"The Missionary Roots of Liberal Democracy", Woodberry 2012 http://www.hillcountryinstitute.org/wp-content/uploads/MissionaryRootsOfLiberalDemocracy.pdf ; excerpts:
"This article demonstrates historically and statistically that conversionary Protestants (CPs) heavily influenced the rise and spread of stable democracy around the world. It argues that CPs were a crucial catalyst initiating the development and spread of religious liberty, mass education, mass printing, newspapers, voluntary organizations, and colonial reforms, thereby creating the conditions that made stable democracy more likely. Statistically, the historic prevalence of Protestant missionaries explains about half the variation in democracy in Africa, Asia, Latin America and Oceania and removes the impact of most variables that dominate current statistical research about democracy. The association between Protestant missions and democracy is consistent in different continents and subsamples, and it is robust to more than 50 controls and to instrumental variable analyses.
Previous quantitative research consistently suggests that countries with more Protestants are more democratic and have more stable democratic transitions (Bollen and Jackman 1985; Hadenius 1992; Treisman 2000; Tusalem 2009). However, this earlier research measured Protestant influence less precisely and more recently than my study (i.e., after CP-initiated behavior spread to other groups and become institutionalized) and thus found results less dramatic than those in this article.
Current statistical evidence of an association between Protestantism and democracy matches historical evidence that Protestantism facilitated the development of modern representative democracy (e.g., Bradley and Van Kley 2001; Clarke 1994; Witte 2007). For example, stable democracy first emerged in Protestant Europe and British-settler colonies, and by World War I every independent, predominantly Protestant country was a stable democracy-with the possible exception of Germany.3 Less stable versions of democracy developed in Catholic areas with large Protestant and Jansenist4 minorities, such as France (Anderson 2004; Philpott 2004; Woodberry and Shah 2004). However, democracy lagged in Catholic and Orthodox parts of Southern and Eastern Europe where Protestants had little influence. A similar pattern existed outside Europe (Woodberry 2004c).
some argue that the association between Protestantism and democracy is spurious, even historically (Moore 1966; Rueschemeyer, Stephens, and Stephens 1992; Swanson 1967). For example, in Europe, pre-Reformation class structures, land-holding patterns, and political conditions may have influenced both the spread of Protestantism and the later development of democracy, thereby creating a deceptive association. Unfortunately, differentiating cultural and instrumental causes is always difficult. In any given context, possible causes are so enmeshed that they are difficult to untangle. For every proposed cultural or religious "cause," scholars can find an alternative economic or political "cause," and vice versa. To escape this swamp of indeterminate causality I use several approaches: (1) observing the consistent association between Protestantism and democracy in regions with histories and class structures radically different from those of Europe; (2) showing historically that CPs had a unique role in spreading mass education, printing, civil society, and other factors that scholars argue fostered democracy; and (3) demonstrating statistically that the historic prevalence of Protestant missionaries strongly predicts democracy in 142 non-European societies using (a) controls for alternative explanations and (b) instrumental variable estimation. These different analyses consistently demonstrate a strong link between CPs and democracy, making it extremely difficult to find a consistent alternative explanation.
As mentioned previously, Protestantism is associated with stable, representative democracy in Western Europe (Context 1), although many argue that this association is spurious (Moore 1966; Rueschemeyer, Stephens, and Stephens 1992; Swanson 1967). Because Protestantism spread to many contexts with different class structures, land-holding patterns, and the like, these varying pre-Protestant conditions can be used to help adjudicate between theories. If the association between Protestantism and democracy remains consistent regardless of context, the claim that the association is caused by these pre-Protestant conditions becomes less plausible. We now turn to four other contexts.
Context 2: Among European-settler colonies, "Protestant-based" United States, Canada, Australia, and New Zealand have been more democratic than "Catholic-based" Argentina, Chile, Uruguay, and Costa Rica. Both sets of countries had similar precolonial conditions (e.g., temperate climates, communal land holding, and small indigenous populations), which weakens theories that climate or pre-Protestant class conditions caused the Protestantism-democracy association. Differences between Protestant- and Catholic- settler colonies after the arrival of white-settlers may be influenced by religion and thus be intervening mechanisms rather than competing explanations. Context 2 also weakens theories that secularization causes democracy (e.g., the United States is far more religious than Uruguay). Still, all predominantly "Protestant" areas were British colonies, and all "Catholic"6 areas were Spanish colonies. Thus colonial institutions may be the crucial factor. Yet whatever the mechanisms are, they seem to be transportable from Europe to other countries.
