The Orange Order in the 20th Century: A Comparative Analysis of Northern Ireland, Scotland, Newfoundland and Ontario[1]

Dr. Eric P. Kaufmann, Dept. of Politics, University of Southampton,

Southampton, U.K.

The Orange Order, or Loyal Orange Institution (as it is officially known) is a voluntary association that has played a pivotal organisational role for British-Protestant dominant ethnic groups in Canada, Scotland, Northern Ireland, and, to a lesser extent, northwest England. It continues to make headlines in Northern Ireland due to conflicts arising from the July Twelfth marching season. Indeed, the symbolism of the Orange Order is so interwoven into the fabric of Irish history that the Irish Republic's tricolour proclaims its goal of reuniting Orange and Green. In Canada, the Orange Order once occupied a similar position as it does today in Northern Ireland. Today, however, it is but a shadow of its former self. This discrepancy between two modern societies offers us an important variation within one social movement. Accordingly, the Orange Order serves as a lens through which we can focus on processes of dominant ethnicity and voluntary association as they interact with the forces of late modernity.

The Order has never been the subject of primary social scientific investigation, though Ruth Dudley Edwards' The Faithful Tribe (1999) provides the first journalistic account for Northern Ireland and Cecil Houston & William Smyth's The Sash Canada Wore (1980) is an important historical-geographical work. However, the latter book concentrates on the nineteenth century rise of the Orange Order, and only speculates about the reasons why the Canadian Order declined in the twentieth century. Other historical work on the Canadian Orange Order does not address the question of twentieth century social change (Senior 1972; Pennefather 1984). Some high-quality historical research has also taken place in Britain and Ireland, but here again, the post-1939 period remains neglected (Gray 1972; MacFarland 1990; Walker 1992, 1995; MacRaild 1998).

This paper will address twentieth century Orangeism in comparative perspective. This primarily entails consideration of the pattern of Orange membership in Northern Ireland, Ontario, Newfoundland and central Scotland. Indeed, one of the more pragmatic questions posed by the striking demise of Ontario's Orange Order is whether its sharp decline in the twentieth century can yield any insight into the future of the Ulster Orange Order. This would be of signal importance for the long-term direction of the peace process, since the Orange element provides one of the bulwarks of Unionist resistance to the Good Friday Agreement. Our approach, which attempts to explain the impact of modernising processes, also requires an engagement with theories of social change, ethnic change and social capital. Accordingly, this paper promises to broaden our understanding of how dominant ethnic groups are affected by techno-economic and cultural variants of modernisation.

Comparative Loyalism: Canada and Northern Ireland

It is not the intent of this work to account for the birth of Orangeism. However, few would contest the notion that inter-ethnic conflict and rival versions of the past played a role in the origins of this organisation. An interesting exercise in comparative study is to examine the cultural similarities between the (historically) strongest Orange locales: English Canada and Northern Ireland. Both were settler societies whose dominant group members identified themselves not with any particular British ethnie (ie. Irish, Scottish, English), but as composite, ethnic Britons. In both societies, competition with a Catholic ethnic opponent (French-Canadian, Irish-Catholic) helped to reinforce the Protestant accent of the Loyalist group. Likewise, competition with a Republican foe - the Americans for Canada and Catholic Nationalists in Ireland - fortified the Imperial bond. Furthermore, the connection between the Ulster-Protestants and Anglo-Canadians (or 'British Americans') extended to demography: fully 25 percent of Canada's British population was Irish Protestant in 1867, rendering Irish Protestants the most over-represented British category. (Richard 1991: 44, 48, 83; Burnet 1972: 102-4; Buckner 1998: 11)

Both Loyalist groups also faced similar ontological and political problems. Namely, the task of maintaining a credibly 'British' identity in the face of neglect from the mother country and each group's lack of identifiably 'British' culture. Indeed, the English-Canadian struggle to be un-American parallels the Ulster-Protestant struggle against Irishness. Given the similarity between Anglo-Canadians and Ulster-Protestants, it is not surprising that the Orange Order emerged as the leading social movement within both groups from the mid-nineteenth century onward. Finally, the collapse of the British Empire after the 1950's was a blow which both groups have struggled to come to terms with: Anglo-Canadians have variously embraced multiculturalism and anti-establishment populism, while Ulster-Protestants have increasingly attempted to establish a more 'home-grown' sense of ethnic identity. (Dunn & Morgan 1994; Adamson [1982] 1991; Kaufmann 1997: 130; Craith 2001)

The Rise of the Orange Order in Canada

The Orange Order developed as a fraternal society for the Protestant population in Ireland after 1795. It rose in response to the impending threat of Irish independence posed by the revolutionary coalition between the liberal-Protestant United Irishmen and the Catholic Defender movement. In the following three decades, Irish immigration and its attendant sectarian divisions established Orangeism in the western lowlands of Scotland and the north-west of England. Thus Orangeism existed within a Unionist ethnic and political environment in Northern Ireland, Scotland (particularly in the western part of the central belt) and areas adjacent to Liverpool in north-west England. (Gallagher 1987; MacFarland 1990; McCrone 1992; MacRaild 1998) Nonetheless, since the organisation served instrumental and fraternal functions, one should not assume that the spread of the movement merely reflected the aforementioned cultural-political conditions.

The processes which spawned the rise of the Orange Order in the British Isles operated with equal, or greater, vigour, in Canada. Few contemporary English-speaking Canadians realise that as many as one in three adult Protestant males in Ontario passed through the ranks of the Orange Order between 1870 and 1920 while the influential Ladies' Orange Benevolent Association (LOBA) proved that the principles of the Order were not gender-specific. The Canadian Order was not an Irish organisation, but instead brought together several ethnic components of English Canada's Protestant majority. (Houston & Smyth 1980: 84, 95-6, 104) Founded in Ontario in the early years of the nineteenth century as an association for Irish Protestant immigrants, by the 1860s, the Orange Order had become firmly 'native' in outlook.

Its power was centred in Ontario and New Brunswick, but the Order maintained a strong network of lodges in all provinces[2]. The large-scale immigration of Irish Protestants in 1820-65 gave the organisation its initial impetus in Ontario and New Brunswick. In Newfoundland, however, the Order took root amongst a native-born population of West Country English derivation. First introduced by Prince Edward Islanders and Nova Scotians travelling by ship to the western Port-au-Port peninsula in the early 1860's, Orangeism quickly began to thrive within the active Protestant-Catholic matrix of this British maritime colony. (Senior 1960)

Later, in the nineteenth and early twentieth centuries, both the Canadian and Newfoundland Orders benefited from an imperialist political climate. This was manifested through a Britannic nationalist exuberance for both Empire and the 'civilizing' mission of the Protestant crusade. Britannic nationalism was bolstered by both a romantic Loyalist Revival in the post-Confederation period as well as a series of conflicts in which Orangemen featured prominently as bastions of British Loyalty. These included the Rebellions of 1837-8, the Crimean War (1854-6), the Riel Rebellions (1869-70, 1885), the Manitoba Schools Question (1890), the Boer War (1899-1902) and the First World War (1914-18). (Rasporich 1968: 140-56; Senior 1972: 62, 71, 96; Berger 1969)

Canadian imperialist fervour far exceeded that of the British metropole - where imperialism divided liberals (allied to the cause of Irish home rule) and Tories. For instance, the Royal Tour of Canada in 1901 drew crowds that regularly exceeded the local population. 'Everywhere the crowds were huge and enthusiastic,' writes Philip Buckner. 'In Toronto, between 200,000 and 250,000 people lined the streets.' (Buckner 1998: 12) Meanwhile, Orange expansion in the nineteenth century was so great that the Order had become larger in Canada than in Ireland by 1900. (See figure 1) When we consider that the population of English-speaking Canada was just 3.8 million at this time, we must conclude that English Canada was as Orange a society as Ireland. (Buckner 1998: 14) Furthermore, the concentration of membership in Ontario, New Brunswick and Newfoundland compared favourably, as we shall see, with Orange density in the nine counties of Ulster.

The Order's mainstream nature and political influence as a bastion of popular Toryism is attested to by the many politicians who passed through its ranks, from Sir John A. Macdonald and Oliver Mowat in the mid-nineteenth century to prime minister John Diefenbaker, Newfoundland premier Joey Smallwood and Toronto mayor Leslie Saunders in the mid-twentieth. (Senior 1972) In Ontario, the Order was influential both at Queen's Park and on Toronto city council, enforcing sabbatarian ordinances and temperance laws that influenced the character of 'Toronto the Good' until the Second World War. Indeed, the City of Toronto ('the Belfast of Canada') only acquired its first non-Orange councillor in the 1930s. In Newfoundland, meanwhile, many credit the Order with tipping the 1948 vote in favour of Confederation. (Fitzgerald 1998: 3; Jamieson 1991: 103)

The Decline of the Orange Order in Canada

By the third decade of the new century, Order membership growth in Toronto was failing to keep pace with the city's skyrocketing population. (Houston & Smyth 1980: 154-7, 162-80) And from the 1950s, the organisation began to lose political influence, a change symbolised by the fact that John Diefenbaker proved to be Canada's last Orange prime minister. Its male membership in 1984 stood at just over 14,000 in 616 lodges, a significant drop from the more than 58,000 members and 4000 lodges which made up Canadian Orangeism in 1955. (GOLOWret 1985; ICGW 1955) Thereafter, particularly in its Ontario heartland, a slow but steady decline set in. Today, the organisation is dwindling and is viewed as an interesting survival from another age. (Houston & Smyth 1980: 162-3)

By contrast, owing to stable or rising membership during 1920-65, the Order's presence in Northern Ireland neatly parallels its former influence in Ontario. Thus the Order maintains an influential presence in both civic and provincial politics, with many Belfast city councillors (including mayor Stoker), and nearly all Ulster Unionist Party MPs counting themselves as members. In addition, its 200-year history, 45,000-strong Ulster membership, and its position on the Ulster Unionist Council ensure that the Orange Institution is a significant political and social player in Northern Ireland. This is highlighted annually during its more than 4,000 July Twelfth parades - including the highly controversial Garvaghy Road route at Portadown. Such a profile provides a definite contrast with Ontario, where the Order's July Twelfth parades arouse little excitement, while few of those under fifty are familiar with the organisation.

