Miles Fairburn and Stephen Haslett | Cleavage Within the Working Class? The Working-Class Vote for the Labour Party in New Zealand, 1911–51 | Labour History, 88

Cleavage Within the Working Class? The Working-Class Vote for the Labour Party in New Zealand, 1911–51

Miles Fairburn and Stephen Haslett^*

A convention in New Zealand historiography is that the electoral fortunes of the Labour Party in the first half of the twentieth century were tightly constrained by a particular structure of cleavage. Although by 1919 or 1922 Labour had won the support of the vast majority of urban working-class voters in the leading towns, it could not win office with their support alone. This was demonstrated in the 1930s when it gained electoral domination by winning over small farmers and the urban middle classes; and after the late 1930s when the support of non-manuals ebbed away and it lost office in 1949. This article tests the convention. It takes the ten largest provincial towns, determines the class composition of their streets in 11 general elections, estimates the distribution of the votes for each party in each street in each town for every year, and correlates the estimated percentages of Labour vote with percentage of working class. The results are the inverse of the trends claimed by the historiography. The article then examines the social geography of the towns and finds systematic evidence of an unexpected cleavage inside the working class: skilled workers had a much weaker tendency to vote Labour than the unskilled and no tendency at all to reside in the same residential areas as the unskilled.

Introduction

The term 'cleavage', although rather loose and lacking precision, generally refers to the influences of a country's underlying social and cultural differences and conflicts on the party preferences of individual voters at elections. By extension, it also refers to the effect these structural influences have on the ideologies and programs of a country's political parties.¹ Thus in a society where religious divisions are the dominant form of cleavage, each major political party will tend to align itself to a particular religious group and voters will tend to prefer the party that represents their particular faith over one that appeals to a rival. In most modern democracies, however, it is rare for any one form of cleavage to dominate. Usually, many cleavages operate simultaneously, the most common and powerful being those based on class, regional, linguistic, ethnic, religious, and demographic differences. All will shape individual voter choices but in complex ways, some mutually reinforcing, others cross-cutting. Moreover, as some cleavages are more important than others, their influence greatly varies. Although cleavages structure the ability of political parties to win votes, they do not determine election outcomes, for individual voters still make choices, and the actions of political parties and leaders can substantially diminish or enhance the structuring power of cleavages on voter choices and even create new cleavages.

Cleavages should be of considerable interest to historians, especially those working in the interface between social, political and labour history. More specifically, what labour historians should find interesting about cleavages is how they change over time and affect the electoral fortunes of left wing political parties. In their classic model of the history of cleavages in western democracies, Lipset and Rokkan suggest that they move through two broad phases, an early one where the basic cleavage is territorial and cultural, and a later phase where the basic cleavage is class and sectional.² In the particular case of Britain, the standard account of how its cleavage changed over the twentieth century, says that up to World War I, it was primarily based on region, class and religion (non-conformists versus Anglicans); for the next 50 to 60 years class replaced religion (to the point where from the mid 1940s class exercised more influence on voting than it did in any other democracy), and from 1970s it has increasingly been a mixture of region, ethnicity and housing tenure (a much debated trend known to political scientists as 'class dealignment').³

In the New Zealand historiography, there is likewise a consensus view about what sorts of cleavages have constrained party politics and voter choices over the twentieth century. This view consists of two key claims. The first is that right from the establishment of the modern party system and of the mass franchise in the 1890s, the cleavage structure was dominated by a geographical split between town and country overlaid by a class division between manual workers and non-manual workers. Cleavages based on religion and region also affected voter preferences but these were of minor importance, until the 1970s when, as in other democracies, the power of class rapidly declined.⁴

The second central claim in the historiography, to be discussed in detail later, is that the electoral success of political parties in New Zealand over the twentieth century has depended on their ability to appeal simultaneously to several opposing fractions making up the rural/urban and manual/non-manual divides.⁵ The pattern, the historiography goes on to say, is for parties to lose elections when supported by just one fraction of voters but to stay in office for long periods when supported by a stable coalition of several fractions. Thus the 21 years of electoral domination by the reforming Liberal Party from 1891 to 1912 was based on its success in stitching together a broad coalition of voters made up of the urban working class, small town businessmen and newly settled small farmers; the electoral domination of its right-wing successor, the Reform Party, from 1912 to 1928 stemmed from its ability to align itself both to urban non-manual and rural voters; and what enabled the right-wing National Party to win most general elections in the post-war era from 1949 came likewise from its skill in appealing to both urban non-manual and rural voters.

A leading figure in the creation of the consensus in the historiography was R.M. Chapman, an election historian and psephologist, who from 1948 to the 1980s wrote a series of pioneering works on the history of voting behaviour in twentieth century New Zealand, and left his mark on a whole generation of thesis students he supervised.⁶ It is not surprising, therefore, that sometimes the standard historical view about the cleavage structure also finds its way into studies of modern elections. For example, analysing the 1999 general election, one political scientist put Labour's performance in the following long-term perspective:

In close to a century of electoral competition, Labour's electoral heartland has been the low-income inner city and suburban seats of Auckland, Wellington, Christchurch, and Dunedin. With National (and more recently ACT) appealing strongly to those in the countryside and high-income city seats, the real election battle takes place in the highly marginal, though generally National-leaning, mixed-income electorates in the cities and provincial towns. Given the narrowness of Labour's regional appeal, electoral success has been dependent on its ability to reach into the provinces with a view to constructing a broad-based coalition of support.

Labour's first such coalition was assembled in 1935 ... As history shows, however, the party's appeal to rural and provincial town voters is tenuous at best. As the coalition began to unravel in the 1940s, rural voters were the first to desert, closely followed by their associates in the adjacent cities and towns.⁷

The intention of this article is to test what the historiography has said about the relationship between the cleavage structure and the electoral fortunes of New Zealand Labour Party (and its precursors) between 1911 and 1951. The article will begin by sketching the historical background, next it will spell out the claims in the historiography about the trends in Labour's sources of mass electoral support, it will then detail the methodology it uses to test these claims, and lastly it will discuss the results of the analysis.

Historical Background

New Zealand was a democracy from 1893 with a first-past-the-post electoral system in the Westminster tradition and a highly centralised system of government. It had over a million inhabitants by the beginning of the study-period, a highly advanced economy, and a record for progressive legislation. The trade unions began to play an active role in politics from the late 1880s, but after supporting the reformist Liberals, started to break away from them from about 1904. Over the next decade, the union movement established many left parties, with diverse objectives, reflecting its own acute divisions. By the general election of 1914, the left nonetheless had made sufficient inroads to secure 10 per cent of the total vote, and in 1916 finally managed to unite, forming the New Zealand Labour Party, thereafter New Zealand's only mainstream left party. Although in the 1919 general election Labour's share of the vote was double that received by the left in 1914, the Party made few gains in the 1920s. Its electoral doldrums forced it in 1927 to abandon its radical program of land nationalisation and, in effect, its socialist objectives (the first mainstream socialist party in the western world to do so without splitting). But its losing streak continued in the 1931 general election, when it secured just 35 per cent of the vote, despite its move to the centre, the advent of the Great Depression and the coalition of the two conservative parties which ended 20 years of three-party vote splitting. Not until the next election in 1935, did the Party's long period in the electoral wilderness come to an end, when it won a landslide victory with over 46 per cent of the vote. A spasm of reformist legislation and strong economic recovery then helped the Party to reach its peak level of voting support in the 1938 general election. It went on to rule continuously for the next 11 years, establishing the welfare state and a full employment regime based on intensive state economic regulation. Amidst post-war austerity, however, its support declined, to the point where it lost office in the 1949 general election to the right wing National Party. It then experienced a disastrous election defeat in 1951, when National held a snap election over the bitterest industrial dispute in New Zealand history.