Contexts 3 and 4: After the fall of communism, Eastern European Catholic and Protestant countries (Poland, Hungary, East Germany, Slovakia, Slovenia, the Czech Republic) had earlier, more stable democratic transitions than did Orthodox Christian and Muslims ones (Albania, Romania, Bulgaria, Serbia, Bosnia). Similarly, Protestant and Catholic former Soviet republics (Latvia, Lithuania, Estonia) had earlier, more stable transitions than did Orthodox and Muslim ones (Anderson 2004; Woodberry 2000). None of these countries were British colonies or had mass immigration from Northwest Europe; this weakens nonreligious explanations for the Protestantism-democracy association in Context 2. Moreover, all the countries in Contexts 3 and 4 had similar pre-transition institutions and entered a similar international environment. All had large secular populations and comparable exposure to Marxist and Enlightenment ideas via monopoly state education. In addition, communists eliminated historic differences in land holdings. Yet in both Contexts 3 and 4, religious differences predict both who mobilized against communism and how smoothly states made the transition to democracy. Catholic and Protestant countries became similarly democratic, but the transitions occurred after the Catholic Church's rapprochement with democracy and in areas where Protestants and Catholics had competed for centuries. "Nonreligious" explanations for the pattern-such as the legacy of Ottoman colonization-might work for Eastern Europe, but not for the former Soviet Union or for Contexts 1, 2, and 5 (see Tables 18 and 19 in the supplemental Online Appendix available at http://www.journals.cambridge.org/psr2012005). Nor is it clear that the Ottoman "influence" was not related
to religion.
Context 5: Finally, if we exclude all European countries and all Protestant European-settler colonies from the sample (i.e., Contexts 1-4) and analyze the remaining countries in Africa, Asia, Latin America, and Oceania, we still find a statistical association between Protestantism and democracy. Thus in at least five distinct contexts the Protestantism-democracy association holds. None of these tests are decisive, and it is possible to think of ad hoc alternative explanations in each context. Yet the consistent association between Protestantism and democracy across all five contexts strengthens the plausibility of causation. It is not clear if any competing theory works in all five contexts or why we should prefer inconsistent explanations over a consistent one.
Contrary to what many theories of "secular modernity" argue, democracy was not a triumph of secularism over religion. From the seventeenth through the mid-twentieth centuries, activist Protestants instituted and spread many of the reforms that made modern representative democracy more likely (see the section, Historical Evidence).
The section, Statistical Evidence, demonstrates that the historic prevalence of Protestant missionaries explains the variation in democracy better than either the prevalence of the nonreligious or of generic Protestants. Moreover, Protestant missions predict democracy, whereas Catholic missions do not. Yet there is no evidence that land-holding patterns (or other theories used to discount the Protestantism-democracy association in Europe and the Americas) shaped the spread of Protestant missionaries but did not shape the spread of Catholic missions (e.g., see Tables 20 and 21 in the supplemental Online Appendix).
Some may argue that the influence of missionaries was too anemic to foster democracy, but before the mid-twentieth century, missionaries were the main source of information about life in the colonies (Fairbank 1985; Hutchison 1987, 1; Tudesco 1980, 56). Moreover, missionaries constituted one of the largest and most educated groups of Westerners in the non-Western world-most had college degrees when few others had them (Daughton 2006; Hutchison 1987). In the Anglo-Protestant world, missionary organizations dwarfed labor unions and other nongovernmental organizations (NGOs).7 In fact, in the nineteenth century in the United States, the largest mission organizations outstripped all but a few commercial banks in size and financial resources (Chabbott 1999, 226-31; Hall 1994; Hutchison 1987). Yet, scholars often argue that labor unions, NGOs, and financial interests influenced democracy.
Nineteenth-century Bible and tract societies were among the largest corporations of any kind (Hall 1994, 34, 44); from 1829 to 1831 the American Bible Society printed and distributed more than a million Bibles at a time when the United States only had about three million households, no railroad system, and a dispersed rural population (Nord 2004, 84).
In fact, arguments for political pluralism, electoral reform, and limitations of state power were originally framed in religious terms (Bradley and Van Kley 2001; Clarke 1994; Ihalainen 1999; Lutz 1988; 1992; Nelson 2010; Witte and Alexander 2008).
For example, Calvinists tried to reconstruct states along "godly" lines and limit sinful human institutions. Perhaps as a result, most Enlightenment democratic theorists came from Calvinist families or had a Calvinist education, even if they were either not theologically orthodox or personally religious (e.g., John Locke, Rousseau, Hugo Grotius, Benjamin Franklin, John Adams, Patrick Henry, James Madison, and Alexander Hamilton),9 and they secularized ideas previously articulated by Calvinist theologians and jurists (Hutson 1998; Lutz 1980; 1988; Nelson 2010; Witte 2007).10 For example, Hobbes' and Locke's social contracts are secular versions of Puritan and Nonconformist covenants, and Locke's ideas about the equality of all people are explicitly religious (Waldron 2002; Woodberry and Shah 2004).
Although stated in secular form, the U.S. Constitution and Bill of Rights derive most directly from earlier colonial covenants, compacts, and bills of rights that were generally justified explicitly in biblical and theological terms; many were written before Hobbes and Locke expounded their ideas. Only 7 of the 27 rights enumerated in the U.S. Bill of Rights can be traced to major English common law documents (Lutz 1980; 1988; 1992; Witte 2007). Even between 1760 and 1805, political writings quoted the Bible more often than either Enlightenment or classical thinkers (34% versus 22% and 9%, respectively; Lutz 1984). Furthermore, the strength of Calvinism and Nonconformism better predicts where democracy emerged than does the strength of Greek and Enlightenment influence. Greek classics were most consistently available in the Eastern Mediterranean and the Muslim world, but democracy did not thrive there; the Roman Empire circled the Mediterranean, and the Renaissance flourished in Southern Europe, but democracy did not thrive in those places either. The "Athenian seed" germinated only after 2,100 years in alien soil: Northwest Europe and North America. Thus, areas with later and weaker exposure to Greek thought would have to have had "stronger pro-democratic effects." At a minimum, some additional catalyst seems likely.
