the enigma of australian manufacturing, 1851–1901

17
THE ENIGMA OF AUSTRALIAN MANUFACTURING, 185 1-1901 * ALLCAN THOMPSON University of Melbourne Australian studies of industrial, labour and economic history in the second half of the 19th century have long relied on manufacturing statistics pub- lished annually in the colonial Statistical Registers. This reliance could become even more pronounced through the influence of Professor N. G. Butlin,l who drew heavily upon these sources. This article suggests, however, that these Statistic Registers are unreliable and indeed-for certain purposes-mislead- ing. If this is so, some popular historical interpretations will have to be reviewed. We shall have to reconsider why Victoria adopted a protective tariff in the 1860s and also reconsider the subsequent effects of the tariff. We shall have to reexamine the effects of the gold-rushes on the growth of manufacturing. We shall have to reject much of Butlin’s pioneering work on the rise of manufacturing in Victoria and Australia between 1861 and 1901, and we shall have to reconsider the role of manufacturing in the boom of the 1880s and the recovery from the depression of the 1890s. The statis- tics and the estimates based on them are also relevant to such questions as how manufacturing influenced the growth of cities, how urban living stan- dards moved, and how the economic and social climate affected the growth of unionism and welfare legislation. Moreover, unless the manufacturing statistics are handled with a new discretion, the infection of error may spread to problems in local history, industrial history, colonial and national history; for these statistics are as basic to industrial history as detailed voting returns are to political history. I The Victorian Statistical Register data are of central importance because Victorian industrial history is so important and because Victoria is the only colony with an annual series covering the full period. Consequently Butlin used Victorian information, in one way or another, in making his manufacturing estimates for every colony. Other economic historians who *The author acknowledges the help that he received by discussing this paper with Dr W. A. Sinclair and Professor G. A. Blainey. 1 N. G. Butlin, Australian Domestic Product, Investment and Foreign Borrowing (London: C.U.P., 1962).

Upload: allan-thompson

Post on 03-Oct-2016

214 views

Category:

Documents


0 download

TRANSCRIPT

THE ENIGMA OF AUSTRALIAN MANUFACTURING, 185 1-1901 *

ALLCAN THOMPSON

University of Melbourne

Australian studies of industrial, labour and economic history in the second half of the 19th century have long relied on manufacturing statistics pub- lished annually in the colonial Statistical Registers. This reliance could become even more pronounced through the influence of Professor N. G. Butlin,l who drew heavily upon these sources. This article suggests, however, that these Statistic Registers are unreliable and indeed-for certain purposes-mislead- ing. If this is so, some popular historical interpretations will have to be reviewed. We shall have to reconsider why Victoria adopted a protective tariff in the 1860s and also reconsider the subsequent effects of the tariff. We shall have to reexamine the effects of the gold-rushes on the growth of manufacturing. We shall have to reject much of Butlin’s pioneering work on the rise of manufacturing in Victoria and Australia between 1861 and 1901, and we shall have to reconsider the role of manufacturing in the boom of the 1880s and the recovery from the depression of the 1890s. The statis- tics and the estimates based on them are also relevant to such questions as how manufacturing influenced the growth of cities, how urban living stan- dards moved, and how the economic and social climate affected the growth of unionism and welfare legislation. Moreover, unless the manufacturing statistics are handled with a new discretion, the infection of error may spread to problems in local history, industrial history, colonial and national history; for these statistics are as basic to industrial history as detailed voting returns are to political history.

I

The Victorian Statistical Register data are of central importance because Victorian industrial history is so important and because Victoria is the only colony with an annual series covering the full period. Consequently Butlin used Victorian information, in one way or another, in making his manufacturing estimates for every colony. Other economic historians who

*The author acknowledges the help that he received by discussing this paper with Dr W. A. Sinclair and Professor G . A. Blainey.