Historiographical Conventions

According to the historiography, Labour's electoral fortunes over the study-period go through five phases and each is strongly related to the cleavage structure. In the first phase, the period covering the 1911 and 1914 general elections, the historiography says that the vast bulk of the vote for Labour's left-wing precursors came from manual workers in the leading towns and in exceptional rural seats (where there were large concentrations of unionised miners and timber workers), though many such voters continued to be loyal to their traditional party, the Liberals.⁸ Although the historiography acknowledges that the left was divided at this stage, there is general agreement that unity at the institutional level had largely been achieved by 1913 and that the divisions did not prevent the bewildering variety of socialist and labour candidates from increasing their overall share of the vote between the general elections of 1905 and 1914. In the second phase, from 1916 to 1919, the historiography relates the party's surge of support to its ability to win over the vast majority of the manual workers in the leading urban centres and in exceptional rural areas.⁹ In the third phase, from the early 1920s to the early 1930s, the historiography, with recent exceptions, maintains that manual workers in the major urban centres and in special rural seats remained loyal to Labour, but, it says, Labour failed to increase its overall share of the vote, because it failed to align itself with rural dwellers and urban non-manuals.¹⁰ Then in the fourth phase, from to the mid-1930s the 1ate 1930s, the historiography links Labour's electoral breakthroughs to its success in broadening its sources of voter support, winning not only its traditional support (urban workers and workers in exceptional rural seats), but also capturing for the first time mass support from urban non-manuals and small farmers.¹¹ Finally, according to the historiography, Labour's electoral decline and loss of office in 1949 are partly linked to National's acceptance of the welfare state but mainly to the loss of support from urban non-manual and small farmers, and thus to the breakdown of the coalition of voters it had built in the mid-1930s.¹²

This summary of the historiography, it should be added, excludes discussion of the many other explanations that have been advanced for the outcomes of particular elections and for trends in Labour's share of the vote. It also glosses over certain differences in interpretation. In an interesting parallel with the big debate that has gone on amongst British labour historians, the most important of these differences is whether the working-class shift to the left was greater during World War I than in the decade or so preceding it.¹³ But generally such differences in interpretation are not over the structure of cleavage. On this matter, there is virtual consensus, except as mentioned above, about whether Labour was winning most working-class votes in the 1920s.

To test these claims the article will analyse the relationship between the distribution of manuals and Labour's share of the vote at street level for 10 leading provincial towns in 11 general elections over the study-period. Special attention will be paid to a possible source of cleavage within the working class not mentioned in the New Zealand historiography and of considerable relevance to international research on the electoral history of the left.

Methodology

(a) The Cases

The ten towns selected for the study were New Zealand's second tier, ranking below the four main centres of Auckland, Wellington, Christchurch and Dunedin. They were selected as the cases partly because they were regarded by contemporaries as a unitary group, partly because each constituted a single electorate, partly because they are small enough to be manageable for research purposes, and partly because they embraced New Zealand's limited range of diversity. They were widely distributed across the country's regions and its two main islands. They also varied considerably in their principal economic traits and functions. Thus seven were port towns (Napier, New Plymouth, Wanganui, Gisborne, Nelson, Timaru, and Invercargill) while three were not (Hamilton, Hastings, Palmerston North); two were relatively new and expanding rapidly (Hamilton and Palmerston North), others were older and relatively stable (Invercargill and Nelson); three were large railway centres (Hamilton, Wanganui, and Palmerston North) the others were not; the proportion of the workforce engaged in government administration in New Plymouth was double that of Timaru while Hastings had an exceptionally large seasonal labour force; and so on. Despite their diversity, however, none was a one-industry town or diverged radically in its economic and social characteristics from the others.

The available evidence indicates that the cleavage structure of the ten towns was more like that for the four main centres than it was for New Zealand in its entirety. The towns had much smaller proportions of farmers and rural workers than New Zealand as a whole. The level of mean personal incomes across the towns was almost the same as that for the four main centres but considerably higher than it was for the rest of New Zealand.¹⁴ In addition, the proportions of males in each of the six employment status categories for the towns were closer to their counterparts in the four main centres (despite their larger populations) than to those for New Zealand as a whole (which includes the rural areas and all the smaller towns).¹⁵ However, the relative size of the male manual sector (unskilled/semi-skilled and skilled combined) in the male workforce in the ten towns, the four main centres, and New Zealand as a whole differed little (all being around 60 per cent), and their proportions in the unskilled/semi-skilled sub-category were also very similar (around 32–35 per cent).¹⁶

As far as Labour's share of the vote is concerned, Table I shows Labour performance in the towns was remarkably comparable to its national performance in each year over the study period, except in 1911 and 1914. The two series are never more than a couple of percentage points apart during the period of Labour's electoral stagnation of the 1920s; both rise in 1931 and 1935 although a small gap of about three points develops between the two; both series peak in 1938 at virtually the same level; and subsequently both decline together, the New Zealand series falling rather faster than that for the towns.¹⁷ Contrary to what might be expected, there is no evidence that the ten towns lagged behind the four main centres in the development of class consciousness. In the seats it contested in the general elections for 1911 and 1914, the left received almost the same proportion of the vote on average in the ten towns as it did in the four main centres (30.17 versus 30.89 per cent); and although the proportion varied substantially between the towns, the same is true for the four main centres.¹⁸

Table I: Percentage of total vote won by Labour Party in New Zealand compared with the percentage of total vote it won in the ten towns taken as a whole for general elections in study period, 1911–51

Year/Place	1911	1914	1919	1922	1925	1928	1931	1935	1938	1946	1951
New Zealand	8.8	9.6	24.2	24.5	27.5	26.9	35.0	47.4	55.9	51.3	45.8
Ten Towns	16	7	25	25	25	26	39	44	57	54	49

The percentages are based on all 10 towns (inclusive of rural polling booths), and on all European electorates for New Zealand up to 1943 but including the Maori electorates thereafter. The 1911 and 1914 data are for precursors of Labour Party. Town percentages are rounded. The 1911 data are for first ballot only.

New Zealand-wide data from S. Levine, The New Zealand Political System: Politics in a Small Society, George Allen & Unwin, Sydney, 1979, pp.191–95. Town data from N.D. Stevens, 'Labour Candidates for the New Zealand House of Representatives, 1890–1916', Political Science vol. ix, 1957, pp. 61–71; Lyttelton Times; and Press.

(b) Primary Sources and Social Categories

To reconstruct the social composition of the streets in the towns, the study used the raw data in Henry Wise's New Zealand Post Office Directory. Compiled annually over nearly all the study-period, the directory listed the name, address and occupation of every household head in order of appearance on each side of every street in each town. The study gave an occupational code to each household head in every town for each of the 11 years, then aggregated the heads belonging to each category by street and by year, dividing long streets into sections treated as separate units. Altogether the study assessed over 400,000 household heads (although not all these heads changed in consecutive years). The average number of streets/street sections for all the towns went from 149 in 1911 to over 270 in 1951, with an average of 16 households per street/street section in 1911 and 20.5 in 1951. The coding system had seven categories:

unskilled/semiskilled manual male household heads;
skilled manual male household heads (exclusive of employers and self-employed) and supervisory workers;
lower white collar male household heads;
higher white collar male heads;
all female household heads;
male farmer household heads;
male household heads given no occupation (the 'undescribed').

Categories i–iv are defined as economic classes in Weber's sense.¹⁹ In other words, we follow Weber in assuming that wealth, education and skill determine the market power of individuals, that persons with similar levels of market power belong to the same economic class, and that market power is a function of occupation and determines life chances. We certainly do not deny the possibility, however, that the four economic classes were also social classes experiencing different degrees of what Anthony Giddens, the neo-Weberian, has called 'structuration'. That is to say, we do not exclude the possibility that they were classes for themselves at different stages of development according to the extent the members of each occupied segregated neighbourhoods, engaged in common consumption patterns, and exercised authority in the work-place.²⁰

This Weberian concept of the class structure is the same as that adopted by two leading and very large research programs of historical research on mobility and class structure in New Zealand.²¹ In consequence, the allocation of occupations across the four class categories broadly follows their practices.²² In particular we, like them, made every effort when allocating skilled workers to distinguish those who were employees from those who were self-employed and employers, placing the former in category (ii) and the latter in category (iii). The distinction was greatly assisted by the procedure in Wise's directory to denote self-employed household heads (tradesmen and otherwise) with the initials 'pr' (proprietor) and to give the name of the business in square brackets for managers and owners of large businesses. There is no guarantee that Wise's did this with absolute consistency but there are good reasons to believe that the amount of inconsistency would not have been large.²³

Unlike the research programs on mobility mentioned above, however, we used fewer class categories, aggregating several of theirs (for example, lumping the semi-skilled manual into the unskilled manual). A four class schema was preferred to a larger one partly because the information implied by many occupational labels is inadequate to allow fine distinctions to be made, partly because elaborate schemas tend to generate cells which are too small for analytical purposes, and partly because the additional information gained from generating data for an elaborate schema would have been less than the transaction costs.