CPs' catalytic effect on printing is clear from the shifts in the printing centers of Europe. Before the Reformation, Italy had the largest printing industry, but Protestantism made little headway there and printing did not increase rapidly or birth either an early public sphere or mass literacy (Graff 1987, 112-19). In contrast, England initially had little printing activity (Graff 1987, 115), but CPs used printing to mobilize ordinary people, forcing their elitist enemies to respond in kind. This activity spurred the growth of newspapers, printed debates, and an early public sphere (Zaret 2000). CPs had similar effects in continental Europe (e.g., Germany, France, Switzerland, and the Netherlands; Eisenstein 1979, 312-452; Febvre and Martin 1976, 287-319; Melton 2001). However, non-state Protestants were weaker there, and extended religious wars destroyed their early gains. Still, from the 1600s on, even in continental Europe, Protestant areas consistently printed more books per capita and exported more printed material per capita than Catholic areas (Eisenstein 1979, 403-23).
When Muslims, Hindus, and Theravada Buddhists engaged in printing, it was usually a response to mass printing by Protestant missionaries or by those trained by them (e.g., Lebanon, Syria, Sudan, Persia/Iran, Malaysia, India, Burma, Sri Lanka, Thailand; Green 2009; Woodberry 2011c).15 CPs printed so many vernacular texts that it forced elite response. For example, within 32 years of importing a press to India in 1800, three British missionaries printed more than 212,000 copies of books in 40 languages and, along with other missionaries, created the fonts and paper that dominated South Asian printing for much of the nineteenth century (Ghosh 2003, 27; Khan 1961; Ross 1999, 40-77, 118). This spurred both Hindus and Muslims to respond, but the earliest Indian printers learned their skills at mission presses, and most early non-Christian Indian imprints were religious-often rebuttals to missionary tracts (Aggarwal 1988, 2; Ghosh 2003; Khan 1961; Robinson 2000, 77; Shaw 2007). In fact, in most Asian societies early indigenous printers gained their skills and equipment from Protestant missionaries (Green 2009; Woodberry 2011c). Conversionary Protestantism also shaped printing's "consequences." If printing was a sufficient cause for mass literacy, newspapers, and the public sphere, then we would expect these developments to have originated in China, Korea, or Japan, but they did not. Printing occurred in China, Korea, and Japan 600-800 years earlier than in Europe. China and Korea had movable-font metal type before Europe; Japan had movable-font metal type starting in 1590; and Korea and Japan had phonetic alphabets, which facilitated literacy and made movable font efficient. All three countries had a high level of economic development and thriving mercantile classes. However, until CPs arrived in the nineteenth century, printing never supplanted handwritten manuscripts, newspapers did not develop, and literacy remained primarily the prerogative of elite men (Davis 1994; Duchesne 2006, 82-83; Reed 2007; Su 1996).
...Protestant missionaries printed the first East Asian newspapers (in Chinese);16 Chinese reformers copied them (Dunch 2001, 78; Zhang 2007), as did Japanese and Koreans, who read mission publications in Chinese and visited missionaries while their countries were still closed to Westerners (Altman 1966, 23-27, 37, 41-42; Inglehart 1959, 40; Lutz 2008, 92-96; K. Shin 1999; Y.H. Shin 1984). Christian converts published the first privately printed Japanese and Korean-language newspapers (Davis 1988; Huffman 1997, 30-31, 410).17 Protestant missionaries also reintroduced movable-font metal type (which Asians had abandoned) and developed the fonts and techniques that dominated nineteenth- and early-twentieth-century printing throughout East Asia (Altman 1966, 100-7; Lutz 2008, 173-75; Su 1996; Zhang 2007). Similarly, Protestant missionaries initiated newspapers in most other non-European societies (Woodberry 2011c).
However, missionaries were different. First, many nineteenth- and early-twentieth-century Protestant missionaries came from politically activist traditions. In much of Northwest Europe and in English-settler colonies (excluding slave-holding U.S. states), the Protestant missions movement was closely tied to social reform movements such as abolition and temperance. Thus many missionaries perceived societal reform as a natural extension of their faith (Etherington 2005; Masters and Young 2007; Young 2006). Second, the abuses made mission work more difficult because they angered indigenous people, turning them against Christianity, which many indigenous people associated with the colonizers. Finally, missionaries had the power to fight abuses because they wrote regularly to supporters in colonizing states. During the nineteenth and early twentieth centuries Europeans and North Americans got most of their news about colonized territories from missionary periodicals, which encouraged people to care about distant people they otherwise could have ignored (Miller and Stanczak 2009; Stamatov 2010; Woodberry 2004c; 2006a; 2006b).