1 N. G. Butlin, Australian Domestic Product, Investment and Foreign Borrowing (London: C.U.P., 1962).

1970 THE ENIGMA OF AUSTRALIAN MANUFACTURING 77

checked the Victorian series also concluded that they were reasonably re- liable.? Now the earliest year in which their reliability can be fully checked is 1891. In April of that year one of the ten-yearly censuses was held, and indeed this is the first census which has been generally accepted by historians.3 The other advantage of 1891 is that the census coincides with an independent set of manufacturing statistics collected for the Reports of the Chief Inspector of Factories, Workrooms and Shops. We have, then, three sets of figures for Victoria, collected independently and almost simultaneously. My conclusions from the analysis of these figures are that the Register data, which Butlin used, are not a reliable source, that the census is a more authoritative source, and that the censuses can be used for the whole of the crucial period 1861- 1901. A set of manufacturing figures emerges from these conclusions which is very different from that normally used.

Before comparing the three sources, it is as well to note that I compare the census with the 1890 Register and not with the more usual 1891 Register. This is because the 1890 Register data were probably collected in January or February 1891, shortly before the census. It seems likely that information collected referred only to those early months of 1891, because manufac- turers were not required to keep employment records for the whole year and the questionnaire did not ask for average employment for the previous year.4 The choice of the 1890 Register is clearly endorsed by Henry Hayter, the Victorian statist, in his own comparison between census and Register.5

The three estimates of manufacturing employment are as follows :

50,320 56,369 96,013

Factory Inspector’s Report (collected late in 1890) Statistical Register (collected early in 189 1 ) Census (April 189 1 )

With the Register figures a little above the Factory Inspector’s figures, most observers have assumed that the Register was “fairly cornprehen~ive”,~ but the similarity of these figures may be mere coincidence. This possibility must

2 See J . D. Bailey, Groic,th and Depression: Contrasts in Australian and British Eco- nomies 1870-1880 (A.N.U. Social Science Monographs, 1956).

3 See Brian Fitzpatrick, The British Empire in Australia, 2nd edn. (Melbourne: M.U.P., 1948), pp. 197 et seq.; N. G . Butlin, Investment in Australian Economic Develop- ment 1860-1900 (London: C.U.P., 1964), p. 196; and E. C. Fry, The Conditions of the Urban Wage-earning Class in Australia in the 1880s, unpublished thesis submitted for the degree of Ph.D. in the A.N.U. (1956), pp. 535-536.

4 Chief Secretary’s Correspondence, State Library Archives (hereafter S.L.A.) . ij Census of Victoria, 1891: General Report, p. 261. GThe number given is simply the crude total of manufacturing sub-orders. A better

estimate can be derived from the census by carefully aggregating individual occupa- tions. See footnote 12 below.

7 Bailey, op cit.

78 AUSTRALIAN ECONOMIC P A P E R S JUNE

be given serious consideration when we realize that the coverage of the two sources was quite different:

1. The Factory Inspector’s Report, unlike the Register, excluded all em- ployers, all family businesses, all government workshops, all factories in shires, and a group of industries of which the most notable were gasworks, limeworks and shipyards. The Factory Inspector had to exclude these because of the circumscribed scope of the Factory Acts under which he worked.

2. Both recorded all enterprises using mechanical power, but the mini- mum size of other enterprises differed: the Reports listed only those with at least six employees, whereas the Register minimum was four employees, or five or more including a working proprietor. The Register actually recorded even smaller enterprises than this formal minimum and Butlin believes that so many of these were included that it is improbable that much was omitted at any stage.R If he is correct on this point then the Register was even more deficient than my final estimate shows, since I work on the assumption that the Register minimum for manually operated factories was four or more employees as against six or more in the Factory Inspector’s Reports.

We can make estimates for each of these deficiences with the unfortunate exception of family businesses, which are not listed separately in any of the available figures. The additions to the Report indicate that there must have been at least 3,900 factories employing at least 64,000 persons in 1891. Some of these estimates are based on the Register, which is almost certainly de- ficient, and the figures could be as high as 4,220 factories with 68,500 persons.