Therefore, the Order's elderly Canadian alumni hold the key to a puzzle of

(post?) modernisation: what caused the decline of the Orange Order, a structural backbone of Anglo-Canadian dominant ethnicity? Secularisation would appear to be a promising explanation, and it is among the factors listed by Cecil Houston and William Smyth in their speculations about the reasons for the Order's demise. (Houston & Smyth 1980: 170-71) Yet this can serve at best as a partial answer, since Orange decline preceded the decline in Canadian Protestant religiosity by some twenty to forty years. Value change, in a liberal-egalitarian direction, offers a competing explanation behind the decline, as does a more general decline in the prestige of British Loyalism. To some extent, this has been borne out by recent survey research concerning the decline of Loyalism in Canada. (Schwartz 1967: 74-6, 106-123; Cheal 1980)

On the other hand, techno-economic rationality, immigration (or some other variable) might turn out to be critical - and all are offered as competing hypotheses by Houston and Smyth in their concluding chapter. In many ways, therefore, the puzzle of Orange decline in Canada touches upon the very questions which lie at the heart of debates in empirical political and social theory.

Theoretical Context

Social and Political Theories

This research intersects with three major theoretical discourses: social/empirical political theory, social capital/social movements theory, and ethnicity/nationalism theory. The analysis attempts to explain why a 150-year old organisation at the centre of one society can enter into sudden decline, while remaining stable in other places. It also tries to determine, more generally, which forces drive the growth and decline of a large-scale social movement over place and time. The Orange Order has been much more stable over the past two centuries than more radical organisations like the Scottish Protestant League, American Protective Association or (Canadian) Equal Rights Association.

Furthermore, the Orange Order's existence as a traditional fraternity exactly parallels that of the modern period, thereby providing an interesting window into the interplay of tradition and modernisation. A stable organisation like the Orange Order likewise provides a good yardstick for assessing the transition from industrial modernity to post-industrial or 'high' modernity (Giddens 1991). If Orange membership and political influence rose with modernisation but declined with the advent of post-industrial modernity, this suggests that there may be something qualitatively different about this 'post-modern' phase.

Another implication of this project for social theory concerns the cause of fluctuations in Orange membership and power. Here, one axis of debate centres around whether the engine of social change is the techno-economic 'base' of society (the position held by orthodox Marxists and many rational choice theorists) or its cultural-symbolic 'superstructure,' the explanation favoured by many in the Durkheim-Weber tradition (Inglehart 1990). For these theorists, changes in norms and worldview are the leading driver of change. To these 'economistic' and 'culturalist' theories, one must add the more recent empirical political theories which ascribe independent causal power to institutional processes (i.e. March & Olsen 1984). Finally, empirical political theorists also highlight the causal impact of events like the First World War or Watergate scandal in recasting established worldviews or reframing economic and political structures.[3]

Social Capital and Social Movements Theory

Debates in social theory have their correlates in social movements theory, particularly in the work of Robert Putnam. Putnam's recent research, using American data, attempts to explain the decline in social capital, or voluntary association, in the United States. Fraternal societies (like the Orange Order or Freemasons), Putnam notes, are amongst the hardest hit by recent developments. Putnam traces a membership profile for thirty-two American chapter-based associations which charts a trend of steadily rising membership through 1900-1957, a plateau during 1958-68, and a period of steady decline post-1969. (Putnam 2000: 54, 438-45)

In the first half of the century, only the depression interrupted a steady growth in voluntary organisations, as Americans became ever more involved in associational life. However, Putnam posits that half the general post-sixties decline in organisational vitality is attributable to the replacement of a 'civic' generation of joiners born before 1940 with more privatised cohorts born after that date. Thus a generation reared during a time of depression, war and optimism proved more willing to join associations than its post-modern progeny. Television, changing gender roles and suburban sprawl - with their attendant privatising effects - contribute a further 25-40 percent towards Putnam's explanation of decline. (Putnam 2000: 284)

Other Departures in the Theory of Fraternal Organisation

A number of challenges to Putnam's thesis have emerged in the literature since he published his ground-breaking article in 1995. One of the more important with respect to fraternalism is David Beito's recent analysis of life insurance fraternities in the U.S. Beito's exhaustive analysis of the social welfare provisions of numerous fraternal orders suggests that decline set in far earlier than the 1960's. For life insurance fraternal orders - a category which encompasses most of the larger Orders - such as the Moose, Eagles and Masons, the sustained trend of post-Civil War membership growth began to slow by the 1920's. Beito suggests that rival entertainment sources such as radio, the automobile and the movies were one source of the slowdown. Changing cultural values which derided core fraternal idioms like thrift, mutual aid and character-building, were also responsible. Finally, predatory state legislation, which favoured state and commercial insurers, drove fraternities from the social welfare field. (Beito 2000: 204-5)

Meanwhile, immigrant-dependent associations lost membership as the restrictive 1924 National Origins Quota Act took effect. According to John Higham, the new shared American working-class culture of the thirties acted as an assimilating solvent. 'All the institutions of ethnic culture weakened,' Higham remarks. 'Lodges declined, ethnic ceremonies and theatres faded…and in their churches and newspapers a younger generation of priests and editors began to encourage a greater use of English.' (Higham 1999: 53, emphasis added)

Whereas Putnam views the sixties as the plateau before the decline, Beito considers the twenties as the membership hiatus which heralded a subsequent avalanche of decline. As evidence, Beito notes that the ranks of the six largest orders stressing sick and funeral benefits fell from 7.2 million in 1930 to 5.9 million in 1935 and - despite the end of the depression - continued to tumble to 4.7 million in 1940. Meanwhile, life insurance orders saw their total membership slide from 10.1 to just 7.8 million Americans during the same period. (Beito 2000: 222-23) Even the mighty Masons peaked at their staggering 12.1 percent of native-born white American adults in 1930. (Dumenil 1984: 225)

The case of the largest association in Canada, the Independent Order of Oddfellows (IOOF), is instructive: J.C. Herbert Emery's analysis of lodge registers in British Columbia and Ontario shows that the IOOF grew rapidly until the 1920's, but during this decade, initiations slumped from 11 to 3 percent. In contrast to Beito, Emery discounts fraternal insolvency, government legislation and both union and commercial insurance as contributing factors. Instead, rising income among younger men, Emery suggests, allowed them to insure themselves without having to join a fraternity. Together with recreational competition, this spelled the beginning of the end for large fraternities like the Masons, IOOF or the Orange Order. The trend is particularly noticeable in the cities, where insurance and recreational competition was keenest and IOOF membership rates lowest. (Emery 1999: 40, 44)

Though differing widely in their interpretations of decline, most analysts of fraternal organisation point to a combination of cultural factors (i.e. changing attitudes or recreational tastes, generational change), economic influences (i.e. rising income, new technologies like the car or television, new recreational outlets), institutional forces (i.e. growth of welfare state, fraternal scandal) and events (i.e. world wars). However, as with Houston and Smyth in the case of Canadian Orangeism, few analysts have made a concerted attempt to weigh the relative importance of various factors.

Ethnicity and Nationalism Theory

Some of the most important work on fraternalism stresses the importance of changing conceptions of gender, class and race in accounting for the rise and fall of fraternal lodges. (Carnes 1989; Clawson 1989) Though admittedly a culturalist statement, it is nevertheless a truism that one of the hallmarks of ritualistic fraternalism in the first half of the twentieth century has been its ethnic homogeneity. As agents of 'bonding' social capital, fraternal organisations - whether immigrant/ethnic, black or WASP maintained a remarkably homogeneous membership base.[4] In the United States, as in Canada, IOOF members were largely - if not exclusively - drawn from the declining Anglo-Protestant majority. (Dumenil 1984: 115-47; Emery 1999: 31)

Meanwhile both the IOOF and Freemasons endorsed the defensive ethno-nationalism which gripped WASP America between 1890 and 1925. The rise and fall of the largely northern-based, multi-million member 'second' Ku Klux Klan between 1915 and 1928 - a more moderate and mainstream organisation than either its predecessor or successor - should give us pause when dismissing shorter-term explanations of fraternal change. The same must be said for other 'nativist' ethno-national movements like the Immigration Restriction League, million-plus member American Protective Association or the Native American Party - which won several state elections and threatened to win at the federal level in the 1850's. (Kaufmann 2002; Higham 1955 [1986])

The Orange Order, too, was at the forefront of immigration restriction and defensive nationalism, this time on the Canadian prairies in the 1920s, a stance which continued unabated into the 1970s.[5] (Houston & Smyth 1980) Indeed, it is unsurprising that Protestant fraternities should manifest dominant ethnic tendencies given that these sentiments were all but universal among the Protestant middle and lower strata. Moreover, at every level, Protestants maintained higher rates of associational activity than Catholics, so fraternalism provided an obvious outlet for defensive middle and working-class ethnic sentiment. (Anderson 1970: 118-19)

At a more middle-range level of analysis, ethnicity/nationalism theorists are concerned with the question of whether ethnicity, dominant or otherwise, tends to decline in high modernity. This scholarly discourse cuts to the question of whether ethnicity is a transitory phenomenon, thrown up by improved communication systems between smaller clan units, that will disappear with the extension of those communication systems beyond the confines of the ethnie. 'Evolutionists contend that ethnic groups will be superseded in the future, while 'revivalists' see ethnic groups as more persistent.

One school of thought, modernisation theory, holds that ethnic decline forms part of the general trend away from kinship-based forms of 'traditional' social organisation. (Durkheim 1893; Parsons 1951; Deutsch 1953; Kerr 1960; Hechter 1975: 22-9). To some degree, Steve Bruce, a leading authority on comparative Protestant politics, subscribes to this viewpoint, claiming that Protestant hegemony in Britain and North America has declined due to the combined impact of secularisation, Protestant schism and structural differentiation (Bruce 1998). Might these forces have led to Orange decline in the latter half of the twentieth century? Our analysis will shed important light on this question.

