">Key Takeaways and Summary Points</h3><ul><li><p>Core empirical finding: Predominantly Muslim countries tend to score lower on political openness (FH and Polity) even when controlling for GDP per capita, growth, ethnic fractionalization, colonial heritage, communist heritage, and oil endowment.</p></li><li><p>The Islam variable remains robust across multiple specifications and dependent variables, indicating a substantive relationship beyond simple confounds.</p></li><li><p>A major mechanism is the status of women: literacy gaps, sex ratio imbalances, and low women’s representation in government help explain part of the Islam–democracy link, though not all of it.</p></li><li><p>Other explored explanations (poverty, oil rents, ethnic diversity, and violence propensity) do not fully account for the observed pattern; many fail to hold once robust controls are added.</p></li><li><p>The findings invite a cautious, non-essentialist approach to religion and politics, recognizing substantial intra-Islam variation and the potential for policy and institutional change to alter trajectories toward democracy.</p></li><li><p>The paper ultimately underscores the complexity of democratization, the need for nuanced cross-national analysis, and the importance of addressing gender inequality as part of broader democratic consolidation.</p></li></ul><h3 id="73771ee6-493a-4ab2-a233-e95fdff217ba" data-toc-id="73771ee6-493a-4ab2-a233-e95fdff217ba" collapsed="false" seolevelmigrated="true">Notable Data References and Formulas (LaTeX)</h3><ul><li><p>Dependent variables:</p><ul><li><p>Freedom House score (FH):$FH_i^{(ten ext{-}yr)} ext{ averaged across } t ext{ in } [1991 ext{--}1992, 2000 ext{--}2001]$; higher is more open.</p></li><li><p>Polity score:$Polity_i^{(8 ext{ yr})} ext{ averaged across } t ext{ in } [1991,1998]$; higher is more democratic.</p></li></ul></li><li><p>Islam dummy:$Islam_i \in {0,1}$(1 = predominantly Muslim).</p></li><li><p>Controls (illustrative notation):</p><ul><li><p>$GDPpc{i,1990}^{ ext{log}} = \log\big(\text{GDP per capita}{i,1990}\big)$</p></li><li><p>$ELF_i \in [0,1]$(Ethnolinguistic Fractionalization)</p></li><li><p>$gi = \text{avg annual growth of } GDPpc{i,t} \text{ for } t \in [1975,1998]\, (\% )$</p></li><li><p>$Briti, Commi, OPEC_i \in {0,1}$</p></li><li><p>Alternative development:$AgrarianShare_i = \frac{\text{Agrarian employed}}{\text{Total employed}}$</p></li></ul></li><li><p>Key regression form (FH as DV):</p><ul><li><p>$FHi = \beta0 + \beta1 Islami + \beta2 \log GDPpc{i,1990} + \beta3 ELFi + \beta4 gi + \beta5 Briti + \beta6 Commi + \beta7 OPECi + \varepsilon_i$</p></li></ul></li><li><p>Key regression form (Polity as DV):</p><ul><li><p>$Polityi = \gamma0 + \gamma1 Islami + \gamma2 \log GDPpc{i,1990} + \gamma3 ELFi + \gamma4 gi + \gamma5 Briti + \gamma6 Commi + \gamma7 OPECi + \eta_i$</p></li></ul></li><li><p>Female-status mechanism (illustrative effects reported):</p><ul><li><p>Literacy gap effect on FH: approximately$\Delta FH \approx -0.04 \times (\text{literacy gap in percentage points})$per percentage point gap.</p></li><li><p>Women in government effect: each 1$+0.08$$ FH points.

  • Sex ratio effect: higher male/female ratios correlate with lower openness; controls for literacy gap and development reduce some of this association.