In British and American colonies, religious liberty and private mission financing weakened officials' ability to punish missionaries and freed missionaries to critique abuses (Greenlee and Johnston 1999, 34-38), while popular support allowed missionaries to punish colonial officials and settlers. For example, colonial magistrates and governors were reprimanded or removed, military officials were put on trial for murder, confiscated land was returned to indigenous people, and so on (Etherington 2005; Hincks 2009, 181; Oddie 1978; 1996; Stocking 1987, 240-54, 272; Turner 1998; Woodberry 2004c). Thus, Protestant missionaries spurred immediate abolitionism (Stamatov 2006; Woodberry 2004c; 2006a; 2006b), as well as movements to protect indigenous land rights, prevent forced labor, and force the British to apply similar legal standards to whites and nonwhites (Chaudhuri 1998; Clements 1999; Etherington 2005; Gladwin 2007; Grant 2005; Knaplund 1953; Oddie 1978; Turner 1998; Woodberry 2004b; 2004c; 2006b; 2011a). Although others participated in these movements, it was the missionaries who provided detailed information and photographs that documented atrocities. Missionaries also provided emotional connections to distant people and mobilized large groups through church talks and mission presses. Without missionaries, mobilizing mass protests would have been difficult (Grant 2005; Etherington 2005; Stamatov 2010; Woodberry 2004b; 2004c; 2006a; 2006b). The missionary-enabled mobilization made it more difficult for the British to sustain colonial violence or to apply different legal standards to whites and nonwhites. It helped create a cocoon in which nonviolent, indigenous political movements could develop and increased the incentives for colonial officials to allow gradual democratization and decolonization.
R-squareds range from .439 to .849, suggesting that the models predict the distribution of missionaries well. In both tables, the sign and significance levels of coefficients predicting the prevalence of Protestant and Catholic missionaries are almost always consistent, except for variables related to colonizers. Protestant missions had greater prevalence in "Protestant" colonies, and Catholic missionaries had greater prevalence in "Catholic" colonies.
Documentary sources also suggest more missionaries went to places with higher conversion rates, more ethnolinguistic diversity, a greater history of slave trading, and more media attention (e.g., that given to Captain Cook's voyage). However, these factors seem either unrelated to democracy (media attention) or negatively related to it. Conversions were higher among poor, marginalized groups, and poverty, discrimination, ethnolinguistic diversity, and extensive exposure to slave raiding seem unlikely to promote democracy and thus unlikely to increase the association between Protestant missions and democracy.
Both historical and statistical evidence suggest that CPs promoted democracy, although often through indirect means. In all five contexts analyzed - Western Europe, Eastern Europe, the former Soviet Union, European-settler colonies, and mission territories - Protestantism is associated with democracy. Comparative historical analyses show that CPs consistently initiated and spread factors that past research suggests promote democracy: mass printing, mass education, civil society, and colonial rule of law. In cross-national statistical analysis Protestant missions are significantly and robustly associated with higher levels of printing, education, economic development, organizational civil society, protection of private property, and rule of law and with lower levels of corruption (Woodberry 2004a; 2004c; 2006c; 2011b; 2011c; and Table 22, Online Appendix). Moreover, wherever they have been tested, these patterns repeat at the subnational level (Bai and Kung 2011; Gallego and Woodberry 2010; Lankina and Getachew n.d.; Nunn 2010; Woodberry 2004a). Finally, statistical analysis suggests that Protestant missions are strongly and robustly associated with democracy. In fact, missions seem to explain about half the variation in democracy outside Europe and survive dozens of controls and robustness checks.
If omitted variable bias caused the entire association between Protestant missions and democracy, the omitted variable(s) would need to be strongly correlated with both democracy and Protestant missions, but not correlated with Catholic missions (even though Protestant and Catholic missions are highly correlated). More concretely, the cumulative correlation between Protestant missions and democracy is .707.47 If Protestant missions did not cause democracy, then in a properly specified model the correlation would be zero, and the .707 correlation would be the product of the omitted variable(s)'s correlation with democracy and the omitted variable(s)'s correlation with Protestant missions. This requires a mean correlation of .841 with each (.841 ∗ .841 = .707). If true, the omitted variable(s) should be nameable because it would virtually dictate both democracy and Protestant missions. None of the variables used in this article have correlations close to .841 except for the two measures of democracy used in Table 7 (BP and Polity IV), which have a correlation of .844. The three measures related to Protestant missions have correlations with each other of between .332 and .390. Thus, the omitted variable(s) would have to be as correlated with democracy as another measure of democracy is and simultaneously far more correlated with Protestant missions than any of the other measures of Protestant missions are. That seems highly unlikely.
In addition, the relationship between Protestant missions and democracy holds in widely different samples (i.e., sample sizes vary between 26 and 142). The relationship holds if we change regions of the world, if we limit the sample to British colonies, if we drop Muslim societies and Caribbean islands, if we change measures of democracy, or if we do instrumental variable estimation in nine different ways. Moreover, the Protestant mission coefficients are extremely consistent between models, even using different excluded instruments (especially the coefficients for Years of Protestant Missions and Percent evangelized by 1900). This implies that the omitted variable(s) has little correlation with any of the 52 control variables48-which makes it hard to imagine what the omitted variable(s) might be. It also implies that the excluded instruments all have comparable undetected influence on democracy, which is easy to imagine if the excluded instruments have little or no correlation with the error term and hard to imagine otherwise."