Since the 1970s, modernisation theory has been eclipsed as the main paradigm in ethnicity and nationalism theory by more conflict-based approaches (Smith 1981; Horowitz 1985). These advance the notion that ethnic groups either resist decline or increase their social importance in post-industrial modernity. This is prompted by improved methods of communication, which facilitate greater group consciousness. This occurs because ethnic media and education networks become both more intensive and more extensive, all within an atmosphere of increased awareness of other groups. This paradigm would construe Orange decline as either temporary, or as an aspect of the 'translation' of Protestant tradition from the Orange Order to new social constellations like sports teams or Unionist marching bands (Bell 1990; Bairner 1997). Here again, the analysis will try to assess the relative merits of these two main approaches.

Research Methodology

The quantitative portion of this research is based on an examination of Orange Order membership figures for all years available since 1900, broken down geographically (nationally, and by province, county, and district). The quantitative analysis is complemented by a qualitative dimension, including interviews with Orangemen and women, lodge reports, lodge newsletters and city newspapers. The goal being to grasp some of the more complex influences on Orange membership as well as to provide a subjective perspective on the dynamics of Orange political power.

It is important to recognise that Orange Order data has generally not been available to academic researchers and that the data presented here have never previously been published. Due to privacy agreements with several Grand Lodges, certain membership figures will be omitted from the analysis. It should also be stressed that this research is still at an early stage, with a significant body of both Orange and census data still awaiting collection, digitisation, and - where boundaries change or are incongruent - GIS interpolation. Consequently, missing data restricts the scope of our present model. Even so, current data has already furnished a wealth of important material, some of which will be presented here.

Current Findings

Since I have already published material from my qualitative research[6], I will concentrate on quantitative trends in Orange membership in relation to variables from the census, historical events and, in the case of Northern Ireland, political fatalities.

International Overview

Figure two presents a time-series portrait of membership trends in Newfoundland, Ontario, Scotland and the three Northern Ireland counties for which we have data. The first point to notice is that membership density, i.e. the proportion of the Orange target population (male British Protestant population over the age of eighteen) that is in membership, fluctuates greatly over both time and place. Secondly, it is clear that membership density in Scotland is relatively weak, despite popular perceptions. This is not merely an artefact of Scottish Orangeism's concentration on the West Coast since even in the most heavily Orange zone comprising Glasgow, Lanarkshire, Ayrshire and Renfrewshire, membership densities for this period do not exceed two percent. Only in a few selected areas, like Glasgow's Govan ward or Rutherglen district, to the southeast of the city, do densities approach those of Ontario.

The upshot is that Canada was a far more Orange place than anywhere else outside of Northern Ireland. Moreover, while Ontario's membership density was similar to Belfast's until World War I, Newfoundland's far outstripped that of the three eastern Ulster counties for which we have good time-series data. In fact, only Co. Fermanagh and parts of Co. Tyrone and Co. Armagh can approach Newfoundland's peak 1920's density of more than one-third of adult male Protestants in membership. (See fig. 2)

Source: Grand Lodge and County returns and reports.

A second feature of Orange membership trends is the pattern of twentieth century decline. In Canada, this dates to the 1920's. However, in Northern Ireland and Scotland, absolute declines set in much later. Northern Ireland membership only began to trend downward in the mid-sixties, while Scotland did not see a downturn until the mid-eighties. Furthermore, local variation is significant: Belfast resisted decline until the early seventies, then dropped sharply while Antrim and Down crested in the early fifties, and slipped gradually downward thereafter. In the rest of Northern Ireland - particularly mid-Ulster - decline was, and is, far less noticeable. At the macro level, decline occurred first and most severely in Ontario, followed by Newfoundland, Northern Ireland and Scotland.

Source: Grand Lodge and County returns and reports. Note that these figures are raw membership figures, unlike those in figure 2, which are adjusted for male, adult, British-Protestant population.

Explaining Membership Fluctuation

'Native born Canadians,' declared Canadian Grand Master Gordon Keyes, 'and indeed new Canadians from the British Isles…find difficulty in devoting time to lodge affairs. Perhaps the ever increasing attitude of materialism, or selfish attainment contributes to neglect of our heritage….' (Keyes IGCW1964: 30) Commentary such as this, along with interviews I have conducted with both elite and rank-and-file members, helps to define the parameters of the problem.

Indeed, combining feedback from respondents, Orange documents and secondary historical literature, the list of possibilities for explaining recent decline includes the following: 1) cultural: decline of Anglo-Protestant ethnic hegemony, rise of non-British immigration, decline of religiosity, ecumenism/decline of sectarianism, changing gender roles, rise of Canadian/Scottish nationalism, declining support for Monarchy/Britishness, and rise of liberal-egalitarianism; 2) social: decline of social capital/connectedness, rise of alternative forms of recreation and conviviality, declining appeal of ritualistic activities; 3) economic: rise in shift work, suburbanisation and geographic mobility, increase in income, spread of television/radio/automobile/telephone, depression, unemployment, industrial decline, slum clearance; 4) institutional: increases in lodge dues, organisational apathy, organisational schism, growth of government functions; 5) event-driven: World Wars, Depression, political violence, diplomacy and constitutionalism (i.e. Good Friday Agreement, Royal Bilingualism and Biculturalism Commission), and cross-community political events (such as new Northern Ireland Police Force or papal visit to John Knox's statue) .

The richness of these theories is evident: they could not have been generated by quantitative techniques. However, in order to winnow down the large number of possible drivers of Orangeism's rise and decline in the twentieth century, I believe that it is imperative that we next to turn to statistical strategies. We cannot address all of the above possibilities just yet, but the analyses to follow will go some of the way toward underlining the most promising hypotheses.

Ontario

Our first port of call on the route to understanding twentieth century Orangeism's trajectory is Ontario. There is a fairly good set of annual data for many socio-economic indicators in the province, especially for the post-1945 period. Moreover, some indicators reach back into the early decades of the century. Figure 1 shows the results of multiple regression analysis across all 52 available years for 1910-1961. Here adult male Orange density (y) is regressed on the following available data: 1) Proportion of the provincial population that is British-origin and Protestant [gbprot] - this provides an indirect measure of the Catholic population as well; 2) Passenger auto permits per capita [auto]; 3) Proportion of eligible pupils attending K-12 education [propsch]; 4) Proportion of eligible pupils attending K-12 education ten years ago [propschlag10]; 5) World War I [WWI]; 6) Great Depression [Depr]; 7) World War II [WWII]. Variables for lags of 1, 5, 10 and 20 years for war and depression have been included as well. Finally, I have created a variable for the interaction between war and education [impact of wars*proportion in school], along with associated 10 year lags.

SPSS stepwise model total and partial correlations (R-squared) are reproduced below from most important variable to least (all variables are significant in this paper unless otherwise noted):

Independent Variable		Partial Correlation	Standard Error	Significance (Sig)
Propschlag10	Proportion of eligible pupils attending K-12 education ten years ago	-.979	.000	.000
Auto	Passenger auto permits per capita	-.889	.008	.000
WWI	World War I	.449	.001	.000
WWIlag1	World War I - one year aftermath	.430	.000	.010
Deprlag10	Depression - ten year aftermath	-.289	.001	.032
WWIIlag10	World War II - ten year aftermath	.148	.000	.000
Model Total:		.994	.00183

This model for 1911-61 predicts Ontario Orange density almost perfectly (.994), and it is clearly the case that education levels of ten years ago play the most important role in this predictive power, with passenger auto permits running a close second. Both of these modernisation variables have a strongly negative effect on Orange density. Conversely, the World Wars tend to have a stimulating effect - whether this takes place immediately, as in the case of World War I (+.449), or some years thereafter (WWIlag1, WWIIlag10). Finally, the Great Depression is shown to have a weak effect on membership, and this only after ten years.

Removing post-1945 trends from the analysis (N=36) reduces the impact of the ten-year education lag to -.735 and elevates the effect of WWI lag 1 to .836 and depression lag 10 to .567. The British-Protestant proportion of the population takes on importance at .706 while World War I itself correlates with Orange density at .646. Post-1940 analysis (N=23) yields an increase in the proportion of the variation explained by education lag10 to -.812, and the British-Protestant proportion of the population takes on a very high partial correlation of .980. This is interesting, for it demonstrates that, if anything, the Catholic population in Ontario was rising at a time when Orange fervour was in decline - rather than the (expected) reverse relationship. Later, we shall see that this conclusion needs to be qualified by region.

The Role of Structural Forces

Rather than take these results at face value, it is important that we consider the role played by socio-economic data collected only after 1928. Unfortunately, this will reduce our number of cases from 52 to 32 and therefore exclude years up to 1928 when Orange densities rose and then began to fall. The impact of World War I is thereby largely occluded. However, this exercise will increase the number of independent variables in our model considerably. SPSS model total and partial correlations (R-squared) are reproduced below from most important variable to least (all variables significant):

Independent Variable		Partial Correlation	Standard Error	Significance (Sig)
Propschlag10	Proportion of eligible pupils attending K-12 education ten years ago	-.980	.000	.000
GBProt	Proportion of the population of British ethnicity and Protestant faith	.949	.006	.000
Depr	The Great Depression, 1929-33	.679	.000	.000
WW2lag10	World War II - ten year aftermath	.646	.000	.000
War_Inc	Effects of War, interacting with Income per capita	-.537	.442	.006
WW1lag10	World War I - Ten year aftermath	-.514	.000	.009
Warlag10_ed lag10	Interaction of 10-year lags of wars and proportion in K-12 education	.484	.000	.014
War_Ed	Interaction of wars and proportion in K-12 education	.396	.000	.05
Model Total:		.998		.00047

Once again, education of ten years ago (-.980) proves the most important factor in a model that now includes additional variables for per capita social security payments, old age pensions, mothers' allowances, personal income, retail sales, chain store sales, labour wages and salaries, manufacturing wages and an industrial composite employment index.[7] The most important point to notice here is the striking absence of economic variables in our explanation. Only the interaction of war and income plays a role, and a rather meek (-.537) one at that.