Another vital defect was the Factory Inspector’s inability to deal ade- quately with the major manufacturing category of clothing and related trades. Even in the mid-nineties, the Chief Inspector was complaining of the vagaries of this important industry. The industry employed an unusual number of outworkers-working in their own homes and paid by the piece-who were not brought within the scope of the Factory Act until 1894, and even in 1896 one inspector admitted that “a large number must, however, still remain ~nregistered”.~ Moreover, sweating was still widespread in the clothing in- dustry: many owners therefore had an incentive to avoid registration, which would bring them under the supervision of the factory inspectors. This must also have been important in minimizing the Factory Inspector’s statistics in 1891. Thus we find 26,666 employees in the 1891 census and only 15,065 in the Report, a difference of 11,601, and there is also a discrepancy of 1,498

Australian Doniestic Product, etc., op cit., p. 156. Report of the Chief Inspector of Factories, Workrooms and Shops (hereafter re- ferred to as Factory Inspector’s Report) , Victorian Parliarnetitmy Papers (hereafter V.P.P.)., 1899-1900, VOI. 111.

1970 THE ENIGMA OF AUSTRALIAN MANUFACTURING 79

employers. Some of these would have been employed in what Butlin graphic- ally describes as “backyard workshops”-workshops not using mechanical power and having less than four employees-but I estimate that only about 3,000 could have been working in such workshops.1° Since the total dis- crepancy is 13,099, the Report has missed some 10,100 clothing workers. Added to the other estimates, this shows that there must have been at least 74,100 factory workers in Victoria in the summer of 1890-91, and possibly as many as 78,600.

In comparing these results with the census we must realize that the census included not only factory workers but also workers in backyard workshops and workers who were unemployed or self-employed. The last two categories are no problem as they were listed separately in the census and can be excluded easily from the comparison. What remains will be workers in factories and workshops. Surprisingly there are more employees in our esti- mates of factories than in the census. This is because the census did not count “general labourers” as a manufacturing sub-order, although some would have been working in manufacturing (since most general labourers are normally either building or manufacturing workers). My estimatell is that 11,200 of the 34,333 general labourers worked in manufacturing, and, if this is correct, the total manufacturing workforce is 103,700,12 which seems to be compatible with our factory estimates of between 74,100 and 78,600. The difference between census and factory totals is accounted for by 5,562 unemployed, 8,493 self-employed and approximately 3,420 back- yard workshops employing 11,000 to 15,500 persons. These figures show the average number of employees per workshop at between 2.2 and 3.1- a little high perhaps, but quite plausible, since some of the adjustments to the factory estimates are conservative. A similar comparison of the Register and census gives quite impossible results-the difference is 5,040 employcrs and 28,200 employees ( a total discrepancy of 33,240), implying that back- yard employers had an average of 5.6 employees. This is quite impossible when, by definition, a backyard workshop had a maximum of three em- ployees. Obviously a number of factories were missed in the Register col- lection.

Strangely, despite the arithmetical discrepancies between employment totals in the Register and the census, Butlin claims that his estimates of gross out- put and value added are supported by a check against official estimates based 10 This estimate is based on the assumption that the average number of employees

in backyard workshops was two and in factories 19.1 (the average for all recorded factories).

11 See the Appendix. 1” This estimate is got by adding the 11,200 to a carefully computed aggregate of indi-

vidual census occupations rather than to the total of manufacturing sub-orders. The latter include house-painters and plumbers. My estimate of manufacturing workers other than labourers is 92,500, whereas the total of the sub-orders is 96,013.