The schema also varies from the mobility studies in its treatment of farmers, women and the undescribed (categories v–vii). Farmers are generally left out of analyses of urban-class structure but the study chose not to follow suit since they constituted a surprisingly large minority of household heads in the early years (about 12 per cent in some towns). Since farmers must have varied greatly in market power, they would have belonged to many economic classes in the Weberian sense; but they all had to be lumped together in a single category since the variations in their market power are impossible to determine from the limited information provided by their occupational labels.

Male household heads without occupation are also commonly omitted in analyses of class structures but they too were allotted a category of their own since they grew strongly from about six per cent of total household heads by the end of the study period to double that by the end. Close examination indicates that to a substantial extent they were made up of elderly retirees on state pensions.

Recent practice in historical research on mobility and occupational structure places all women in paid work into a class schema of their own. The study instead put all female household heads into a single category, irrespective of their occupations. The study did not follow the convention partly for a practical reason: the directories listed the occupations of only a minority of female household heads, leaving all the others blank, and it is impossible to make up the information from registers of electors since by law female voters during this period were allowed only to designate themselves by marital status. The study also, however, failed to follow the convention for a good historical reason. Most women did not belong to economic classes in their own right in this era for they were the subject of a great deal of employment discrimination which rendered them dependent on the market power of their husbands and fathers. Thus their rates of overall participation in the paid workforce were exceptionally low, traditional male occupations were closed against them over the period (even during both wars), and their opportunities for job promotion, security of employment, pay rates, and access to workforce pension and state unemployment schemes were all much inferior to men's. There was also considerable employment discrimination against women in other countries at this time of course; but in New Zealand it seems to have gone further given that the female rate of participation in the New Zealand paid workforce was lower than that in Australia, the United States and Britain, a reflection perhaps of the exceptional success the New Zealand state enjoyed in implementing its ideological commitment to the male breadwinner wage.²⁴ Ironically, the life chances of some women were not dependent on the market power of their fathers and husbands but on pensions from the state which were intended to keep them out of the workforce and hence from competing with men for jobs. The increasing provision of state pensions for women and its unintended effect in making a growing proportion of women economically independent of their male relations — in conjunction with the effect of two world wars - is reflected in the changing composition of the household heads in the towns: women went from 12 per cent of all heads in all the towns in 1911 to 18 per cent by 1951.

Without question, the directories are biased against unskilled manuals, the class least able to possess their own households. Yet the extent of unskilled under-representation is not likely to have been substantial or to have seriously distorted the class composition of the streets/street sections in each town. There is little difference between the percentages of male unskilled manual household heads in the towns taken together and the percentages for New Zealand, as indicated earlier. Moreover, in one case taken at random — Gisborne in 1935 - the percentage of all male household heads in each of the four class categories (i–iv) according to the directories is little different from the percentage of total males in each of the matching class categories as given in the electoral rolls.²⁵

To be sure, the directories do not have a good reputation for accuracy in the New Zealand historical community.²⁶ Yet their principal sources of alleged error - their misspelling and failure to keep up to date with the changes in the personnel of individual household heads - do not seriously impinge on this study because they would have had minimal effect on the class composition of streets/street sections. Also if the listings had contained a high rate of error, it should have produced major instability in the calculated data from year to year. But there is a remarkable degree of consistency in the percentages of households belonging to each of the seven categories in each town between immediately adjoining years.²⁷

The raw data for the distribution of party vote came from official returns of individual polling booths within the boundaries of each town in each election year. Polling booths were not assigned to wards (as they were in some countries) since there was no ward system, yet in every town a core were always positioned in the same inner-town sites and major suburban centres. The study excluded rural polling booths and their votes in the areas surrounding many of the towns, leaving each town on average with about nine booths at the beginning of the period and roughly 19 at the end.

(c) Technique

The estimation of Labour's ability to win working-class votes involved three procedures. The unit of analysis was the street/street section in every town for each year. First, the total number of working-class household heads (categories i and ii combined) as a percentage of all household heads in the seven categories was calculated for every street/street section in every town in each year. Second, estimates were made of the distribution of votes at street level. Working on the assumption that in the era before the motor car most but not all people voted close to home, the estimates were derived by taking all the votes for each party at every polling booth in each town for each year, and allocating them to every street/street section in the town, but weighting them in inverse proportion to the distance of the booth from the street/street section. They were also adjusted according to the number of households in a street as a proportion of total households in the town. Once the votes from all polling booths were allocated to every street/street section and appropriately weighted, they were then summed, and the votes for each party in each street /street section converted into a percentage of total pooled vote for each party for that street/street section. Lastly, the estimated percentage of votes for each party was correlated with the percentage of working-class household heads across all the streets/street sections for every town for every year.

The resulting correlation values measure the tendency for the Labour Party's voting support to be class specific. Thus a low correlation between the estimated percentage of the Labour vote and the percentage of unskilled/semi-skilled manuals implies that a disproportionately small number of Labour's votes came from the unskilled/semi-skilled and that most of it came from the other six categories. A very high positive correlation implies by contrast that a disproportionately large number of Labour's votes came from that particular category relative to the other six categories. For its part, a correlation close to zero implies that the number of votes coming from a particular category was neither disproportionately high nor low but proportionate to the relative size of the category in the wider household population. The correlations do not, of course, tell us anything about the actual number of people in a particular class category who voted for Labour. But a judgement about this can be made by comparing the strength of a correlation on the one side and the proportion of the total vote going to Labour on the other.

There are good reasons for believing that the technique for estimating the percentage of party vote at street level yields reliable results. The first is that a series of methods for weighting the allocation for votes from polling booth to street were tried and all produced very similar results.²⁸

The second is that if the assumptions underlying the method had been greatly at variance with where voters actually voted on polling day, the anomaly should have manifested itself by throwing up some inherently improbable results. For example, the correlation values would variously would have been consistently zero, extremely high, had no pattern to them at all or contained contradictory patterns, moved in a completely unexpected direction, and so on. By this test none of the values is inherently improbable, though some are surprising.

The third reason is that the overall pattern of correlations fits modern survey data on voting preferences by class in societies where these preferences are dominated by class cleavage. In such surveys, the proportion of respondents favouring the left or right is graded according to class along a continuum. Thus in a four-class scale, the proportion in each class favouring the right is smallest for the unskilled manual, rather higher for skilled manual, larger still for lower middle class, and largest for higher managers and professionals.²⁹ The correlation patterns for the ten towns New Zealand generally have the same shape as the voting surveys. Thus the correlation of estimated percentage of the vote for Reform/National and percentage of working-class household heads across the streets in nearly every town in most years tends to move in a relatively strong negative direction for the unskilled/semi-skilled, in a weaker negative direction for the skilled, a relatively weak positive direction for the lower white collar and a stronger positive direction for the upper white collar. It should be noted that for technical reasons given in Appendix II, all correlations involving estimated party vote are lower than they would be had disaggregated voting data been used.

The obvious advantage of the study's novel research design is that it provides reliable and precise data about the social composition of party votes which would otherwise be unknown given the operation of the secret ballot and the absence in this era of opinion surveys on the party preferences and class background of voters. Another advantage of the design is that it has a very strong comparative dimension generating as it does systematic data for ten cases for each of 11 points in time. A further advantage is that the study covers an unusually large sequence of general elections at abnormally frequent intervals over a 50-year period, and as a consequence has an exceptional capacity to map change over time in the short, medium and long term. In addition, taking the street and street section as the unit of analysis produces a large number of observations for each town for each year, thus enhancing the statistical robustness of the analysis. Also the research design produces a unique and rich data set on the class geography of each town over the study period. Finally, the use of polling booth level voting data in conjunction with social data at the level of individual households minimises a technical problem known to statisticians as the ecological fallacy (it arises when individual behaviour is inferred from aggregate behaviour). Mathematical discussion of the problem and demonstration of how the research design has minimised it are spelt out in Appendix II.