In the previous analysis of the 1910-61 period, we noted that as we approached the more recent period, the proportion of British-Protestants begins to attain significance. Thus it is that we find that for the post-1928 period, this factor becomes very (.949) important. The positive direction of the relationship suggests that the decline in WASP demographic preponderance may have signalled a more general decline of dominant-group ethnicity. This might account for the lack of Orange response to the growth of the non-British, non-Catholic population in this period. It may also suggest that more liberal attitudes among British-Protestant Ontarians helped to lower barriers to Catholic immigration. Evidently, more research is needed to substantiate this hypothesis.

In terms of events, the Great Depression is most important, and it coincides with an improvement of Orange density, at a partial correlation of .679. This is difficult to interpret, but it is probably the case that lodges' practice of making allowance for unemployed members during the depression temporarily reduced the rate of suspensions for non-payment of dues and thus masked the trend of sharp membership decline which characterized the 1920-1939 period.

World War I's ten-year lag effects can be discounted for this post-1928 period due to limited data points, despite the fact that its ten year lag reaches the first few years of this model at a partial correlation of -.514. World War II also shows little effect, but its ten-year aftermath correlates well (.646) with a revival of Orange membership - something noticeable in other provinces as well. Perhaps this phenomenon is related to Robert Putnam's observation that associational life in general picked up strongly in the decade after the second world war. Despite the low number of cases (N=32), it seems that education and dominant-group ethnicity are our best predictor of Orange density while economic and event-driven variables contribute little to our explanation.

The median individual in K-12 education ten years ago is of prime age (21-22) for lodge initiation, and it may be the case that the education system began to act as a socialising agent for values that were antithetical to Orangeism as early as the turn of the century. Given the officially pro-British imperative in Canada around this time, it is doubtful that Canadian nationalism or anti-monarchical sentiment played a role. (Schwartz 1967) Instead, the most likely focus of counter-Orangeism would be religious toleration for Catholicism and, less convincingly, ethnic tolerance of French Canada. Indeed, historical research in the United States and Britain has demonstrated that Protestant sectarianism fell out of favour among the educated middle class and cultural elites by the turn of the century and had begun to pass from school textbooks around this time as well. It would be surprising if English-speaking Canada did not participate in this trend. (Fitzgerald 1979; Ingram 1988; Waller 1988; Bruce 1998: 119)

The high partial correlation for British-Protestant ethnicity and Orange density can be slotted into this explanation, since an ethnically assertive British-Protestant population would be expected to demonstrate strong antipathy to non-British, Catholic population growth. Here we would expect a negative, rather than positive, correlation between the British-Protestant share of the population and Orange density. Yet our findings suggest the opposite: as the Catholic population increased through French-Canadian migration, European immigration and a higher Catholic birthrate, Orangeism actually weakened among British Protestants.

It is certainly possible that the growing Catholic presence may have forced Protestants to be more accommodating. Indeed, this was somewhat evident in the United States, where the second Ku Klux Klan flourished in Protestant-majority sections of Protestant states like Oregon and Indiana, but fared more poorly in mixed regions of more diverse states like Illinois. (Moore 1991) On the other hand, this is a counterintuitive result that our cross-sectional and comparative analyses will cast doubt upon. More than likely, the high correlation has partly been produced because a third factor related to modernisation (be this liberalism or economic growth) prompted a decline in both Orange density and the British-Protestant proportion of the population.

Ontario: Cross-Sectional Analysis

The preceding analysis is consistent with a cultural explanation, but might a different story be told if we examine data cross-sectionally as well as in time series? A limited dataset of census information has been digitised at county level for 1911, 21 and 41. This includes data for counties in Ontario West for 1911, 21, and 41, but - owing to missing Orange records, excludes the third of the province under the Grand Lodge of Ontario East for the years 1921 and 41.[8] 88 cases were generated, though the number of independent variables is a great deal more restrictive than in the time series analysis. Variables include 1) Proportion of the provincial population that is of British origin, 2) Proportion of the population that belongs to a British denomination of Protestantism (i.e. may include German-origin Anglicans but not Scottish-origin Lutherans), 3) Percentage rural, 4) Irish Protestants as a proportion of British-Protestant population, 5) Scottish Protestants as a proportion of British-Protestant population. Variables were also added for major non-British ethnic groups as well as Catholics, Lutherans and Mennonites.

Evidently there is only one variable (urban-rural) that taps an economic factor, so this is as yet an incomplete exercise. Results of our time-series cross-sectional analysis appear below:

Independent Variable		Partial Correlation	Standard Error	Significance (Sig)
ProtIre	Proportion of British-Protestant Population that is of Irish origin	.355	.001	.018
Rural	Proportion rural	.334	.002	.010
Dutch	Proportion Dutch origin	-.294	.007	.000
Lutheran	Proportion Lutheran	-.241	.027	.000
Baptist	Proportion Baptist	-.239	.029	.000
Model Total		.555	.01073	.000

This is a weaker model than we saw in the purely time-series analysis, though this can partly be explained by the large number of cases contained therein, as well as the lower degree of linearity in cross-sectional models. Notice that at the cross-sectional level, there is no variable that predicts the trajectory of Orange density as powerfully as education (lag10) or British Protestant population does in the time series case. Nevertheless, it is interesting that across our 88 cases, the best predictor is Irish Protestant ethnicity. Conversely, Scottish Protestants as a proportion of the British population are not a significant predictor of Orange density. The overall importance of Irish Protestant ethnicity as late as 1941 suggests that there remained an element of 'immigrant association' within the Orange Order despite the fact that the organisation clearly transcended its Irish-Protestant roots. (Houston & Smyth 1980; Senior 1972) Future work with nominal census data and lodge roll books should expose the degree to which the Irish were over-represented within Ontario Orangeism.

The proportion of rural population was next in importance to Irish Protestantism, though it was not used as the initial predictor by the model. Clearly, more variables need to be added in order to determine the importance of this factor in relation to other aspects of structural modernisation. Comparison across 39 cases in 1911, where we have data for the entire province and include variables for population density and population growth, leads to the removal of the rural variable entirely. In fact, besides Irish Protestantism, only Dutch has a partial correlation of any significance, though the model deemed it too statistically suspect to include. Perhaps it is the case that the rural-urban variable takes on greater importance after 1911 or in Ontario West, but it is also possible that its effects may be displaced or dissipated by the addition of further structural variables.

Turning to the less powerful factors, it is a puzzle as to why counties with higher than average Dutch or Lutheran populations might have lower Orange densities within their British-Protestant populations. Perhaps these Protestant minorities may exert a salutary effect on cross-community relations, thereby ameliorating the stark Protestant-Catholic paradigm. It is also possible that Baptist, Dutch and Lutheran populations gain influence over the model due to the skew in the data towards western Ontario. This is particularly true for Baptists since they are over-represented in south-western counties around London where 'late Loyalists' and other immigrants from the United States settled early in the nineteenth century. (Errington 1987) The fact that all three variables disappear from the analysis of 1911 (N=39), goes some way toward endorsing this view.

The above analysis further indicates that the proportion of French and Catholic population in a district does not seem to be as significant as one might have expected. However, when one examines figures for the seventeen counties of Ontario East in 1911 (recall that this is the only year for which we have Ontario East Orange data), the model's only important variable is the proportion of the population of French origin, which correlates at .609 with Orange density. Moreover, narrowing the focus to the 11 easternmost counties in the province raises the model's predictive power to .571, and, most importantly, brings the proportion of Roman Catholics to the fore as the main predictor of Orange density. The correlation is significant at .759.

It is well-known that French-Canadian migration into eastern Ontario was an important precipitator of low-level ethnic conflict in the mid-nineteenth century, and it may well be that this legacy persists most strongly in the easternmost counties of southern Ontario. Yet these are also counties which contain higher than average proportions of Catholics. Could it be that Protestants in counties with Catholic populations above a certain threshold tend to be more likely to respond in an Orange fashion to higher Catholic populations? An analysis of the twenty cases where Catholics form at least 20 percent of the population (irrespective of region) indeed shows this to be the case. For these counties, Catholic percentage is the best predictor of Orange density at a partial correlation of .746, followed by the proportion of British ancestry at .592.

Ontario Summary

First of all, it is important to note that we have yet to build up a comprehensive model that contains cultural, structural and event-driven variables in both time-series and cross-section. Nonetheless, from our evidence to date, it appears that the Irish share of the British Protestant population is our best cross-sectional predictor. It has a consistently high correlation and appears in every large model that was run. As we approach counties with Catholic populations above twenty percent, it appears that the proportion of Catholics supersedes Irish-Protestant ethnicity as the main predictor of Orange density.

Finally, the proportion of British origin appears, if anything, to be linked to higher Orange densities. An interpretation of these patterns might be that Irish Protestant ethnic tradition tends to maintain Orange density in strongly Irish and Protestant western counties like Dufferin, Bruce and Simcoe. In those northern and Quebec border counties of the province where Catholics form a higher proportion of the population, the Protestant imperative supplants Irish-Protestant ethnic tradition as the motor of Orangeism. We also cannot rule out the role of structural modernisation, since the proportion of rural population is important in our model. However, the addition of just a small number of extra structural-demographic variables in 1911 seems to dissipate the impact of this factor, so it is not clear how effective it really is.

Returning to our time-series analysis, it becomes evident that education levels comprise the most important part of the story when it comes to dissecting temporal patterns of Orange membership. They seem to exceed structural variables and events in importance at nearly every turn. World War I and its effects were of moderate importance - stimulating Orangeism during and immediately after the war, but abetting decline ten years later. The rising proportion of Roman Catholic and non-British population in Ontario does not appear to have led to heightened Orange activity among Protestants. On the contrary, it seems to have dampened Orange density. On the other hand, this certainly is not the case at the cross-sectional level. Later on, with the benefit of a comparative perspective, we shall address this seeming anomaly between time series and cross-sectional findings.