80 AUSTRALIAN ECONOMIC PAPERS JUNE

on census returns, not only in 1891 but also in 1881 and 1901.13 However, Butlin is mistaken in assuming that these official estimates were based on the demographic census or an associated production census. Such assumptions are plausible if the source is examined in isolation, because the estimates were made in the census years, and in 1881 were based on data collected by the census sub-enumerators. Nevertheless, these estimates were based on the Register, not the census even in 1881. In that year the census sub- enumerators were used to collect the Register data in addition to the census data. They were told that the Register data must be collected in the ten days prior to the census; and the instructions and questionnaires issued to them were similar in substance to normal Register collection. The experiment of using the census sub-enumerators was not repeated in 1891 and 1901, when the Register collection was made by normal annual tender.“ It is clear that the official estimates to which Butlin appeals were in fact based on Register collections and therefore it is not surprising that they support his own estimates based so heavily upon Register data.

This particular prop not only falls away: it falls on the toe of the user. If Butlin’s estimate is compared with a contemporary estimate which was based on the census, his estimate is shown to be far too low. In 1891 Henry Hayter estimated the annual income of the 96,013 in manufacturing sub- orders at f12.8 million15 and the comparable figure from Butlin would be less than &8 million.l6

These conclusions suggest that Butlin’s estimates of the value of manufac- turing production should be modified, and they can be modified by estimat- ing the productivity of workers excluded from the Register figures. Three basic categories are involved:

1. Factory workers: If we take our estimates based on the Factory In- spector’s Report, with a total factory employment of 74,100 or 78,600, the Register omits either 17,700 (estimate A ) or 22,000 (estimate B ) . The estimated production of this group is based on B utlin’s formula.

2. People in backyard workshops: The number taken depends on which of the two factory estimates in (1) is taken: if estimate A is used, backyard workers would total 15,500; if estimate B, 11,000. Pro- duction in this category is estimated by calculating total wage income

13 Australian Domestic Product, op. cit., pp. 155 and 159. 14 See Chief Secretary’s Correspondence 1881, S.L.A.; also H. Hayter, “The Coming

Census”, Report of the Second Meeting of the Australian Association for the Advance- ment of Science, 1890, p. 580.

15 Census of Victoria, 1891: General Report, pp. 219-220. lo lb id . , pp. 159-175. 17Butlin does not provide separate figures for each colony, but on the basis of his

formula the Victorian figure for 1891 would be f7.96 million.

1970 THE ENIGMA OF AUSTRALIAN MANUFACTURING 81

(computed from Butlin’s wage data) and adding 40 per cent to this to get the value of production. This arbitrary figure is based on the brbad assumption that the value of production of this group would be greater than mere wages but not so great as the production of a factory worker (Butlin’s value added by factory workers is about 80 per cent higher than wages).

3. Self-employed: These numbered 8,493 in the census. Their produc- tion is taken simply as estimated wage income calculated from Butlin’s wage data. This may understate production of this group.

These modifications are summarized in Table I. TABLE I

Estimated Production of Manufacturing Workers in Victoria Omitted from the Statistical Register in 1891

Workers Annual Wage Total Value Omitted Income Added

Estimate A 2.50

1.215 0.666

Factories 17,700 Workshops 15,500 Self-employed I 8,500

Estimate B Factories Workshops Self-employed

22,200 11,Ooo 8,500

1.74 0 .86 0.666

3.13 1.21 0.67

I I I -

! I I 5’01

A possible objection to these calculations is that they ignore the incidence of female employment. This could mean that the factories component of the estimates is overstated, as more than two-thirds of “factory” workers omitted from the Register were females, and female wage rates were con- siderably below male rates. But this will be offset by an understatement in Butlin’s estimate, which arises from applying a wage rate which used 1901 census data for weighting to Register data with a much higher incidence of male workers (84.4 per cent in 1890, as against 67.9 per cent in the 1901 census). Perhaps the simplest way to make the point is to say that Butlin’s wage is not suitable for the unadjusted Register data, but it is suitable for my revised total factory employment figures, which have a similar ma!e/female distribution to that of the 1901 census.