"This article demonstrates historically and statistically that conversionary Protestants (CPs) heavily influenced the rise and spread of stable democracy around the world. It argues that CPs were a crucial catalyst initiating the development and spread of religious liberty, mass education, mass printing, newspapers, voluntary organizations, and colonial reforms, thereby creating the conditions that made stable democracy more likely. Statistically, the historic prevalence of Protestant missionaries explains about half the variation in democracy in Africa, Asia, Latin America and Oceania and removes the impact of most variables that dominate current statistical research about democracy. The association between Protestant missions and democracy is consistent in different continents and subsamples, and it is robust to more than 50 controls and to instrumental variable analyses.
Previous quantitative research consistently suggests that countries with more Protestants are more democratic and have more stable democratic transitions (Bollen and Jackman 1985; Hadenius 1992; Treisman 2000; Tusalem 2009). However, this earlier research measured Protestant influence less precisely and more recently than my study (i.e., after CP-initiated behavior spread to other groups and become institutionalized) and thus found results less dramatic than those in this article.
Current statistical evidence of an association between Protestantism and democracy matches historical evidence that Protestantism facilitated the development of modern representative democracy (e.g., Bradley and Van Kley 2001; Clarke 1994; Witte 2007). For example, stable democracy first emerged in Protestant Europe and British-settler colonies, and by World War I every independent, predominantly Protestant country was a stable democracy-with the possible exception of Germany.3 Less stable versions of democracy developed in Catholic areas with large Protestant and Jansenist4 minorities, such as France (Anderson 2004; Philpott 2004; Woodberry and Shah 2004). However, democracy lagged in Catholic and Orthodox parts of Southern and Eastern Europe where Protestants had little influence. A similar pattern existed outside Europe (Woodberry 2004c).
some argue that the association between Protestantism and democracy is spurious, even historically (Moore 1966; Rueschemeyer, Stephens, and Stephens 1992; Swanson 1967). For example, in Europe, pre-Reformation class structures, land-holding patterns, and political conditions may have influenced both the spread of Protestantism and the later development of democracy, thereby creating a deceptive association. Unfortunately, differentiating cultural and instrumental causes is always difficult. In any given context, possible causes are so enmeshed that they are difficult to untangle. For every proposed cultural or religious "cause," scholars can find an alternative economic or political "cause," and vice versa. To escape this swamp of indeterminate causality I use several approaches: (1) observing the consistent association between Protestantism and democracy in regions with histories and class structures radically different from those of Europe; (2) showing historically that CPs had a unique role in spreading mass education, printing, civil society, and other factors that scholars argue fostered democracy; and (3) demonstrating statistically that the historic prevalence of Protestant missionaries strongly predicts democracy in 142 non-European societies using (a) controls for alternative explanations and (b) instrumental variable estimation. These different analyses consistently demonstrate a strong link between CPs and democracy, making it extremely difficult to find a consistent alternative explanation.
As mentioned previously, Protestantism is associated with stable, representative democracy in Western Europe (Context 1), although many argue that this association is spurious (Moore 1966; Rueschemeyer, Stephens, and Stephens 1992; Swanson 1967). Because Protestantism spread to many contexts with different class structures, land-holding patterns, and the like, these varying pre-Protestant conditions can be used to help adjudicate between theories. If the association between Protestantism and democracy remains consistent regardless of context, the claim that the association is caused by these pre-Protestant conditions becomes less plausible. We now turn to four other contexts.
Context 2: Among European-settler colonies, "Protestant-based" United States, Canada, Australia, and New Zealand have been more democratic than "Catholic-based" Argentina, Chile, Uruguay, and Costa Rica. Both sets of countries had similar precolonial conditions (e.g., temperate climates, communal land holding, and small indigenous populations), which weakens theories that climate or pre-Protestant class conditions caused the Protestantism-democracy association. Differences between Protestant- and Catholic- settler colonies after the arrival of white-settlers may be influenced by religion and thus be intervening mechanisms rather than competing explanations. Context 2 also weakens theories that secularization causes democracy (e.g., the United States is far more religious than Uruguay). Still, all predominantly "Protestant" areas were British colonies, and all "Catholic"6 areas were Spanish colonies. Thus colonial institutions may be the crucial factor. Yet whatever the mechanisms are, they seem to be transportable from Europe to other countries.