Newfoundland

Though Newfoundland's Orange membership was less than a third of Ontario's at its peak, this colony/province had a per capita rate of lodge membership that was 3 to 6 times higher than Ontario. This makes it an important province for consideration. Unfortunately, Newfoundland's late entry into Canadian confederation means that comparable annual statistics do not appear until after 1949. Only data on education, ethnicity, religion and population exists in annualised form for the period 1920-49. Furthermore, we have yet to collect Orange membership data for the post-1965 period. Finally, pooled time-series cross-sectional analysis has been delayed due to the need to correct for changing census boundaries between 1911, 1935 and 1949. Using a limited time-series model for 1920-62 that only includes religion and education as independent variables, we get a model r-square of .692, with partials of -.730 for proportion in K-12 education and -.395 for proportion Protestant. Adding in similar economic data as the Ontario case reduces the number of cases to just 12, limiting the usefulness of the data. Evidently, our best Newfoundland model will be a time-series cross-section based on more complete census and Orange data for the years 1911-1991.

Northern Ireland

Time-Series Cross-Sectional Data for 1891-1971

Northern Ireland is in many ways the most pressing of our cases, given the important political role played by Orangeism in the province. The history of Northern Ireland since partition in 1922 has been punctuated by cycles of peace, protest and violence. However, peace had been the norm until the 1960's. Only the immediate post-partition period and the IRA's modest border campaign of 1956-62 punctuated the relative calm engendered by the Protestant-dominated Stormont regime. This period witnessed sustained economic development and industrialisation in Northern Ireland, including the flourishing of Belfast's famous, Protestant-dominated shipping industry. Demographic change was also a feature of this period as the fertility differential between Catholic and Protestant rose while emigration drew proportionally more Catholics away from the province.

Our county-level model for this period contains only about a third of possible cases due to the paucity of lodge reports retained at Grand Lodge level. Further fieldwork may help to fill the gaps in the record, particularly for the agrarian and strongly divided mid-Ulster region and Londonderry City - both of which are under-represented in this sample. The sample is also skewed somewhat toward 1971, since we have a complete set of Orange data for that census year. Notwithstanding these limitations, this is a solid dataset, with no estimated data, which provides decent temporal and geographic coverage of the province. Results are presented below:

Independent Variable		Partial Correlation	Standard Error	Significance (Sig)
Protpc	Proportion Protestant	-.736	.078	.001
COI	Church of Ireland population as a proportion of Protestants	.579	.116	.019
Model Total:		.757	.03344

Our model generates a score of .757 across 17 cases. The main predictor variables are the proportion of Protestants/Catholics (-.736) and the proportion of adherents of the Church of Ireland (.579). Here we find the pattern we might have expected in an atmosphere of sectarian tension - a pattern generally absent in Ontario. Hence Orange density tends to rise where Protestants feel their hegemony to be weaker.

The reverse polarity attached to Church of Ireland membership is interesting. Together with the strong negative correlation (-.666) between Orange density and proportion Presbyterian, we must conclude that Orangeism is more prevalent in districts where the Church of Ireland is strong. This may reflect a regional bias: the COI is more strongly associated with English-settled regions of mid-Ulster while Presbyterianism is more powerful in the more Protestant, Scottish-settled eastern zones of the province. (Boal 1982)

The absence of any significant impact from variables related to occupation, industry, urbanisation, population density or housing stock is surprising, but does not rule out the role of structural factors. Perhaps the higher number of observations produced by a more detailed time-series cross-sectional analysis at the parish/barony level may highlight such trends. For the present, though, structural factors appear to be strikingly irrelevant when it comes to explaining patterns of Orange activity.

Turning briefly to the question of time-series models, we must take note of the poor state of available membership data for Northern Ireland in the 1890-1966 period. What we have is largely confined to the least powerful Orange areas in the eastern part of the province. Meanwhile, data for structural variables is in the process of digitization. Even so, we can use available and estimated membership data for 119 cases in the counties of Belfast, Antrim and Down to compare across a minimal set of variables for political events (war, depression) and religious composition. Doing so, we find that the proportion of Protestants is the most powerful predictor at -.688, followed closely by the events of the partition period of 1919-23, which have a positive impact on Orange density of .656.[9] Here we find that, in contrast to Ontario, the religious mix is important in both time series and cross-section, though firm conclusions must be withheld until structural variables are added to the model.

1966-Present

The contemporary period of the 'Troubles' in Northern Ireland begins in 1968 with the advent of the Northern Ireland Civil Rights movement for Catholics, which led to a crisis due to the response of the Protestant-dominated Stormont regime and the violent counter-involvement of the IRA. Fortunately, we have obtained access to Northern Ireland membership data for the critical period 1966-95. Time series comparison data for Northern Ireland is still being collected and digitised, but is less comprehensive than in the Canadian case. Currently, the model contains annual data on religious affiliation, Protestant fatalities caused by Republican (i.e. Catholic) terrorist organisations and male unemployment. There are also dummy variables for three important historical events: the civil rights movement (1968-74), direct rule (1969-72), and peace processes (1974-Sunningdale, 1985-Anglo-Irish Agreement, 1993 - Downing Street Declaration). Once again, we must caution that this model is missing key elements of socioeconomic and opinion poll data. However, it is interesting to examine the results of even this, preliminary, model, shown below.

Independent Variable		Partial Correlation	Significance (Sig)	Standard Error
Civrt	Civil Rights Period, 1968-74	.707	.020	.000
Protfatal	Protestant Fatalities caused by Catholic organisations	.565	.002	.001
MaleUE	Male Unemployment	-.556	.018	.000

The most glaring result of this exercise is the relationship between religious composition and Orange density. The Protestant proportion of the population did not emerge as an important predictor of Orange membership. However, the zero-order correlation between proportion Protestant and Orange density is actually strongly positive (.770) in this period. What we find therefore is that, in this limited model of the 1973-1995 period, Orange density tends to fall even as the proportion of Catholics rises. This pattern matches that of Ontario and contradicts the results of our analyses of Northern Ireland for the 1891-1971 period. Perhaps we can infer that when Orangeism is in decline (as it is in this period in Northern Ireland), demographic changes which increase the Catholic population do not stimulate Orange growth.

On the other hand, events and economic factors appear more critical in the recent period since the 'Troubles' began. The civil rights movement, for example, serves as the most powerful predictor of Orange density in contemporary Northern Ireland at a partial correlation of .707. One might surmise that this movement energised Orangeism - indeed, membership rolls broke trend and began to turn upwards with the advent of this Catholic movement. Protestant fatalities (.565) runs second to the civil rights movement in importance as a positive predictor of Orange density. This is in many ways unsurprising: the idea that Orangeism would be energised by sectarian conflict seems intuitive. In some ways, one might have expected a stronger relationship, though such a trend will be more clearly illuminated as more cultural and economic variables are added to the model.

Lastly, male unemployment (-.556) registers as a vital component of our analysis: it vastly improves the predictive power of our model and serves as an important force in dampening Orange membership. It questions the 'Protestant alienation' thesis concerning the reduced upward mobility of Protestant workers due to industrial decline since one might expect disaffected workers to swell Orange ranks in protest. (Dunn & Morgan 1994) On the other hand, it is certainly possible that Protestant alienation may have translated itself into other forms, including paramilitaries or marching bands. (Bryan 2000; Fraser 2000)

Northern Ireland Summary

Overall, our analysis seems to indicate that the relationship between Orange density and religious composition changes some time around 1970. Prior to this date, Orangeism is strengthened as Protestants become a smaller share of the population. This is similar to what we found for high-Catholic counties in Ontario. At first glance, the relationship is stronger in Northern Ireland since it holds across the entire dataset, rather than merely over a subset of it. On the other hand, when we account for the low number of cases and the higher Catholic population in Northern Ireland, the picture in Ontario and Ulster can be amalgamated into a general theory that Orange density is strongly affected by the presence of a Catholic minority of more than twenty percent.

In time-series, Northern Ireland and Ontario part company since the rising Catholic population in Northern Ireland during 1891-1971 prompted an increase in Orange membership. However, after 1971, the relationship between Orange vitality and Protestant demography reverses itself and assumes the Ontario pattern. Here, Orange decline and Protestant demographic decline proceed in tandem. Events appear more important in the Northern Ireland case, where the immediate post-World War I/Partition period led to a strong surge in Orange membership. The Civil Rights movement of 1968-74 had a similar, dramatic, result. Finally, structural factors are generally muted in Northern Ireland - as in Ontario - with the caveat that male unemployment is of importance in the recent period. It should also be noted that it is at present unclear whether education levels might serve as a strong predictor of Orange density in Northern Ireland, as they have in the Canadian case. The absence of a census question on this matter means that we cannot rule out this possibility.

Once again, caution is advised when it comes to investing these results with final significance. The addition of census, household facility and opinion survey data is needed to round out our post-1968 time series. Moreover, further field exploration is required in order to expand upon the county-level time series. Finally, GIS areal interpolation will allow us to combine pre- and post-1966 datasets in such a way as to surmount the administrative boundary changes of 1974. In combination with a finer-grained, parish and barony-level analysis, this will generate a far higher number of cases for analysis - particularly for the critical post-1966 period.

Scotland

As in the case of Northern Ireland (time-series excepted), we have a comprehensive set of census data with which to generate independent variables for analysis of Scottish Orange density. Scottish Orangeism generally did not attain high levels of density. Indeed, there is no membership at all in the Highlands or Borders regions and only limited representation outside the west-central belt of the Lowlands. Inside this area, though, particular Glasgow wards and certain districts of Lanarkshire and Renfrewshire approach or somewhat exceed levels attained by Ontario during its membership peak. It is also worth noting that the organisation resisted decline for longer than other Orange jurisdictions, so it forms an interesting case for comparison.

Our model in this case is quite comprehensive in terms of independent variables, as it includes the complete set of census, income, electoral and registrar general data gathered by Michael Hechter for his important work on internal colonialism in the Celtic periphery. The relative completeness of our Scottish Orange data also helps to ensure a high number of cases. This data thus forms our most complete time-series cross-section, with over 50 initial independent variables prior to stepwise regression. However, this panel data is restricted to the 1901-61 period because of boundary changes and Orange data for which I am still negotiating access.