Results of the Analysis

The intention of the article, it will be remembered, is to test the claims in the historiography that Labour's electoral fortunes were generally tied to its ability to broaden its appeal outside the working class in the leading towns and exceptional rural seats where there were abnormal concentrations of unionised workers (miners and timber workers). Thus Labour (and its precursors) did badly from 1911 to 1931 because it failed to appeal to other fractions in New Zealand's cleavage structure; it performed brilliantly in 1935 and 1938 because it succeeded in forging a voter coalition consisting of its traditional manual supporters, small farmers and urban non-manuals; and it declined after 1938 largely because of diminishing support from the two cleavage fractions it had won in the mid-1930s: small farmers and urban non-manuals. The historiography, it should be emphasised, not only asserts that Labour's electoral fortunes were generally related to its ability to appeal across the cleavage divisions. It also, with a partial exception, takes for granted that there were no cleavages within the ranks of manuals. The partial exception was the so-called rift between the (moderate) craft unions and the (radical) industrial unions, 1908–12, during the first phase; but there is general agreement that the rift was largely bridged at the institutional level with the Unity Congress in 1913, and the possible social and cultural basis of this rift has never been systematically researched, let alone on a comparative basis.

The results of the analysis do not support the claims. Table A in Appendix I provides the data. The first proposition is that in the leading towns from 1911 to the Slump of the 1930s, Labour and its predecessors received the vast bulk of its voter support from the working class, though in much larger numbers after 1919 than in 1911 and 1914. Left candidates contested five of the town seats in 1911 and two in 1914. If the historiography is right and the only or virtually the only votes they received were from manuals, we would expect the estimated percentage of left vote to have a correlation of, at least say, +0.7 if not more with the percentage of both manuals combined (categories i +ii). It is not an extremely high value in itself but it makes some allowance for the effect of the model in smoothing the variations in the Labour vote across streets/street sections and thus of pulling the correlations downwards. In fact, the correlation values, which vary between -0.045 and +0.378, are lower than expected. For the general elections from 1919 to 1928, although the correlations tend to rise in all the towns, they nonetheless fail to go beyond the +0.3 level in New Plymouth, Palmerston North, Invercargill and Hastings; and never exceed the +0.4 to +0.5 range in the six other towns. In other words, contrary to what the historiography says about the 1920s, in the ten towns at least the vast majority of the Labour Party's votes did not come from the urban working class and the vast majority of the urban working class did not vote for Labour.

Turning to the 1930s, the correlations again show a pattern which is the inverse of that claimed by the historiography. Whereas the convention says that Labour broadened its sources of mass support as it became more popular (more non-manuals relative to manuals) the rising overall level of the Labour vote and correlations show the opposite. In the ten towns, as the percentage of the Labour vote shot up, the proportion of the votes Labour received from manuals must have continued to increase at a faster rate than from non-manuals because in every town both the levels and the correlations rose. Indeed they were higher in these years in almost every town than in the rest of study-period. These statistics do not exclude the possibility that Labour won an unparalleled number of non-manual votes in the 1930s; but they imply that Labour's increase in manual votes outstripped its increase in non-manual votes. Labour's surge of popularity in the ten towns in the 1930s came more from the working class than from the other social categories.

What about the phases of Labour's electoral decline, from the late 1930s to its landslide defeat in 1951? The pattern here too is the inverse of that postulated by the historiography. According to the historiography, the proportions of the Labour vote that was working class in addition to the correlations between percentage of Labour vote and the percentage of total household heads who were working class should have gone up. What we find, in fact, is that both the levels and the correlations tend to decline, implying that the size of Labour's working-class vote went down as well as becoming less aligned.

Contrary to the assumption in the historiography, then, Labour was not a working-class-specific party in terms of its voting support for most of the study-period. A reasonably strong alignment of working class and Labour vote did not occur until the 1930s and decreased after the 1930s. On top of this, the gap between Labour's share of the vote and the relative size of the working class in the electorate before the 1930s clearly indicates that perhaps most of the working class did not vote Labour at that point. Without doubt many factors explain Labour's difficulties in attracting working class votes. The rest of the article will explore just one; namely, the party's persistent failure to bridge a particular cleavage within the working class itself, between skilled and unskilled/semi-skilled manual workers.

To investigate this possibility we will disaggregate the two manual-class categories and re-do the correlations but analysing separately the percentage belonging to each of the manual categories with the estimated percentage of the total votes won by Labour across all the streets/street sections of each town in each year.

Dealing first with the correlations between percentage skilled and percentage vote Labour, we find that in most of the towns the correlation co-efficients tend to rise over the study-period. Table B in Appendix l gives the data. The exceptions are Hamilton where they fall consistently, New Plymouth where they are consistently negligible, and Hastings where they collapse after 1938. (Nelson is a special case since the seat was not contested by Labour in most elections.) It is notable, however, that despite their tendency to rise, the correlations are very low compared with those for the two manual categories combined. Indeed, over the whole period most of the correlations for the skilled are close to zero and seldom rise beyond about +0.25 in six towns (New Plymouth, Palmerston North, Hamilton, Hastings, Invercargill, and Timaru), and are not much higher in Gisborne. Only in two towns — Wanganui and Napier — are they normally more than 0.2; but even in these two cases they never exceed +0.4, which is well beneath the peak levels in the same towns for the correlations between the percentage Labour vote and the percentage of the two manual categories combined.

It would be an exaggeration to say that the weak support from skilled manuals determined Labour's electoral fortunes. They were far from a majority of all household heads and did not outnumber unskilled/semi-skilled manual household heads in the working class. Yet they were not a negligible fraction of either the entire electorate or the working-class segment of the entire electorate. As Table II indicates, they constituted about a fifth of total household heads and about 40 per cent of all working-class household heads in all towns taken as a whole over the study period. Hence their weak support at least partly explains Labour's chronic electoral failures up to the 1930s; it also contributed little to the surge in Labour's electoral popularity in the 1930s; and it did nothing to help prevent Labour from declining from the 1930s.

Table II: Percentage of total household heads who were skilled manuals compared with percentage who were unskilled/semi-skilled manuals in all the towns taken as a whole over the study period, 1911–51

Year/category of household head	Skilled as % of all household heads	Unskilled/semi-skilled as % of all household heads
1911	21.7	26.5
1914	21.7	25.0
1919	20.3	24.9
1922	19.8	26.4
1925	20.5	24.2
1928	20.6	25.1
1931	20.4	24.6
1935	19.5	25.9
1938	19.1	25.8
1946	17.1	23.9
1951	17.6	22.5

Unskilled/semi-skilled is category i; skilled is category ii. The base is categories i–vii. For sources see text.

The picture is quite different when we look at the correlations between percentage Labour vote and the percentage of unskilled/semi-skilled household heads. These can be seen in Table C of Appendix I. Apart from Hastings, the unskilled correlations are mostly above 0.3, with Palmerston North, Invercargill and Gisborne tending to have the lowest (fluctuating primarily or mainly between the +0.3 and +0.4 range), and Napier, Hamilton and Timaru with the highest (hovering generally between +0.4 and +0.5, sometimes reaching +0.6). Thus, in almost every town, with the exception of Hastings, the unskilled/semi-skilled correlations are appreciably higher than the skilled. Although in general they are at least one 0.1 point of a co-efficient stronger, in three towns they are many times more powerful. In New Plymouth they are twice as strong in most years, in Hamilton from 1928 they are between two to six times stronger, and in Timaru from 1919 they usually more than three times stronger.

But it is not only in comparison to the levels of the correlations between percentage skilled and percentage Labour vote that we can say that the corresponding unskilled/semi-skilled correlations are high. They are also high compared to the correlation values for percentage Labour vote and the variable for the percentage of the two manual categories combined (Table A in Appendix I). Although not larger than these, they are usually less than 0.1 points smaller, and tend to behave like them over time, rising slowly over the 1920s, peaking in the 1930s, and falling thereafter. In sum, the alignment of the Labour Party (and its predecessors) with the unskilled/semi-skilled fraction of the working class is consistently much stronger than it is with the skilled fraction over the whole study period. To the extent to which Labour is a class-specific party, it is a party predominantly of the lower echelons of the working class not of the working class in general.

The single exception to this pattern is Hastings: in contrast to all the other centres, its unskilled/semi-skilled correlations are very close to zero and are also lower than their skilled counterparts. But Hastings is a maverick in almost everything: of all the towns over the whole period, it seems to have had the largest seasonal labour force, and the lowest correlations between percentage vote Labour and percentage of the aggregate of skilled and unskilled/semi-skilled (which seldom go beyond +0.2).