These calculations suggest that Butlin’s estimates of value added by manu- facturing in Victoria in 1890-91 are understated by almost f 5 million. Butlin’s estimate would be f7.96 million on his formula. His estimate of manufacturing production in Victoria must therefore be increased by about

82 AUSTRALIAN ECONOMIC PAPERS JUNE

56,369 51,410 14,542

} 22,162

62 per cent to €12.9 million. The error in the Victorian figures suggests that for Australia as a whole manufacturing production estimates must be increased by about 23 per cent. It is also possible that there were similar difficulties with Register enumerations in other colonies. This is certainly suggested by the quick comparison of Register and census figures in Table 11.

96,013 74,559 21,795 20,494 2,769

TABLE I1 Manujacturing Work Force 1891

Victoria New South Wales Queensland South Australia Western Australia

1 Register 1 census . _

1- 153,103 1 223,090

Note: Butlin regards the South Australian and Western Australian figures as unknown, and he assumes their combined total is equal to the combined total of Queensland and Tasmania.

If the productivity of those omitted from the Register is equal to that of those actually listed, Butlin’s Australian estimates must be increased by €9.9 million. This may be too high if the productivity of non-factory workers is markedly below that of factory workers, but on the other hand the Victorian figures suggest that these crude sub-order totals may understate the manu- facturing work force. It is clear that a very substantial revision is essential if the censuses are accepted.

A rather unexpected benefit of this revision is that it helps to eliminate an apparent conflict in Butlin’s analysis. He rightly emphasises the rapid growth of manufacturing in the period 1861-91, but at the same time his figures show that manufacturing produced only 10.5 per cent of gross product in 1892 although it employed 16.7 per cent of the workforce.18 The differ- ence suggests that the productivity of manufacturing was comparatively low, but how do we reconcile the rapid growth with the low productivity at the end of the period? A possible explanation is the high proportion of females in manufacturing. We can see how far this may alter the comparison by using one of Butlin’s assumptions-that the productivity of female workers was one-third of males-to estimate a “male equivalent” of the manufac- turing work force, and we can then compare this with a “male equivalent” of the work force as a whole. However, this only reduces the manufacturing work force to 15.6 per cent of the total, which is well above Butlin’s esti- mated contribution to production of 10.5 per cent. Another explanation is that manufacturing was less capital-intensive than other industries; but would it have been less so than the construction industry which, Butlin

Investment in Australian Economic Development, 1861-1900, op. cit., p. 196.

1970 THE ENIGMA OF AUSTRALIAN MANUFACTURING 83

estimates, had precisely the same share of gross product as its share of the work force? Moreover, the productivity of Butlin’s manufacturing sector is so low that, if true, it should have affected comparative wages; but there is no evidence that manufacturing wages were markedly below other wages. In fact, Butlin himself uses a manufacturing wage index as a general wage index in this period19-a practice which would be dangerous if manufac- turing productivity were really so low at the end of the period. A more likely explanation of the discrepancy is simply that Butlin has underestimated manu- facturing production in 189 1 . The revision suggested above would bring manufacturing production to about 14.6 per cent of the Australian total. The share of the work force is above this figure, though possibly not enough to call the estimates into question. If the suggested adjustments are reason- ably accurate, manufacturing was not the extraordinarily unproductive sector which Butlin’s estimates suggest.20 If historians continue to use Butlin’s figures, they could easily draw inaccurate conclusions about the living stan- dards of the manufacturing and urban work force and develop inaccurate explanations of union activity and industrial legislation in Australia in the eighties and nineties.