Furthermore, unlike the other jurisdictions under scrutiny, the Scottish membership figures combine male, female and junior figures. The latter two categories are much more important in Scotland than elsewhere, accounting for as much as 50 to 65 percent of membership after 1930. Adult male membership must be estimated from the data, and though I have done this for the time-series analysis, some caution is necessary when considering these figures. Results from our stepwise regression analysis for the 1901-61 period, using total rather than male figures, appear below:

Independent Variable		Partial Correlation	Significance (Sig)	Standard Error
Ire50lag*cath	Interaction of Irish-born population of 50 years ago and present Catholic share of population	.754	.000	.000
Nonc_prot	Interaction effect between non-conformist Protestant population and Protestant share of population	-.604	.000	.023
Propeng	Proportion English	-.525	.000	.034
Marrate	Marriage rate	.495	.000	.003
Celt	Proportion Celtic speakers	-.294	.039	.050
Model Total		.750

Notice that these results appear to combine findings from the Ontario time-series of 1901-61 and panel analyses in that they emphasise the importance of the contemporary Catholic and Irish heritage proportion of the population but also demonstrate a limited significance for structural variables. Catholic and Irish heritage both tend to predict higher Orange densities. This may point toward the maintenance of a tradition of Orangeism within Irish-Protestant descended families, a phenomenon documented in the literature on Scottish Orangeism in the nineteenth century. (MacFarland 1990; Marshall 1996)

The proportion of Protestants, in interaction with the proportion of nonconformists, tends to dampen Orangeism. Bearing in mind that nonconformists are vastly over-represented in earlier decades, it is nonetheless interesting that non-conformity seems to be associated with lower levels of Orange activity. This might be due to the Orange link with established institutions like the monarchy, but it is also probably related, like Celtic language ability, to Scottish highland migrants, who would have had no exposure to Orange traditions.

The importance of the English population, which does not exceed seven percent in any of our cases, and its negative correlation with Orange density, is uncertain. Englishness seems to correlate moderately well with the proportion of civil servants, those over 65 and those with higher incomes. It has a negative relationship with birthrates, religiosity and the proportion of people in manufacturing. Thus the influence of the English could be due to their less sectarian worldview. They may thereby have been influential in institutions of religion, commerce and government where they were better represented. On the other hand, it may arise because the English proportion of the population is somehow related to economic development. We know that the cases with the highest proportion of English are mostly from the more recent 1951-61 period. Meanwhile, the English are over-represented in the East and Mid-Lothians and Fifeshire counties, which contain relatively few members.

Our most important structural variable is marriage rate, which correlates well with population density and proportion urban. This variable does not seem well-correlated with economic indicators like income or occupation, so any broadbrush conclusions about the role of industrialization must be put in context. In addition, no data on education appears to have been collected for Scotland in this period, hampering a direct comparison with Ontario, where this factor was extremely significant. Finally, if we isolate male adult membership and run the same model, we get a similar set of results across 55 cases, with the current proportion of Irish-born (.750) and nonconformist Protestants (-.594) as the leading explanatory variables and total population (.526) and marriage rate (.352) as the most important structural factors.

The impact of the 'ethnic' or 'immigrant' factor, along with sectarianism, is highly pronounced for 1901-21 (28 cases), and the best predictors are the current proportion of Catholics in interaction with Irish-born proportion of 50 years ago (.886), and the proportion of Celtic language speakers in interaction with Irish-born proportion of 50 years ago (-.601). Indeed, these factors predict Orange density to an R-squared of .836. The negative effect of Celtic language, which we noticed in our previous analysis, probably reflects the limited appeal of Orangeism among Highland Scots and Highland migrants - a finding which makes sense given the absence of any Orange lodges in the Highlands. Our cases include no data for the Highlands, and when we add in data for Argyll in 1961, where Celtic speakers are well-represented, the importance of this variable increases. No doubt we would find a stronger effect from this variable if we expanded our range of cases to include the Highland counties.

Finally, the contemporary Irish population does not appear significant, thus it is probably the case that early-to-mid twentieth century immigrants from Ireland contained fewer Protestants. Probing more deeply, we find that for 1901-31, the main factors are a) interaction of Irish-born of 50 years ago and Catholic today (.884); b) nonconformity

(-.485); and c) celtic-speakers (-.446). The R-squared of the model is .812 across 36 cases.

An analysis of the more recent period of 1931-61 also flags up the Irish factor: the Irish-born population of today, interacting with the current proportion of Catholics (partial = -463) and that of 50 years ago (.683) are selected by SPSS as the important variables. No structural variables appear. Evidently, Irish immigrants no longer excited a Protestant reaction, and/or failed to bring Orange traditions with them by this time. However, our two factors only predict Orange density at a lower R-squared of .599. On the other hand, isolating the post-war 1951-61 period (N=19) raises the importance of the Irish-born proportion of 50 years ago to (.909) and that of the current proportion of Catholics interacting with Irish to (-.838).

This strong inverse pattern between traditional and contemporary Irish-born populations reinforces the earlier claim that traditions bequeathed by Irish Protestant immigrants of the 1880s and 1890s - possibly within families or local networks - rather than living anti-Catholicism, kept Orangeism vital during this period. This fits in with patterns of twentieth century sectarian voting behaviour detailed by David McCrone and Tom Gallagher, among others. (McCrone 1992; Gallagher 1987) Partial correlations with population growth (-.822) and infant mortality rate (.678) suggest that structural variables appear to take on more importance in the post-war period.

Time Series Analysis

Given the excellent annual record of Scottish Orange membership, it was interesting to construct a time series which employs an interpolated variant of Hechter's variables to bridge inter-censual periods.[10] This makes sense insofar as many variables (income, urban/rural, nonconformity, etc) tend to trend in the same direction over time. On the other hand, these results need to be taken with a great deal of caution given the large degree of estimated data. The result of this exercise appears below:

Independent Variable		Partial Correlation	Significance (Sig)	Standard Error
Popdens	Population Density	.866	.001	.000
WW II	World War II	-.617	.000	.000
WW1lag5	World War I - five-year aftermath	.696	.000	.000
WWIlag10	World War I - ten year aftermath	.580	.000	.000
Deprlag1	Depression - One year aftermath	-.463	.000	.000
WW2lag1	World War II - one year aftermath	.430	.000	.001
Model Total		.883		.00040

Despite the large amount of estimation, it is interesting that across the 61 cases examined here (1901-61), structural variables become more important. Population density, which is well-correlated with urbanisation and the marriage rate, gives us our best partial correlation (.866). Something about the move to towns and cities - more so than any changes in the mode of production - seems to have encouraged Scottish Orange membership. Next in importance are factors related to war and depression. World War II itself had a dampening effect on Orangeism (-.617), but war's end brought a positive (.430) pressure to bear on the organisation. World War I seems to have had limited effect - this can be explained partly by lodges keeping serving members on the books - though both five (.696) and ten years (.580) after the British victory, Orangeism received a boost in fortunes.

The 'ethnic' factor only makes its appearance in the 1945-61 period, when Irish population of 50 years ago, in interaction with current Catholic share of population, explains .750 of the variance for these 16 cases. Conversely, in the earlier 1900-45 period, the lags for 5 (.765) and 10-years (.655) of World War I, together with population growth rate (-.609) are important. However, before we conclude that cultural factors are silent in time-series analysis, we need to examine figures for adult male Protestants. Main explanatory variables here include: Catholic % of 50 years ago (-.854), World War I (-.364), World War II one year lag (.443) and World War II's five year lag (-.398). Notice that these figures re-establish the importance of cultural explanation and confirm the positive stimulus for Scottish Orangeism of the end of World War II.

Theoretical Implications

What are the theoretical implications of these findings in comparative perspective? To some extent, firm conclusions must be suspended until further data becomes available. Nonetheless, our single most important finding is unlikely to be affected by additional data. Namely, that cross-sectional analysis flags 'modernising' factors (population density, population growth, marriage rate, education) to be more important while cultural ones (proportion of Catholics, proportion of Irish 50 years ago, proportion of Irish Protestants, Protestant denominational mix) have a stronger weight in cross-sectional analysis.[11] Northern Ireland is largely an exception, Scotland less so.

How to explain this disjuncture? One interpretation might be that the meaning of Irish or Catholic in time series differs slightly from that in cross-section. After all, the change in ethnic or religious composition over time often owes more to slow-moving fertility differentials than it does to the rapid social change wrought by settlement and immigration. For instance, it may be that the initial Irish and Catholic migratory patterns which established cross-county differences in religious makeup generated considerable Orange dynamism in Canada and Scotland. However, this is everywhere a heavily nineteenth century phenomenon. The twentieth century witnessed a slowing of both Catholic,[12] and especially Irish Protestant, immigration to Scotland and Canada. In Northern Ireland, geographic differences in religious composition were established (Belfast excluded) as early as the seventeenth century. So it is probably not surprising that this factor predicts powerfully in cross-sectional terms but weakly - if at all - in time series.

It could be the case, therefore, that once the initial Irish migratory patterns became established, community and family tradition - more so than sectarian or ethnic conflict - maintained Orangeism's vitality. This recalls Stinchcombe's historicist causal explanatory scheme, whereby historical templates impart path-dependency to subsequent social organization, regardless of latter-day sources of stimuli. (Stinchcombe 1968) In such an environment, steady demographic change would not provide the ammunition for Orangeism's expansion.

A similar story might be adduced from the data on religious denomination. Nonconformity, which tends to be negatively related to Orangeism, crested in the middle nineteenth century in Scotland, Northern Ireland and Ontario. Settlement patterns were being established at this time which may have entrenched the cross-sectional effects we observe. As with Catholic/Protestant migration, it was only under the impact of rapid change that these differences produced an Orange response. Thereafter, Orangeism became more tradition-dependent and its membership strength tended to march more strongly to the tune of political events or modernising variables like education or population density.