It would be going too far to say that the correlation values for percentage vote Labour and the unskilled/semi-skilled variable are radically different from those for percentage vote Labour and the skilled variable. They could only be considered radically different — different in kind as opposed to different in degree — if they moved in diverging directions, with one set being generally negative and the other generally positive. Even so, the differences between the levels of each set are sufficiently large in degree and systematic to indicate that the working class was not a homogenous body of voters but one divided into two fractions. The attenuation of all these correlations by virtue of modelling error does not alter the conclusion. To put this another way, the two manual classes have a sufficiently different voting relationship with the Labour Party to suggest they were not just different classes in an economic sense but in a social sense as well: the people in them belonged to dissimilar sub-cultures.

The differences, it must be emphasized, cannot be attributed to the notion that the skilled manual category, in contrast to the unskilled/semi-skilled, contained substantial portions of small proprietors, whether self-employed or small employers. As explained earlier, the number of such petit-bourgeois elements in category (ii) would have been minimal.

Unfortunately, we know very little about what traits these sub-cultures might have possessed since the bulk of the research — which has been quite extensive - on the history of the labour movement in twentieth century New Zealand has focussed on leaders, imported formal ideologies and institutions.³⁰ Thus although the split during most of the period between the industrial unions of the unskilled and the craft unions of the skilled is a prominent theme in the literature, the latter tells us very little about how the split might have reflected contrasting life-styles and norms of wage earners at the local community level. Instead the literature has looked at the split from the top down, as a consequence of personality clashes between union leaders, struggles for power and of the divergent effects of the state wage regulation on the economic interests of the skilled and the unskilled. In addition, although the literature has dealt with the conflict of ideas between the pragmatism of the craft unions and the syndicalist doctrines frequently embraced by some of the industrial unions, this does not help us explain why Labour appealed more to the unskilled than the skilled because according to the historiography the syndicalist unions were usually anti-Labour whereas the craft unions supported the Party.

The exceptions to this preoccupation with leaders, ideologies and institutions include Libby Plumridge's study of the social context for the rise of the labour movement in early twentieth century Christchurch, and Len Richardson's history of the coal miner's union which had much to say about the union's relationship to the miners' occupational and local communities.³¹

The key exceptions, however, are Erik Olssen's ground-breaking studies of the rise of syndicalism amongst the unskilled (the 'Red Feds') before the First World War, and of an artisan-dominated working-class community called Caversham in Dunedin (one of the four main centres) from the 1880s to the 1920s.³² He claims that what helped to make the unskilled receptive to syndicalism was that their 'rough' normative code alienated them from the New Zealand mainstream. In his later work on Caversham he found much evidence of tensions between its skilled workers and unskilled over religion, liquor, respectability and at work where the skilled often exercised authority over the unskilled. But while both sets of studies have some potential to explain the cleavage between skilled and unskilled, they are mutually inconsistent on one crucial point. In his 'Red Fed' studies Olssen claims that the unionisation of the unskilled workers provided the electoral foundation for the Labour Party and its precursors; in his later Caversham work he says it was the skilled. In his Caversham studies, furthermore, he argues that despite the tensions there was no cleavage between the skilled and unskilled since their strong tendency to live cheek-by-jowl in the same neighbourhoods led the unskilled to assimilate the norms of the skilled, producing an egalitarian community, a single status group.³³ Finally, if, as Olssen suggests, the rest of urban working class New Zealand followed the Caversham egalitarian model, this does not explain why in nearly all the ten towns Labour drew proportionately more votes from the unskilled/semi-skilled than the skilled.³⁴

The best clue about what might have underpinned working-class cleavage in the towns comes from an analysis of their social geography. In none of them did the skilled and unskilled have a strong tendency to co-reside in the same neighbourhoods, a necessary condition for the egalitarian community postulated by the Caversham model. Instead in all the towns, the relationship between the spatial distribution of the skilled and of the unskilled/semi-skilled was closer to what would be expected in the 'Aristocracy of Labour' model put forward by Gray and Crossick for nineteenth century Britain.³⁵ In every town in every year the correlations between the percentage of skilled and the percentage of unskilled/semi-skilled across the streets/street sections were all virtually zero or generally moved in a negative direction (except for Napier). Table D in Appendix I gives the results of the correlation analysis, which is completely unaffected by the modelling of voter behaviour and reflects correlations between actual variables (rather than the correlation between an actual and a modeled variable to which attenuation applies). The conclusion is that in each town in each election year, there was a balance between residential areas with a high concentration of skilled but a low one of unskilled/semi-skilled (or the other way round), and areas where there were high or low concentrations of each.

What the zero correlations between proportion of unskilled/semi-skilled and the proportion of skilled suggest is that a not insubstantial portion of the skilled considered themselves to be of higher status than the unskilled/semi-skilled, and took advantage of their superior market power to live in the same residential areas as the non-manuals, distancing themselves from the unskilled. The weaker identification of the skilled with the working class, hence, is revealed both in the greater difficulty Labour had in winning their votes and in the comparative reluctance of the skilled to reside in the same areas as the unskilled. It should be stressed that not all the skilled identified with the non-manuals. Like the working class as a whole, they were not a homogenous but a heterogenous class in a social sense. Indeed, more must have had a sense of identity with the working class than otherwise given that the correlations between percentage vote Labour and percentage of skilled tended to be mildly positive; and given that the correlations between the two manual categories at street level of each town in each year were close to zero. Had the balance been the other way round, pairings in the sets of correlations with Labour vote (for each year) would have moved in a moderately or strongly negative direction.

Unfortunately, there are no comparable international data that allow the New Zealand data to be put in perspective. Still, there is another — internal - standard of comparison which can be used. Although there is no space to discuss the subject here, the correlation analysis shows that the tendency to reside together in the same street/street section was generally weaker for the two manual categories of household heads than it was the two white collar categories of household heads. In addition, the correlations between the percentage of skilled manual category with the estimated percentage of the vote for Labour tended to be much lower than the corresponding correlations between the percentage of lower white collar heads and the estimated percentage vote for the mainstream right wing party. In other words, in relation to the level of cohesion between the two white-collar classes in the ten towns over the study period, the degree of solidarity between the two working classes was low.

Conclusion

This paper has several implications for studies of working-class 'homogenisation' in other advanced societies in the late nineteenth and early twentieth centuries.

There is a tendency in many of these studies to assume, as their New Zealand counterparts have done, that at some point by the early twentieth century the historic cleavage between the skilled and unskilled/semi-skilled disappeared or greatly diminished as class consciousness grew. It is difficult to reconcile these claims with what has been found with nearly all the ten towns. Their cleavage persisted over the whole period, as we have seen, even though conditions in New Zealand were far less favourable to cleavage than they were in almost every advanced society. In other societies, economic disparities between the skilled and unskilled within the working class were frequently overlaid and reinforced by ethnic, religious, linguistic and other cultural divisions. But the sources of such divisions did not exist or barely existed in New Zealand. For one thing, it did not have a legacy of imported slave, convict and indentured labour. For another, because it was one of the newest of the new societies, its inherited stock of norms about social distinctions were more egalitarian than they were in nearly every other society in the old and new world. Furthermore, unlike some new world societies, its indigenous people, the Maori, were never forced to work in mines or plantations; and although predominantly operating as casual labourers by the turn of the century, the number of Maori unskilled workers and the amount of contact they had with the Europeans were far too small to affect adversely the status of the unskilled as a whole until after the second world war. On top of this, by comparison with Britain, Australia and the United States, New Zealand's unskilled workforce was less tainted by its associations with Irish immigrants. By the turn of the century, New Zealand had smaller proportions of people of Irish parentage and extraction than Australian and the United States; in contrast to these societies and Britain, its Irish Catholic population did not form ghettoes; and, in contrast to them, by the early 1920s class differences between people of Irish Catholic background and New Zealanders as a whole had disappeared. Certainly in New Zealand, as in other advanced capitalist societies, there were strong historical disparities in the market position between the skilled and unskilled/semi-skilled. Yet over the study period these market differences were diminishing as a consequence of the state's tight control over industrial relations after 1894, and of the welfare and full employment programs established by the first Labour Government from 1935.

The point, in other words, is that if working-class cleavage in New Zealand persisted over the whole study period in the face of such weak countervailing pressures, then how is it that the cleavage affecting the working classes in other societies supposedly disappeared or greatly declined when their countervailing pressures were so much stronger, or at least no weaker?