I1

Comparisons with other data suggest that there are dangers in using the Statistical Register in 1890-9 1. For earlier years historians have assumed that the Registers are a reliable source. They are, of course, very easy and convenient to work with, whereas the census information is bundled up in nineteenth-century packages and must be re-sorted before it can be used. This may be a reason why censuses have never been used as a check against the Registers. But can the censuses be used safely before 1891? The early censuses may be in disfavour because some are clearly unsatisfactory. The records of the New South Wales census of 1881 were destroyed by fire before they had been fully processed. There are more general objections, perhaps based on Coghlan’s statement that “no exact comparison can be made between the occupations of the people in 1891 and at previous census periods”.z1 The earlier censuses were certainly classified under a different system. In Victoria in 1871 and 1881 the system used was one that had been devised by Dr. William Farr for the censuses of England and Wales of 1851-71. The essence of Farr’s system was to classify occupations by the raw materials worked with. For example, wool merchants were grouped with

lelbid., pp. 276 and 396. 20 My adjustments may still be insufficient. Contrast the resulting estimates with

Hayter’s, which showed that manufacturing accounted for 19.4 per cent of the work force and 24.2 per cent of total income.

21 Census of New South Wales, 1891: General Report, p. 283.

84 AUSTRALIAN ECONOMIC PAPERS J U N E

woollen textile manufacturers. However, we do not have to use these strange classifications, because the figures for individual occupations were also pub- lished and we can re-classify them. This is quite valid if the occupational figures in early censuses are comparable with the 1891 census occupational figures. Coghlan thought that they were comparable except in one important respect-that the earlier censuses did not systematically distinguish between workers and dealers.22 But a careful scrutiny of the earlier returns suggests that there are few occupations where it is impossible to draw a distinction and that if individual occupation figures are used carefully there is little error involved.

This is confirmed by a comparison which Hayter made of the censuses of 1871, 1881 and 1891,23 using the Farr method of classification. Since Hayter had been a leading official for all three censuses, there is no reason to doubt the validity of this compari~on.?~ If we use the 1891 section of Hayter’s table, removing all obviously mercantile occupations from the manufacturing sub- orders, and if we compare the result with a careful estimate of the manufac- turing work force from the census proper for 1891, we find a high degree of similarity. For the whole manufacturing work force the error in using the Farr classification was less than five per cent. Much of this error would probably have been in residual categories such as “others working in drinks and stimulants”; and since all these censuses give a breakdown of these residual categories in footnotes the error can be reduced even further. My estimates suggest that of 15 industrial sectors only two have more than five per cent error: paper and paper products has a seven per cent error, and a very minor sector which includes jewellery, watchmaking, instruments and toys errs by 31 per cent. This large error in an insignificant group is quite understandable because so many in this group would be both workers and dealers. The small overall error leaves little doubt that these censuses can be used in Victoria.

The headings of the sub-orders were too broad and the different industries were jumbled.25 Separate occupational figures were not published, so that we cannot correct the sub-orders as we can with the later enumerations. This means that estimates from these censuses are less accurate than the later ones,

The Victorian censuses of 1861 and earlier are less satisfactory.

ZZIbid., p. 271.

23 Census of Victoria, 1891: General Report, pp. 395-405.

24 Coghlan agreed that a comparison of this kind was valid. Census of New South Wales, 1891: General Report, p. 275.

25 For example, bakers, brewers and distillers were placed in the same sub-class as greengrocers, grocers and produce merchants. A list of the occupations included in each sub-order is given in Census of Victoria, 1861 : Part I1 (Occupations of the People), Appendix, pp. 351-355.