In terms of our theoretical mission, we might say that cultural factors appear to take precedence over structural effects within our time-series cross-sectional analysis. In time-series, however, our task is more difficult, since we cannot as yet discern whether the economic, demographic or cultural aspects of modernisation loom largest. In Ontario, where we have the best set of data, education appears to be the premier agent of Orange time-series decline. In Scotland, religious composition was most important for male lodges, but across the full spectrum of Scottish membership, factors related to demographic modernisation (marriage rate, birth rate, population growth, population density) took precedence. In Northern Ireland, where the time-series data is weakest, religious composition was important only for the pre-1971 period when Orangeism developed in response to Catholic growth. After that point, political events became the most potent predictors. War and depression, along with their after-effects, are of moderate importance in all locations.

Evidently political and economic variables play their part in our story. Yet were we to try and draw an interim assessment, we need to highlight the fact that where the data is strongest, cultural factors (education, religious composition, ethnic composition) are the best predictors of adult male Orange membership density. The limited impact of economic indicators (especially occupational change and income) is another important finding. Political events and demographic changes lie between cultural and economic variables in their predictive power. Finally, we have seen that there is an uneven set of causal factors operating across the four national locations: each of our jurisdictions has a particular historical and geographic matrix which moderates the universalising impact of our model's main explanatory variables.

Turning from social theory to social capital theory, our analysis appears to lend little support to Beito's argument regarding the replacement of mutual aid with government welfare. Only in Ontario did the timing of Orange decline correspond to the advance of the welfare state. In Scotland and Northern Ireland, Orangeism shows little conjunction with the rise of the post-World War II British welfare state. Even in Ontario, where the timing is right, our time-series analysis indicates that the expansion of social security, old age pensions and family/mother's allowances is of only limited importance in explaining the decline in Orange density. This is not to dismiss Beito's main premise since he is primarily addressing life insurance fraternal orders. Yet it should serve as a cautionary note to those who would generalise his findings beyond these organisations.

J.C. Herbert Emery's thesis that rising income might explain fraternal decline should also be questioned since income plays little role in our model. The parallel trajectory of Orange and Oddfellow membership in Ontario raises the additional question of whether Emery's model is even approporiate for the Oddfellows.[13] On the other hand, both Beito and Emery's observations regarding changing norms seem plausible due to the primacy of education over structural factors like income in explaining Orange decline. Education might be acting as a proxy for some form of cultural change (i.e. liberalism, nationalism) that is independent of structural modernisation. This affirmation of Emery and Beito's ideas should not extend to changes in technology or taste, however, since Ontario Orange membership does not decline with the rate of per capita auto permits or correspond to Canadian movie theatre attendance growth.

Finally, the absence of a strong dataset for the post-1961 period means that we cannot, as yet, adequately assess Robert Putnam's thesis. Certainly the pattern of membership in all four of our study locations seems to illustrate a steady Putnamite pattern of decline in the post-1950's period. Yet the most critical declines in Canada seemed to take place in the 1920's and 30's while Scotland and parts of Northern Ireland did not experience their modern phase of decline until the 1960's, 70's or 80's. This may be due to differing rates of television penetration and generational turnover, but initial models for Ontario and Newfoundland for the 1951-65 period suggest that television is less important than the telephone in explaining Orange decline (and actually has a positive impact on membership!). Needless to say, much more data is needed before we can confirm this finding. Generational change likewise stands outside the ambit of the present model, though it shall be included once we amass a dataset that extends through to the present period.[14]

Conclusion

The Loyal Orange Association is the most important voluntary association for the Protestant community in Northern Ireland. This used to be the case in many English-speaking provinces of Canada as well, particularly Ontario, New Brunswick and Newfoundland. Indeed, our analysis of surviving Orange membership returns suggests that the Orange Order was generally most powerful in Newfoundland, followed by Northern Ireland, Ontario and then Scotland. Decline has been the order of the day for the organisation in all jurisdictions since the eighties, and in most areas since the 1950's. This process has resulted in the elimination of the Order's political influence in Ontario and Newfoundland, but has yet to affect the Order in Northern Ireland, where nearly all Ulster Unionist Party MPs and MLAs are members.[15]

Our analysis to date, though incomplete, suggests that cultural factors related to religious and ethnic composition, as well as education, are the best predictors of Orange membership patterns. Economic factors generally correlate poorly with Orange membership density. Political events lie in between the cultural and economic in their importance. This picture, however, conceals great diversity. Cultural factors are more important across space than across time. Economic factors play a greater role after World War II than before it. Finally, political events are more important in Northern Ireland than elsewhere. Depression and war had little systematic effect on membership. The only generalisable finding is that the end of World War II buoyed membership in all locations.

Beyond these general effects, we can specify some surprising particular ones. First of all, Orangeism has been in decline across all locations since the 1980's, and across most locations since the 1960's. Northern Ireland is not an exception to the rule. Secondly, since the advent of the 'Troubles,' religio-ethnic division appears to have abated as a spur to Orange growth in Northern Ireland. In fact, the decline in Ulster Orangeism is now positively associated with the relative decline of the Protestant population - much as was the case in Ontario during 1901-61.

Finally, the role of ethnic and religious factors (notably Catholicism or Irish-Protestant ethnicity) is more powerful in cross-sectional than in time-series analysis. This suggests that while sectarianism and Irish-Protestant migration flows helped to stimulate Orange membership in the eighteenth and nineteenth centuries, this pattern no longer holds true today. Geographic differences bear the fossils of these earlier dynamics, but now it is community and family tradition, not Protestant-Catholic antagonism, that keeps the Orange flame from flickering out.

REFERENCES

Primary

TO-1999 - Interviews conducted with Orange Order and Ladies Orange Benevolent Association members, ex-members, and their children in the Greater Toronto Area, July-August 1999.

BFST-2000 - Interviews in Belfast and Co. Tyrone, N. Ireland, September 2000

GLAS-2000 - Interviews conducted at Glasgow, Scotland, September 2000

GOLI - Grand Lodge of [Northern] Ireland county reports

GOLIret - Grand Lodge of [Northern] Ireland county returns

GOLS - Grand Lodge of Scotland reports and media files

ICGW - Imperial Grand Council of the World reports

CGL - Canadian Grand Lodge reports

GOLOW - Grand Lodge of Ontario West reports and media files

GOLOWret - Grand Lodge of Ontario West provincial returns

LOBA - Ladies Orange Benevolent Association, Canada, annual reports

LOBA Ont. E - Ladies Orange Benevolent Association of Ontario East, provincial returns

Secondary

Adamson [1982] 1991. The Identity of Ulster: the Land, the Language and the People (Bangor: Pretani Press)

Bairner, Alan. 1997. ' 'Up To Their Knees?': Football, Sectarianism, Masculinity and Protestant Working-Class Identity,' in P. Shirlow and Mark McGovern (eds.) Who Are the People?: Unionism, Protestantism and Loyalism in Northern Ireland (London: Pluto Press)

Beito, David T. 2000. From Mutual Aid to the Welfare State: Fraternal Societies and Social Services (Chapel Hill & London: University of North Carolina Press)

Bell, Desmond. 1990. Acts of Union: Youth Culture and Sectarianism in Modern Ireland (London: MacMillan)

Berger, Carl. 1969. Imperialism and Nationalism, 1884-1914: A Conflict in Canadian Thought (Toronto: Copp Clark Publishing)

Berger, Carl. 1970. The Sense of Power: Studies in the Ideas of Canadian Imperialism, 1867-1914 (Toronto and Buffalo: University of Toronto Press)

Bibby, Reginald W. 1987. Fragmented Gods: The Poverty and Potential of Religion in Canada (Toronto: Irwin)

Blake, Donald. 1968. Canadian Census and Election Data, 1908-68. Dataset held at Inter-University Consortium for Political Research.

Boal, Frederick and J. Neville Douglas (eds.) 1982. Integration and Division: Geographical Perspectives on the Northern Ireland Problem (London: Academic Press)

Bruce, Steve. 1998. Comparative Protestant Politics (Oxford: Oxford University Press)

Buckner, Philip. 1998. 'Migration and National Identity in Canadian History.' Paper delivered at Association for the Study of Ethnicity & Nationalism seminar, London School of Economics

Burnet, Jean. 1972. Ethnic Groups in Upper Canada. Ontario Historical Society Research Publication no.1.

Carnes, Mark C. 1989. Secret Ritual and Manhood in Victorian America (New Haven, CT: Yale University Press)

Cheal, David J. 1981. 'Ontario Loyalism: A Socio-Religious Ideology in Decline,' Canadian Ethnic Studies, 13, 2: 40-51.

Clawson, Mary Ann. 1989. Constructing Brotherhood: Class, Gender and Fraternalism (Princeton: Princeton University Press)

Cole, Douglas L. 1970. 'Canada's 'Nationalistic' Imperialists', Journal of Canadian Studies, Vol.V, No. 3

Coulter, Colin. 1997. 'The Beliefs of the Unionist New Middle Classes,' in P. Shirlow and Mark McGovern (eds.) Who Are the People?: Unionism, Protestantism and Loyalism in Northern Ireland (London: Pluto Press), pp. 117-38

Craith, M. Nic. 2001. 'Politicised linguistic consciousness: the case of Ulster-Scots,' Nations & Nationalism, Vol. 7, Part I, pp. 21-37

Deutsch, Karl. 1953. Nationalism and Social Communication: An Inquiry into the Foundations of Nationality (Cambridge, MA: M.I.T. Press).

Dumenil, Lynn. 1984. Freemasonry and American Culture 1880-1930 (Princeton, NJ: Princeton University Press)

Durkheim, Emile. [1893] 1984. The Division of Labour in Society, Introduction by Lewis Coser, translated by W.D. Halls (Houndsmills & London: Macmillan).