Contexts 3 and 4: After the fall of communism, Eastern European Catholic and Protestant countries (Poland, Hungary, East Germany, Slovakia, Slovenia, the Czech Republic) had earlier, more stable democratic transitions than did Orthodox Christian and Muslims ones (Albania, Romania, Bulgaria, Serbia, Bosnia). Similarly, Protestant and Catholic former Soviet republics (Latvia, Lithuania, Estonia) had earlier, more stable transitions than did Orthodox and Muslim ones (Anderson 2004; Woodberry 2000). None of these countries were British colonies or had mass immigration from Northwest Europe; this weakens nonreligious explanations for the Protestantism-democracy association in Context 2. Moreover, all the countries in Contexts 3 and 4 had similar pre-transition institutions and entered a similar international environment. All had large secular populations and comparable exposure to Marxist and Enlightenment ideas via monopoly state education. In addition, communists eliminated historic differences in land holdings. Yet in both Contexts 3 and 4, religious differences predict both who mobilized against communism and how smoothly states made the transition to democracy. Catholic and Protestant countries became similarly democratic, but the transitions occurred after the Catholic Church's rapprochement with democracy and in areas where Protestants and Catholics had competed for centuries. "Nonreligious" explanations for the pattern-such as the legacy of Ottoman colonization-might work for Eastern Europe, but not for the former Soviet Union or for Contexts 1, 2, and 5 (see Tables 18 and 19 in the supplemental Online Appendix available at http://www.journals.cambridge.org/psr2012005). Nor is it clear that the Ottoman "influence" was not related
to religion.
Context 5: Finally, if we exclude all European countries and all Protestant European-settler colonies from the sample (i.e., Contexts 1-4) and analyze the remaining countries in Africa, Asia, Latin America, and Oceania, we still find a statistical association between Protestantism and democracy. Thus in at least five distinct contexts the Protestantism-democracy association holds. None of these tests are decisive, and it is possible to think of ad hoc alternative explanations in each context. Yet the consistent association between Protestantism and democracy across all five contexts strengthens the plausibility of causation. It is not clear if any competing theory works in all five contexts or why we should prefer inconsistent explanations over a consistent one.
Contrary to what many theories of "secular modernity" argue, democracy was not a triumph of secularism over religion. From the seventeenth through the mid-twentieth centuries, activist Protestants instituted and spread many of the reforms that made modern representative democracy more likely (see the section, Historical Evidence).
The section, Statistical Evidence, demonstrates that the historic prevalence of Protestant missionaries explains the variation in democracy better than either the prevalence of the nonreligious or of generic Protestants. Moreover, Protestant missions predict democracy, whereas Catholic missions do not. Yet there is no evidence that land-holding patterns (or other theories used to discount the Protestantism-democracy association in Europe and the Americas) shaped the spread of Protestant missionaries but did not shape the spread of Catholic missions (e.g., see Tables 20 and 21 in the supplemental Online Appendix).
Some may argue that the influence of missionaries was too anemic to foster democracy, but before the mid-twentieth century, missionaries were the main source of information about life in the colonies (Fairbank 1985; Hutchison 1987, 1; Tudesco 1980, 56). Moreover, missionaries constituted one of the largest and most educated groups of Westerners in the non-Western world-most had college degrees when few others had them (Daughton 2006; Hutchison 1987). In the Anglo-Protestant world, missionary organizations dwarfed labor unions and other nongovernmental organizations (NGOs).7 In fact, in the nineteenth century in the United States, the largest mission organizations outstripped all but a few commercial banks in size and financial resources (Chabbott 1999, 226-31; Hall 1994; Hutchison 1987). Yet, scholars often argue that labor unions, NGOs, and financial interests influenced democracy.
Nineteenth-century Bible and tract societies were among the largest corporations of any kind (Hall 1994, 34, 44); from 1829 to 1831 the American Bible Society printed and distributed more than a million Bibles at a time when the United States only had about three million households, no railroad system, and a dispersed rural population (Nord 2004, 84).
In fact, arguments for political pluralism, electoral reform, and limitations of state power were originally framed in religious terms (Bradley and Van Kley 2001; Clarke 1994; Ihalainen 1999; Lutz 1988; 1992; Nelson 2010; Witte and Alexander 2008).
For example, Calvinists tried to reconstruct states along "godly" lines and limit sinful human institutions. Perhaps as a result, most Enlightenment democratic theorists came from Calvinist families or had a Calvinist education, even if they were either not theologically orthodox or personally religious (e.g., John Locke, Rousseau, Hugo Grotius, Benjamin Franklin, John Adams, Patrick Henry, James Madison, and Alexander Hamilton),9 and they secularized ideas previously articulated by Calvinist theologians and jurists (Hutson 1998; Lutz 1980; 1988; Nelson 2010; Witte 2007).10 For example, Hobbes' and Locke's social contracts are secular versions of Puritan and Nonconformist covenants, and Locke's ideas about the equality of all people are explicitly religious (Waldron 2002; Woodberry and Shah 2004).