Another implication of this study for the labour history of other societies is that it complicates the explanations which are advanced for the homogenisation of the working class and the diminishing gap between the skilled and the unskilled that supposedly took place by the early twentieth century. In Britain, for example, these explanations variously focus on the effect of mechanisation in degrading skill and eliminating certain trades, on the rise of new occupations and boom towns, on successful attempts by employers to gain control over the work process, on the effect of World War I in levelling the market situations of the skilled and the unskilled, the replacement of apprenticeships with other forms of training and so on.³⁶ Although some of these factors might have affected the ten towns in some degree or another, what this study shows is that none of them changed the comparatively low propensity of the skilled to live in the same streets as the unskilled/semi-skilled, since this tendency was basically static over the whole study period. Moreover, while some of these factors might have stimulated both the unskilled/semi-skilled and the skilled to identify with the Labour Party, a stronger stimulus seems to have been the common experience of the economic slump and the benefits of Labour's interventionist legislation in the 1930s — and even then the effect was stronger on the unskilled and semi-skilled than on the skilled.

The final implication of this study is methodological. The study has used level of vote and correlations to test for cleavage between the skilled and the unskilled/semi-skilled — the extent to which the skilled and unskilled/semi-skilled voted for the same mainstream left party and the extent to which they lived in the same streets. Other statistical measures are possible but have not been found necessary here. Whether other studies with different data, different variables and different levels of aggregations will find these statistics equally useful as well as theoretically adequate must remain a matter for further debate.

Endnotes

^* This article is part of a large scale project on the history of working class voting behaviour in New Zealand financed by a grant administered by the Royal Society of New Zealand. The authors also wish to thank Erik Olssen, Melanie Nolan, and Graeme Dunstall for their helpful criticisms on earlier drafts of the paper. The paper has been peer reviewed for Labour History by two anonymous referees.

^1. Moshe Maor, Political Parties and Party Systems: Comparative Approaches and the British Experience, Routledge, London, 1997, p. 19ff.

^2. Seymour Lipset and Stein Rokkan, 'Cleavage structures, party systems, and voter alignments: an introduction', in Seymour Lipset and Stein Rokkan (eds), Party Systems and Voter Alignments: Cross-National Perspectives, Free Press, New York, 1967, pp. 1–64.

^3. See, for example, David Butler, and Donald Stokes, Political Change in Britain: the Evolution of Election Choices, Macmillan, London, 2^nd edn, 1974, chs 4, 7, 9; Anthony Heath et al, Understanding Political Change: the British Voter, 1964–1987, Pergamon Press, Oxford, 1991.

^4. Len Richardson, 'Parties and political change', in Geoff Rice (ed.), The Oxford History of New Zealand, Oxford University Press Auckland, 2^nd edn, 1992, pp. 201–29; David Hamer, The New Zealand Liberals: the Years of Power, 1891–1912, Auckland University Press, Auckland, 1988, pp. 150, 184; Robert Chapman, The Political Scene, 1919–1931, Heinemann Educational Books, Auckland, 1969, pp. 6–8; Robert Chapman, William Jackson and Austin Mitchell, New Zealand Politics in Action: the 1960 General Election, Oxford University Press, London, 1962, 238ff, 252ff; A.D. Robinson, 'Class voting in New Zealand', in Lipset and Rokkan (eds), Party Systems and Voter Alignments: Cross-National Perspectives, Free Press, New York, 1967, pp. 95–114; Jack Vowles and Peter Aimer, Voter's Vengeance: the 1990 Election in New Zealand and the Fate of the Fourth Labour Government, Auckland University Press, Auckland, 1993, pp. 27–8.

^5. Chapman, The Political Scene; Michael Bassett, Three Party Politics in New Zealand 1911–1931, Historical Publications, Auckland, 1982, p. 40.

^6. R.M. Chapman, The significance of the 1928 General Election: A study in certain trends in New Zealand politics in the Nineteen-Twenties, MA thesis, University of Auckland, 1948; Chapman, Jackson and Mitchell, Politics in Action; Chapman, The Political Scene; R.M. Chapman, 'From Labour to National', in Rice (ed.), Oxford History of New Zealand. His other articles on elections, past and contemporary, have been collected in New Zealand Politics and Social Patterns: Selected Works, E.M. McLeay (ed.), Wellington, Victoria University Press, 1999. The Introduction by McLeay also gives some idea about his influence on the research of a whole generation of politics post-graduate students.

^7. R. Miller, 'Labour', in R. Miller (ed.), New Zealand Government and Politics, Oxford University Press, Auckland, 2001, p. 236. Miller cites Chapman as his authority in the passage quoted. Vowles and Aimer, Voter's Vengeance, pp. 28–29 do not go as far as Miller in saying that Labour could not win without forging a coalition between inner-city working class voters and rural/small town voters. But like Miller they too cite Chapman as their authority on the historical context.

^8. Erik Olssen, 'The origins of the Labour Party: a reconsideration', New Zealand Journal of History, vol. 21, 1987, pp. 83–6; Barry Gustafson, Labour's Path to Political Independence: the Origins and Establishment of the New Zealand Labour Party 1900–19, Auckland University Press, Auckland, 1980, ch. 3.

^9. Erik Olssen and Len Richardson, 'The New Zealand Labour Movement, 1880–1920', in Eric Fry (ed.), Common Cause: Essays in Australian and New Zealand Labour History, Allen & Unwin, Wellington, 1986, pp. 14–5; Hamer, The New Zealand Liberals, p. 185; Gustafson, Labour's Path to Political Independence, chs. 12, 13.

^10. Erik Olssen, 'The New Zealand Labour Movement, 1920–40', in Fry (ed.), Common Cause, 1986, pp. 16–7; Chapman, The Political Scene, pp. 16, 32, 55; Bruce Brown, The Rise of New Zealand Labour, Price Milburn, Wellington, 1962, pp. 41–2, 104–5; Richardson, 'Parties and Political Change', pp. 219, 220, 222; Keith Sinclair, Walter Nash, Auckland University Press, Auckland, 1976, p. 75; cf. Miles Fairburn, 'Why did the Labour Party Fail to Win Office Until 1935?', Political Science, vol. 37, 1985, pp. 101–24. Note that Olssen has recently moved away from the position he held in 1986; see eg E. Olssen and H. James, 'Social Mobility and Class Formation: the Worklife Social Mobility of Men in a New Zealand Suburb, 1902–1928', International Review of Social History, vol. 44, 1999, pp. 419–49.

^11. Brown, The Rise of New Zealand Labour, pp. 180–1; Olssen, 'The New Zealand Labour Movement', p. 22; Chapman, 'From Labour to National', pp. 353–4; Richardson, 'Parties and Political Change', pp. 228–9; Sinclair, Walter Nash, p. 269.

^12. Chapman, 'From Labour to National', pp. 368, 370, 373; Sinclair, Walter Nash, pp. 266–7.

^13. The first serious critique of the view that the great war was the turning point in the raising of working class consciousness and the rise of the left in New Zealand, was by Olssen, 'The Origins of the Labour Party'. Useful summaries of the British debate include G.R. Searle, The Liberal Party: Triumph and Disintegration, 1886–1929, St Martins, New York, 1992; D. Tanner, 'Class Voting and Radical Politics: the Liberal and Labour Parties, 1910–31' in J. Lawrence and M. Taylor (eds), Party, State and Society: Electoral Behaviour in Britain since 1820, Scolar Press, Aldershot, 1997, pp. 106–30; K. Laybourn, 'The Rise of Labour and the Decline of Liberalism: the State of the Debate', History, vol. 80, 1995, pp. 207–26.

^14. In the 1926 census, the first to record average incomes in detail, the mean income for gainfully employed males for New Zealand as a whole was £205 a year, for the four main centres it ranged from £220 to £240, for the ten towns the spread was between £210 to £230, while for all the areas outside the fourteen largest towns the mean was £185. New Zealand Census, 1926, vol. XI, Incomes of the People, p. 3. Mean annual incomes for females in the ten towns were almost identical to those in the four main centres.

^15. Distribution of the male workforce by occupational status in New Zealand, the ten towns, and the four main centres, 1936 (%):

Status/place Employers Self-employed Wage and salary earners Apprentices Unemployed (partly and wholly) Relatives assisting without pay

New Zealand 10.6 12.7 62.7 1.4 10.3 2.1

Four main centres 6.2 7.6 68.3 2.2 15.5 0.3

Ten towns 8.7 9.5 66.6 2.0 13.5 0.5

Data from New Zealand Census, 1936, Occupational Status, iv.
Percentages are rounded so may not add up to 100 per cent.