Dunn, Seamus & Valerie Morgan. 1994. Protestant Alienation in Northern Ireland: A Preliminary Survey (Coleraine: Centre for the Study of Conflict)

Edwards, Ruth Dudley. 1999. The Faithful Tribe: An Intimate Portrait of the Loyal Institutions (London: Harper Collins)

Emery, J.C. Herbert. 1999. A Young Man's Benefit: The Independent Order of Oddfellows and Sickness Insurance in the United States and Canada, 1860-1929 (Montreal & Kingston: McGill-Queens University Press)

Errington, Jane. 1987. The Lion, the Eagle, and Upper Canada: A Developing Colonial Ideology (Kingston and Montreal: McGill-Queen's University Press)

FitzGerald, Frances. 1979. America Revised: History Schoolbooks in the Twentieth Century (Boston & Toronto: Little, Brown & Company)

Fitzgerald, John. 1998. 'Newfoundland Referenda Campaigns of 1948,' The Beaver, February-March, p.3

Gallagher, Tom. 1987. Edinburgh Divided: John Cormack and No Popery in the 1930s (Edinburgh: Polygon)

Giddens, Anthony. 1991. Modernity and Self-Identity (Cambridge: Polity Press)

Gray, Tony. 1972. The Orange Order (London: Bodley Head)

Hechter, Michael, County Data for the British Isles, 1851-1961. Dataset held at UK Data archive.

Hechter, Michael. 1975. Internal Colonialism (London: Routledge and Kegan Paul)

Higham, John. [1955] 1986. Strangers in the Land: Patterns of American Nativism, 1860-1925 (2nd ed.) (New Brunswick: Rutgers University Press)

Horowitz, Donald L. 1985. Ethnic Groups in Conflict (Berkeley: University of California Press)

Houston, Cecil and William J. Smyth. 1980. The Sash Canada Wore: A Historical Geography of the Orange Order in Canada (Toronto; Buffalo: University of Toronto Press) 1980

Inglehart, Ronald. 1990. Culture Shift in Advanced Industrial Society (Princeton, NJ: Princeton University Press)

Ingram, P. 1988. Sectarianism in the North West of England with special reference to class relationships in the City of Liverpool (Unpublished PhD Dissertation, Lancashire Polytechnic)

Jamieson, Donald, 'I saw the Fight for Confederation,' in J. Smallwood (ed.), 1991. The Encyclopedia of Newfoundland (St. John's, NF: Joseph R. Smallwood Heritage Foundation, Inc.), vol. 3, pp. 70-104

Kaufmann, Eric. 1997. 'Condemned to Rootlessness: The Loyalist Origins of Canada's Identity Crisis,' Nationalism and Ethnic Politics, vol.3, no.1, (1997), pp. 110-135

Kaufmann, Eric. 1998. ' 'Naturalizing the Nation': the Rise of Naturalistic Nationalism in the United States and Canada,' Comparative Studies in Society and History, vol. 40, no. 4 (October), pp. 666-695

Kaufmann, Eric. 2000. 'Dominant Ethnie,' The Companion Guide to Nationalism (London: Transaction Publishers)

Kaufmann, Eric. 2002. ‘The Decline of Sectarianism in the West?: A Comparison of the Orange Order in Canada, Ulster and Scotland,' in N. Singh and T. Vanhanen (eds.) Ethnic Violence and Human Rights (Delhi: KIRS, 2002) - forthcoming.

Kerr, Clark, John R. Dunlop, Frederick Harbison, and Charles Myers. 1960. Industrialism and Industrial Man (Cambridge: Harvard University Press)

MacFarland, Elaine. 1990. Protestants First: Orangeism in Nineteenth Century Scotland (Edinburgh: Edinburgh University Press)

MacRaild, Donald. 1998. Culture, Conflict and Migration: The Irish in Victorian Cumbria (Liverpool: Liverpool University Press)

March, J.G. and J.P. Olsen. 1984. 'The New Institutionalism: Organisational Factors in Political Life,' American Political Science Review, vol. 78, no. 3.

Marshall, William S. 1996. The Billy Boys: A Concise History of Orangeism in Scotland (Edinburgh: Mercat Press)

McCrone David. 1992. Understanding Scotland: The Sociology of a Stateless Nation (London: Routledge)

Moore, Leonard J. 1991. Citizen Klansmen : the Ku Klux Klan in Indiana, 1921-1928 (Chapel Hill & London: University of North Carolina Press)

Moxon-Browne. 1983. Nation, Class and Creed in Northern Ireland (Aldershot: Gower)

McCrone David. 1992. Understanding Scotland: The Sociology of a Stateless Nation (London: Routledge)

O'Dowd, Liam. 1991. 'Intellectuals and Political Culture: A Unionist-Nationalist Comparison,' in Eamonn Hughes (ed), Culture and Politics in Northern Ireland, 1960-1990 (Milton Keynes: Open University), pp. 151-73

Parsons, Talcott. 1951. The Social System (London: Routledge & Kegan Paul).

Pennefather, R.S., The Orange and the Black (Toronto: Orange and Black Publications, 1984)

Putnam, Robert D. 1995. 'Tuning In, Tuning Out: The Strange Disappearance of Social Capital in America,' Political Science and Politics, vol. 28: 664-83

Putnam, Robert D. 2000. Bowling Alone: The Collapse and Revival of American Community (New York: Simon & Schuster)

Rasporich, A.W. 1968. 'Imperial Sentiment in the Province of Canada during the Crimean War, 1854-1856', in Morton, W.L. (ed.), The Shield of Achilles: Aspects of Canada in the Victorian Age (Toronto: McClelland & Stewart Ltd.)

Richard, Madeline A. 1991. Ethnic Groups and Marital Choices: Ethnic History and Marital Assimilation in Canada, 1871 and 1971 (Vancouver: UBC Press)

Schwartz, Mildred A. 1967. Public Opinion and Canadian Identity (Berkeley: University of California Press)

Schwartz, Mildred A., Public Opinion and Canadian Identity (Berkeley: University of California Press 1967

See, Katherine O'Sullivan. 1986. First World Nationalisms: Class and Ethnic Politics in Northern Ireland and Quebec (Chicago & London: University of Chicago Press)

Senior, Elinor Kyte. 1959. The Origins and Political Activities of the Loyal Orange Association in Newfoundland (Unpublished M.A. thesis, Memorial University of Newfoundland)

Senior, Hereward. 1972. Orangeism, the Canadian Phase (Toronto; New York: McGraw-Hill Ryerson)

Smith, Anthony D. 1981. The Ethnic Revival (Cambridge: Cambridge University Press)

Stinchcombe, Arthur L. 1968. Constructing Social Theories (Harcourt, Brace & World)

Walker, Graham. 1992. 'The Orange Order in Scotland Between the Wars,' International Review of Social History, vol. XXXVII, no. 2

Walker, Graham. 1995. Intimate Strangers: Political and Cultural Interaction Between Scotland and Ulster in Modern Times (Edinburgh: John Donald)

Wallace, W. Stewart. 1921. The United Empire Loyalists: the Beginnings of British Canada (Toronto: Glasgow, Brook and Co.)

Waller, P. J. 1981. Democracy and sectarianism: a political and social history of Liverpool 1868-1939 (Liverpool: Liverpool University Press)

---

[1] I am grateful to the Economic and Social Research Council, which has funded the study upon which this work is based. I would also like to acknowledge the advice and assistance of J.C. Herbert Emery, Donald Blake, Cecil Houston and Henry Patterson.

[2] Though the Quebec wing was noticeably weak, even allowing for that province's low Protestant population.

[3] For the latter view, see Marwick, Arthur, War and Social Change in the Twentieth Century: a Comparative Study of Britain, France, Germany, Russia and the United States (London: Macmillan, 1974).

[4] Putnam distinguishes between 'bridging' social capital - networks which link individuals across established social boundaries, and 'bonding' social capital which tends to reinforce links within established social groups of ethnicity or class. (Putnam 2000: 22-24)

[5] In the words of the Grand Master of Ontario East: 'It is surely essential for the future of Canada that British influence remain dominant in this country and that is why we call upon the Government to formulate an immigration policy that will assure at least fifty per cent of immigrants being British.' (Grand Lodge of Ontario East 1955: 7)

[6] Please see Kaufmann, E. P. 2002. ‘The Decline of Sectarianism in the West?: A Comparison of the Orange Order in Canada, Ulster and Scotland,' in N. Singh and T. Vanhanen (eds.) Ethnic Violence and Human Rights (Delhi: KIRS); also Kaufmann, E.P. 2001. 'The Demise of Dominant Ethnicity in English Canada?: An Empirical Examination of Orange Order Decline, 1918-80,' available on project website <http://www.canadian-english.com/OrangeOrder/Orange_order_Project.htm>.

[7] Due to high collinearity among these variables, all but mother's allowances were combined into two principal components using data reduction analysis techniques.

[8] This is not very problematic since we know the ratio of lodges between Ontario East and West for this entire period, as well as membership for the years 1890-1911, 1918, and 1925-35. These show Ontario East membership to exhibit similar trends to that for Ontario West. However, for the present model, it was decided not to use estimated data.

[9] Of interest is the fact that religious composition was not the most important factor at county-level, though it had a strongly negative effect on Orange membership and was the second most important variable. Other variables in the model included World War I (.534), WWI five-year lag (.379), WWII one-year lag (.329), and Great Depression - one year lag (.305).

[10] This involves estimating data in a straight line between decennial censuses. Naturally this builds linearity into the total model, but it should not have a systematic effect on partial correlations. The fact that this model corresponds to the Ontario time series model (where there is no estimated data) in its relationship to cross-sectional findings provides an added measure of reassurance.

[11] Political events are not tested in time-series cross-sectional analysis due to ten-year inter-censual gaps which render their inclusion difficult.

[12] Whether Irish Catholic immigration or French-Canadian Catholic internal migration.

[13] I am grateful to Herb Emery for this data. Oddfellow membership trends in 20th century Ontario are a virtual carbon copy of Orange patterns - both in terms of the number of members and the pattern of growth and decline.

[14] In the absence of a body of data analagous to the American General Social Survey, we will need to specify a dummy variable for generation to cope with this problem. Additional insight will emerge through an analysis of surviving lodge roll books.

[15] It should be noted that many UUP politicians are not active Orangemen and that there have been attempts to sever the official link between the Orange Order and the UUP which runs through the Ulster Unionist Council. Such attempts at reform have met considerable resistance, however, and have generally come to naught.