Although stated in secular form, the U.S. Constitution and Bill of Rights derive most directly from earlier colonial covenants, compacts, and bills of rights that were generally justified explicitly in biblical and theological terms; many were written before Hobbes and Locke expounded their ideas. Only 7 of the 27 rights enumerated in the U.S. Bill of Rights can be traced to major English common law documents (Lutz 1980; 1988; 1992; Witte 2007). Even between 1760 and 1805, political writings quoted the Bible more often than either Enlightenment or classical thinkers (34% versus 22% and 9%, respectively; Lutz 1984). Furthermore, the strength of Calvinism and Nonconformism better predicts where democracy emerged than does the strength of Greek and Enlightenment influence. Greek classics were most consistently available in the Eastern Mediterranean and the Muslim world, but democracy did not thrive there; the Roman Empire circled the Mediterranean, and the Renaissance flourished in Southern Europe, but democracy did not thrive in those places either. The "Athenian seed" germinated only after 2,100 years in alien soil: Northwest Europe and North America. Thus, areas with later and weaker exposure to Greek thought would have to have had "stronger pro-democratic effects." At a minimum, some additional catalyst seems likely.
CPs' catalytic effect on printing is clear from the shifts in the printing centers of Europe. Before the Reformation, Italy had the largest printing industry, but Protestantism made little headway there and printing did not increase rapidly or birth either an early public sphere or mass literacy (Graff 1987, 112-19). In contrast, England initially had little printing activity (Graff 1987, 115), but CPs used printing to mobilize ordinary people, forcing their elitist enemies to respond in kind. This activity spurred the growth of newspapers, printed debates, and an early public sphere (Zaret 2000). CPs had similar effects in continental Europe (e.g., Germany, France, Switzerland, and the Netherlands; Eisenstein 1979, 312-452; Febvre and Martin 1976, 287-319; Melton 2001). However, non-state Protestants were weaker there, and extended religious wars destroyed their early gains. Still, from the 1600s on, even in continental Europe, Protestant areas consistently printed more books per capita and exported more printed material per capita than Catholic areas (Eisenstein 1979, 403-23).
When Muslims, Hindus, and Theravada Buddhists engaged in printing, it was usually a response to mass printing by Protestant missionaries or by those trained by them (e.g., Lebanon, Syria, Sudan, Persia/Iran, Malaysia, India, Burma, Sri Lanka, Thailand; Green 2009; Woodberry 2011c).15 CPs printed so many vernacular texts that it forced elite response. For example, within 32 years of importing a press to India in 1800, three British missionaries printed more than 212,000 copies of books in 40 languages and, along with other missionaries, created the fonts and paper that dominated South Asian printing for much of the nineteenth century (Ghosh 2003, 27; Khan 1961; Ross 1999, 40-77, 118). This spurred both Hindus and Muslims to respond, but the earliest Indian printers learned their skills at mission presses, and most early non-Christian Indian imprints were religious-often rebuttals to missionary tracts (Aggarwal 1988, 2; Ghosh 2003; Khan 1961; Robinson 2000, 77; Shaw 2007). In fact, in most Asian societies early indigenous printers gained their skills and equipment from Protestant missionaries (Green 2009; Woodberry 2011c). Conversionary Protestantism also shaped printing's "consequences." If printing was a sufficient cause for mass literacy, newspapers, and the public sphere, then we would expect these developments to have originated in China, Korea, or Japan, but they did not. Printing occurred in China, Korea, and Japan 600-800 years earlier than in Europe. China and Korea had movable-font metal type before Europe; Japan had movable-font metal type starting in 1590; and Korea and Japan had phonetic alphabets, which facilitated literacy and made movable font efficient. All three countries had a high level of economic development and thriving mercantile classes. However, until CPs arrived in the nineteenth century, printing never supplanted handwritten manuscripts, newspapers did not develop, and literacy remained primarily the prerogative of elite men (Davis 1994; Duchesne 2006, 82-83; Reed 2007; Su 1996).
...Protestant missionaries printed the first East Asian newspapers (in Chinese);16 Chinese reformers copied them (Dunch 2001, 78; Zhang 2007), as did Japanese and Koreans, who read mission publications in Chinese and visited missionaries while their countries were still closed to Westerners (Altman 1966, 23-27, 37, 41-42; Inglehart 1959, 40; Lutz 2008, 92-96; K. Shin 1999; Y.
However, missionaries were different. First, many nineteenth- and early-twentieth-century Protestant missionaries came from politically activist traditions. In much of Northwest Europe and in English-settler colonies (excluding slave-holding U.S. states), the Protestant missions movement was closely tied to social reform movements such as abolition and temperance. Thus many missionaries perceived societal reform as a natural extension of their faith (Etherington 2005; Masters and Young 2007; Young 2006). Second, the abuses made mission work more difficult because they angered indigenous people, turning them against Christianity, which many indigenous people associated with the colonizers. Finally, missionaries had the power to fight abuses because they wrote regularly to supporters in colonizing states. During the nineteenth and early twentieth centuries Europeans and North Americans got most of their news about colonized territories from missionary periodicals, which encouraged people to care about distant people they otherwise could have ignored (Miller and Stanczak 2009; Stamatov 2010; Woodberry 2004c; 2006a; 2006b).