^16. The 60 per cent figure for total manual workers in New Zealand is for 1926 and 1936, and calculated from David Pearson and David Thorns, Eclipse of Equality: Social Stratification in New Zealand, Allen & Unwin, Sydney, 1983, p. 46; and Paul Meuli, Occupational Change and Bourgeois Proliferation: a Study of New Middle Class Expansion in New Zealand 1896–1926, MA thesis, Victoria University of Wellington, New Zealand, 1977. The 60 per cent figure for the four main centres is based on Caversham (1922) and Christchurch (1921, 1931 and 1938), and calculated from Michael Smith, 'Some Social Characteristics of Christchurch: 1921–1938', in M. Fairburn and E. Olssen (eds), The Quantitative Turn: Studies of New Zealand Society Over the 20^th Century, Otago University Press, Dunedin, forthcoming; and Tom Brooking, D. Martin, D. Thomson and H. James, 'The Ties That Bind: Persistence in a New World Industrial Suburb, 1902–22', Social History, vol. 24, 1999, p. 65. The 32/35 per cent rate of unskilled/semiskilled for New Zealand are from Meuli and those for the four main centres from Smith. The Caversham unskilled/semi-skilled component in the total male work force seems to have been lower than elsewhere in the four main centres probably because Caversham had an abnormally large skilled manual category.

^17. Indeed, New Zealander's leading historian of elections, Chapman, has maintained that the towns acted as the bell-wether for nation-wide voter opinion containing as they did a balance of the sectional interests making up New Zealand's cleavages. See Chapman et al, New Zealand Politics in Action.

^18. The left candidates for the four main centres were those identified by N.D. Stevens, 'Labour candidates for the New Zealand House of Representatives, 1890–1916', Political Science, vol. ix, 1957, pp. 61–71, as New Zealand Labour Party, Socialist Party and Independent Labour candidates in 1911; and Social Democratic Party, Labour Representation Committee and Independent Labour candidates in 1914. There were 20 such candidates in 1911 and 13 in 1914.

^19. Max Weber, From Max Weber: Essays in Sociology, Kegan Paul, Trench, Trubner & Co. Ltd., London, 1947, pp. 181–8.

^20. Anthony Giddens, The Class Structure of the Advanced Societies, Hutchinson, London, 1973, pp. 107–112.

^21. One is the work by David Pearson; see Pearson and Thorns, The Eclipse of Equality; Olssen and James, 'Social mobility and class formation'; Johnsonville: Continuity and Change in a New Zealand Township, George Allen and Unwin, Sydney, 1980; 'Small Town Capitalism and Stratification in New Zealand 1880–1930', New Zealand Journal of History, vol. 14, 1980, pp. 107–31; 'Marriage and Mobility in Wellington 1881–1980, New Zealand Journal of History, vol. 22, 1988, pp. 135–51. The other and larger research program is the Caversham Project; see Erik Olssen, Building the New World: Work, Politics and Society in Caversham 1880s–1920s, Auckland University Press, Auckland, 1995; E. Olssen and M. Hickey, 'Towards an Occupational Classification of New Zealand, 1901–1926', Caversham Project, History Department, University of Otago, NZ, CWP1998; E. Olssen and H. James, 'Social Mobility and Class Formation: the Worklife Social Mobility of Men in a New Zealand Suburb, 1902–1928', International Review of Social History, vol. 44, 1999, pp. 419–49; Erik Olssen and Maureen Hickey, Class and Occupation: The New Zealand Reality, University of Otago Press, Dunedin, forthcoming 2004. Also T. Brooking et al, 'The Ties That Bind', pp. 55–73.

^22. Where there were uncertainties or lack of precedents the decisions were based on the rich data in the 1926 census on median incomes by employment status.

^23. One reason is that for the national male workforce in 1926, the self-employed skilled manual constituted only 9.06 per cent of all skilled manual (self-employed plus employees) which is surprisingly small; see the reworking of the census occupational status data in Meuli, Occupational Change, p. 177. Another reason is that the proportion of household heads in category (ii) of all heads for each town, is remarkably stable from one year to another over the 1911–1951 period, which is not what would be expected if the coding of skilled tradesmen had been highly erratic; see M. Fairburn and S.J. Haslett, 'Stability and Egalitarianism in New Zealand, 1911–1951', in Fairburn and. Olssen (eds), The Quantitative Turn, Table B, Appendix I.

^24. See Melanie Nolan, Breadwinning: New Zealand Women and the State, Canterbury University Press, Christchurch, 2000, p. 307.

^25. The percentage calculated from the rolls for the unskilled/semi-skilled was 36.8, skilled 22.6, lower white collar 27.8, higher white collar 7.5, farmers 3.2, and undescribed 1.8. The corresponding percentages as calculated from Wise's were 35.3, 20.6, 26.6, 10.0, 3.4, and 3.7.

^26. See, for example, Brooking et al, 'The Ties That Bind', pp. 59–60.

^27. Statistical tests for sharp differences threw up only nine suspects out of the total maximum of 770 cases. The tests were based on a multiple regression (ANCOVA) model and on log-linear models for category, town and year. The models measured the degree of change for each cell in relation to the long-term linear trend for the series.

^28. Thus one method weighted the votes in direct inverse proportion to the distance of each booth from the area; another weighted the votes in inverse proportion to the square of the distance of the booth from the area; and a third weighted the votes in inverse proportion to the cube of the distance. It is also notable that the correlations between votes weighted one way and each of the other two ways for every town in every year were nearly all over the +0.9 level.

^29. See for example, the classic study by D. Butler and D. Stokes, Political Change in Britain: the Evolution of Electoral Choice, Macmillan, London, 1974, 2nd edn, ch. 6.

^30. For biographical approaches see, for example, Patrick O'Farrell, Harry Holland: Militant Socialist, Australian National University, Canberra, 1964; Sinclair, Walter Nash. For ideological see, for example, Gustafson, Labour's Path to Political Independence; Patrick O'Farrell, 'Coal and Politics: the Socialist Vanguard', in Phillip May (ed.), Miners and Militants: Politics in Westland, 1865–1918, Canterbury University Press, Christchurch, 1975, pp. 101–27. For institutional approaches see, for example, Herbert Roth, Trade Unions in New Zealand: Past and Present, Reed Education, Wellington, 1973; R.C.J. Stone, 'The Unions and the Arbitration System, 1900–1937', in Robert Chapman and Keith Sinclair (eds), Studies of a Small Democracy: Essays in Honour of Willis Airey, Blackwood and Janet Paul Ltd, Auckland, 1963, pp. 201–20; J. McAloon, 'Radical Christchurch', in J. Cookson and G. Dunstall (eds), Southern Capital, Christchurch: Towards a City Biography, Canterbury University Press, Christchurch, 2000. An excellent discussion of the limitations of these approaches is Olssen, 'The origins of the Labour Party: a reconsideration'; and 'Some reflections about the origins of the "Red" Federation of Labour, 1909–13', in Fry (ed.), Common Cause, 1986, pp. 27–41.

^31. L. Plumridge, 'The Necessary but Not Sufficient Condition: Christchurch Labour and Working Class Culture', New Zealand Journal of History, vol. 19, 1985, pp. 130–50. L. Richardson, Coal, Class and Community: the United Mineworkers of New Zealand, 1880–1960, Auckland University Press, Auckland, 1995.

^32. Olssen, 'Some Reflections About the Origins of the 'Red' Federation of Labour'; Erik Olssen, The Red Feds: Revolutionary Industrial Unionism and the New Zealand Federation of Labour 1908–1914, Oxford University Press, Auckland, 1988; Olssen, Building the New World.

^33. His claim that the skilled and unskilled intermingled in Caversham is consistent with a study of its residential segregation by one of his research team, C. Griffen, 'The New World Working-Class Suburb Revisited: Residential Segregation in Caversham, New Zealand', Journal of Urban History, vol. 27, 2001, pp. 420–44.

^34. Olssen, Building the New World, p. 261.

^35. Robert Gray, The Labour Aristocracy in Victorian Edinburgh, Clarendon, Oxford, 1976; Geoffrey Crossick, An Artisan Elite in Victorian Society: Kentish London 1840–1880, Croom Helm, London, 1978.