In British and American colonies, religious liberty and private mission financing weakened officials' ability to punish missionaries and freed missionaries to critique abuses (Greenlee and Johnston 1999, 34-38), while popular support allowed missionaries to punish colonial officials and settlers. For example, colonial magistrates and governors were reprimanded or removed, military officials were put on trial for murder, confiscated land was returned to indigenous people, and so on (Etherington 2005; Hincks 2009, 181; Oddie 1978; 1996; Stocking 1987, 240-54, 272; Turner 1998; Woodberry 2004c). Thus, Protestant missionaries spurred immediate abolitionism (Stamatov 2006; Woodberry 2004c; 2006a; 2006b), as well as movements to protect indigenous land rights, prevent forced labor, and force the British to apply similar legal standards to whites and nonwhites (Chaudhuri 1998; Clements 1999; Etherington 2005; Gladwin 2007; Grant 2005; Knaplund 1953; Oddie 1978; Turner 1998; Woodberry 2004b; 2004c; 2006b; 2011a). Although others participated in these movements, it was the missionaries who provided detailed information and photographs that documented atrocities. Missionaries also provided emotional connections to distant people and mobilized large groups through church talks and mission presses. Without missionaries, mobilizing mass protests would have been difficult (Grant 2005; Etherington 2005; Stamatov 2010; Woodberry 2004b; 2004c; 2006a; 2006b). The missionary-enabled mobilization made it more difficult for the British to sustain colonial violence or to apply different legal standards to whites and nonwhites. It helped create a cocoon in which nonviolent, indigenous political movements could develop and increased the incentives for colonial officials to allow gradual democratization and decolonization.
R-squareds range from .439 to .849, suggesting that the models predict the distribution of missionaries well. In both tables, the sign and significance levels of coefficients predicting the prevalence of Protestant and Catholic missionaries are almost always consistent, except for variables related to colonizers. Protestant missions had greater prevalence in "Protestant" colonies, and Catholic missionaries had greater prevalence in "Catholic" colonies.
Documentary sources also suggest more missionaries went to places with higher conversion rates, more ethnolinguistic diversity, a greater history of slave trading, and more media attention (e.g., that given to Captain Cook's voyage). However, these factors seem either unrelated to democracy (media attention) or negatively related to it. Conversions were higher among poor, marginalized groups, and poverty, discrimination, ethnolinguistic diversity, and extensive exposure to slave raiding seem unlikely to promote democracy and thus unlikely to increase the association between Protestant missions and democracy.
Both historical and statistical evidence suggest that CPs promoted democracy, although often through indirect means. In all five contexts analyzed - Western Europe, Eastern Europe, the former Soviet Union, European-settler colonies, and mission territories - Protestantism is associated with democracy. Comparative historical analyses show that CPs consistently initiated and spread factors that past research suggests promote democracy: mass printing, mass education, civil society, and colonial rule of law. In cross-national statistical analysis Protestant missions are significantly and robustly associated with higher levels of printing, education, economic development, organizational civil society, protection of private property, and rule of law and with lower levels of corruption (Woodberry 2004a; 2004c; 2006c; 2011b; 2011c; and Table 22, Online Appendix). Moreover, wherever they have been tested, these patterns repeat at the subnational level (Bai and Kung 2011; Gallego and Woodberry 2010; Lankina and Getachew n.d.; Nunn 2010; Woodberry 2004a). Finally, statistical analysis suggests that Protestant missions are strongly and robustly associated with democracy. In fact, missions seem to explain about half the variation in democracy outside Europe and survive dozens of controls and robustness checks.
If omitted variable bias caused the entire association between Protestant missions and democracy, the omitted variable(s) would need to be strongly correlated with both democracy and Protestant missions, but not correlated with Catholic missions (even though Protestant and Catholic missions are highly correlated). More concretely, the cumulative correlation between Protestant missions and democracy is .707.47 If Protestant missions did not cause democracy, then in a properly specified model the correlation would be zero, and the .707 correlation would be the product of the omitted variable(s)'s correlation with democracy and the omitted variable(s)'s correlation with Protestant missions. This requires a mean correlation of .841 with each (.841 ∗ .841 = .707). If true, the omitted variable(s) should be nameable because it would virtually dictate both democracy and Protestant missions. None of the variables used in this article have correlations close to .841 except for the two measures of democracy used in Table 7 (BP and Polity IV), which have a correlation of .844. The three measures related to Protestant missions have correlations with each other of between .332 and .390. Thus, the omitted variable(s) would have to be as correlated with democracy as another measure of democracy is and simultaneously far more correlated with Protestant missions than any of the other measures of Protestant missions are. That seems highly unlikely.
In addition, the relationship between Protestant missions and democracy holds in widely different samples (i.e., sample sizes vary between 26 and 142). The relationship holds if we change regions of the world, if we limit the sample to British colonies, if we drop Muslim societies and Caribbean islands, if we change measures of democracy, or if we do instrumental variable estimation in nine different ways. Moreover, the Protestant mission coefficients are extremely consistent between models, even using different excluded instruments (especially the coefficients for Years of Protestant Missions and Percent evangelized by 1900). This implies that the omitted variable(s) has little correlation with any of the 52 control variables48-which makes it hard to imagine what the omitted variable(s) might be. It also implies that the excluded instruments all have comparable undetected influence on democracy, which is easy to imagine if the excluded instruments have little or no correlation with the error term and hard to imagine otherwise."