^36. See for example, M. Childs, 'Labour Grows Up: the Electoral System, Political Generations, and British Politics 1890–1929', Twentieth Century British History, vol. 6, 1995; Duncan Tanner, Political Change and the Labour Party, Cambridge University Press, Cambridge, 1990, pp. 352–61; Charles More, Skill and the English Working Class 1870–1914, Croom Helm, London, 1980, 230ff; Searle, The Liberal Party, pp. 136ff.

Appendix I

Table A: Pearson correlations between estimated percentage of vote for Labour and percentage of total manuals (skilled and unskilled/semi-skilled) across the streets/street sections of each town over the study period, 1911–51

Year/Town	1911	1914	1919	1922	1925	1928	1931	1935	1938	1946	1951
Hamilton	-	-	-	0.540 ***	0.552 *	0.578 *	0.521 ***	0.600 ***	0.576 ***	0.436 ***	0.462 ***
Gisborne	-	-	0.240 **	0.319 *	0.297 ***	0.429 ***	0.413 ***	0.349 ***	0.473 ***	0.295 ***	0.309 ***
New Plymouth-	-	-	-	-	0.295 ***	0.164 *	0.326 ***	0.332 ***	0.347 ***	0.256 ***	0.397 ***
Palmerston North	-	-0.045	0.229 **	0.258 **	0.116	0.270 ***	0.461 ***	0.423 ***	0.301 ***	0.287 ***	0.380 ***
Wanganui	0.373 ***	0.378 ***	0.387 ***	0.372 ***	0.504 ***	0.534 ***	0.566 ***	0.504 ***	0.521 ***	0.452 ***	0.429 ***
Napier	0.059	-	0.456 ***	0.559 ***	0.500 ***	0.523 ***	0.490 ***	0.611 ***	0.546 ***	0.504 ***	0.563 ***
Hastings	0.022	-	0.206 **	0.157 *	0.130	0.048	0.137 *	0.140 *	0.211 **	0.152 *	0.292 ***
Nelson	-	-	0.300 **	-	0.327 ***	-	-	-	-	0.569 ***	0.505 ***
Timaru	0.350 ***	-	0.312 ***	0.392 ***	0.448 ***	0.558 ***	0.608 ***	0.589 ***	0.602 ***	0.529 ***	0.515
Invercargill	0.350 ***	-	0.321 ***	0.047	0.305 ***	-	0.344 ***	0.414 ***	0.419 ***	0.308 ***	0.393 ***

Notes: ***p=<.0001, **p=<.01, *p=<.05 assuming normality. Manuals are categories i–ii. Base is all categories i–vii. Data for 1911 are for first ballot. For sources see text. Spearman correlations produce very similar values.

Table B: Pearson correlations between estimated percentage of vote for Labour and percentage of skilled manuals across the streets/street sections of each town over the study period, 1911–51

Year/Town	1911	1914	1919	1922	1925	1928	1931	1935	1938	1946	1951
Hamilton	-	-	-	0.393 ***	0.342 ***	0.209 **	0.172 **	0.055	0.016	0.018	0.075
Gisborne	-	-	0.157 *	0.055	0.136	0.219 **	0.198 **	0.246 **	0.243 **	0.218 **	0.252 **
New Plymouth	-	-	-	-	0.090	-0.054	0.068	0.121	0.087	0.030	0.271 ***
Palmerston North	-	0.101	0.136	0.281 ***	-0.014	0.094	0.220 **	0.135 *	0.107	0.224 **	0.214 **
Wanganui	0.342 ***	0.242 **	0.166 **	0.201 **	0.265 ***	0.378 ***	0.339 ***	0.279 ***	0.303 ***	0.225 ***	0.216 ***
Napier	0.179 *	-	0.294 ***	0.232 **	0.181 *	0.241 **	0.332 ***	0.388 ***	0.371 ***	0.349 ***	0.377 ***
Hastings	0.196 **	-	0.170 *	0.138	0.147 *	0.182 **	0.123	0.206 **	0.270 ***	0.008	-0.021
Nelson	-	-	0.060	-	0.106	-	-	-	-	0.321 ***	0.252 **
Timaru	0.225 **	-	0.094	0.124	0.105	0.161 *	0.198 **	0.189 **	0.207 **	0.188 **	0.220 **
Invercargill	0.176 **	-	0.101	0.001	0.192 **	-	0.134 *	0.198 **	0.238 ***	0.258 *	0.239 ***

Notes: ***p=<.0001, **p=<.01, *p=<.05 assuming normality. Skilled manual is category ii.
Base is all categories i–vii. Data for 1911 are for first ballot. For sources and procedures see text.
Spearman correlations produce very similar values.

Table C: Pearson correlations between estimated percentage of vote for Labour and percentage of unskilled/semi-skilled manuals across the streets/street sections of each town over the study period, 1911–51

Year/Town	1911	1914	1919	1922	1925	1928	1931	1935	1938	1946	1951
Hamilton	-	-	-	0.390 *	0.477 *	0.590 *	0.485 ***	0.602 ***	0.591 ***	0.466 ***	0.455 ***
Gisborne	-	-	0.183 **	0.333 ***	0.262 **	0.326 ***	0.340 ***	0.249 **	0.347 ***	0.177 **	0.251 **
New Plymouth	-	-	-	-	0.311 ***	0.244 **	0.339 ***	0.294 ***	0.327 ***	0.262 ***	0.282 ***
Palmerston North	-	-0.150	0.165 *	0.114	0.156 ***	0.231 **	0.432 ***	0.413 ***	0.293 ***	0.195 **	0.320 ***
Wanganui	0.207 *	0.324 ***	0.361 ***	0.300 ***	0.427 ***	0.393 ***	0.455 ***	0.405 ***	0.405 ***	0.386 ***	0.374 ***
Napier	-0.231 *	-	0.369 ***	0.530 ***	0.500 ***	0.507 ***	0.394 ***	0.500 ***	0.434 ***	0.392 ***	0.457 ***
Hastings	-0.155 *	-	0.098	0.082	0.041	-0.097	0.074	0.010	0.060	0.169 **	0.358 ***
Nelson	-	-	0.300 **	-	0.314 **	-	-	-	-	0.480 ***	0.509 **8
Timaru	0.279 ***	-	0.300 ***	0.367 ***	0.444 ***	0.544 ***	0.597 ***	0.572 ***	0.592 ***	0.522 ***	0.518 ***
Invercargill	0.258 ***	-	0.304 ***	0.056	0.204 **	-	0.323 ***	0.365 ***	0.344 ***	0.183 **	0.301 ***

Notes: ***p=<.0001, **p=<.01, *p=<.05. Unskilled/semi-skilled is category i. Base is all categories i–vii.
Data for 1911 are for first ballot. For sources and procedures see text

Table D: Pearson correlations between percentage of unskilled/semi-skilled manuals and percentage of skilled manuals across the streets/street-sections of each town over the study period, 1911–51

Year/Town	1911	1914	1919	1922	1925	1928	1931	1935	1938	1946	1951
Hamilton	-0.079	0.061	-0.008	0.050	0.094	0.087	-0.089	-0.185 **	-0.196 **	-0.111 *	-0.161 *
Gisborne	-0.210 **	-0.125	0.0005	-0.110	-0.032	-0.174 **	-0.103	-0.039	-0.217 **	-0.185 **	-0.105
New Plymouth	0.167	-0.013	-0.174 *	-0.026	-0.017	-0.047	-0.047	-0.129	-0.153 **	-0.146 *	-0.063
Palmerston North	-0.048	-0.015	-0.130	0.036	-0.033	-0.190 **	0.081	-0.022	0. 026	-0.006	0.028
Wanganui	0.079	0.149 *	-0.012	-0.058	-0.010	0.027	0.005	-0.052	-0.065-0.	029	0/005
Napier	0.086	0.036	0.072	0.039	0.048	0.136	0.099	0.075	0.076	0.082	0.109
Hastings	-0.178 *	-0.162 *	-0.187 **	-0.109	-0.046	-0.067	-0.040	-0.038	-0.023	-0.118	-0.026
Nelson	-0.017	-0.002	-0.139	-0.126	-0.031	-0.009	0.156	-0.041	0.092	0.028

Cleavage Within the Working Class? The Working-Class Vote for the Labour Party in New Zealand, 1911–51