1970 THE ENIGMA OF AUSTRALIAN MANUFACTURING 85

and also that a comparison by industrial sectors is limited to the few cate- gories which seem to have been reasonably treated in the sub-orders. For- tunately, the sectors which were probably most important - the clothing trades and metal industries-were more satisfactorily grouped than most; and as Farr’s classification was not used, the sub-orders contained fewer mercantile groups than similar sub-orders in the 1871 and 1881 returns. Even so, there were undoubtedly errors for which we must allow. As a first step, a list was prepared of occupations which were not manufacturing occu- pations but which appear in “manufacturing~’ sub-orders in 1861. A second list showed those “manufacturing” occupations which appear in other sub- orders. A calculation was made from the 1871 census of the percentage error involved by including those in the first list and excluding those in the second. As a check, a similar calculation was made from the 1891 census, which gives a more detailed breakdown of occupations. Each calculation was used to reduce the totals of “manufacturing” sub-orders in the 1861 census; there was no serious discrepancy between the two estimates. Of course it would be a mistake to think that the result is an accurate count, but it is probably a reasonable indication of the order of magnitude of the manufacturing work force in 1861.26

The census most difficult to compare with the others is that of 1901, again a departure from the common view of census usage. On a superficial examina- tion, there was a substantial increase in the manufacturing work force between 1891 and 1901, but the basic difference of method in the two censuses makes this difficult to prove. Whilst the 189 1 census was essentially occupational in approach, the 1901 census tried to classify “industrially”, so that for the first time “ancillary” workers were included in the manufacturing work force.27 The schedule for the 1901 census was improved also, permitting a number of general labourers to be identified with particular industries, although this was not completely successful. The 1901 census is thus more satisfactory for measuring the work force than the earlier enumerations, but for the same reason it is very difficult to compare with the earlier ones. As there was no occupational classification in 1901, there is no way of allowing for the inclusion of ancillary workers or of estimating the number of general labourers included in the manufacturing class.

With these reservations in mind, let us examine the census estimates of manufacturing work forces. These estimates are given in Table 111, which

26 Most of the non-manufacturing workers included were retail traders, but the number was small. The overall error is estimated to be less than ten per cent, so that it i s unlikely that any alternative treatment would produce markedly different results.

27 Census of Victoria, 1901, p. 401. “Ancillary” workers are non-operatives such as clerks, carters and cleaners.

86 AUSTRALIAN ECONOMIC PAPERS JUNE

Census Including

Labourers in Manufacturing

Year Census Estimate of General

also shows the discrepancies between the census estimates and Register em- ployment data.28 Note the marked increase in the proportion of the total manufacturing work force covered by the Register data.

TABLE I11 Estimates of Victorian Manufacturing Work Force 1861-91

% Excess of Statistical Census Over Register Register

1861 1871 1881 1891 1901

26,000 40,960 59,580 86,960 99,140

29,190 46,290 66.760 98;190 56;369 74,2

104.210 66.529 I 56.6

6,405 355.8 17,630 162.6 38.141 1 75 .0

Note: The census figures used here are net of unemployed.

It may be thought that a movement of the kind shown in the last column could be expected if factories quickly replaced workshops in the maturing Victorian economy after 1861. Such an assumption would be to mistake the content of the Register in the early decades, because the early Registers did not record factories. It is true that what were recorded were described as “extensive” establishments, but this only means that the Victorian Statist recognized the conceptual difference, common at the time, between “handi- craft” and “modern” manufacturing methods. In England a similar distinc- tion was drawn, but even there it was not drawn on the modern basis of numbers employed and/or whether mechanical power was used, but only on the basis of power used. The English usage was too rigid for the rela- tively primitive Victorian economy of the early 1860s, but the distinction was still drawn, though it appears to have been a qualitative distinction and was certainly not based on numbers employed. Since small establishments, such as brickyards each employing only one man using manual power, were recorded in the sixties, and it was only in the eighties that the modern concept of a factory began to emerge, we cannot explain the greater coverage of the Register in terms of a relative increase in the importance of “factories”. The main point arising from the comparisons is that the greater coverage was simply the result of improved collection of Register data.

Apart from comparisons already made with Census and Factory Inspector Report data there is a great deal of direct evidence that the Registers were d e f e c t i ~ e , ~ ~ and that the problems of collecting this new type of information

a*The small apparent improvement in Register coverage in the 1880s may have been partly the result of a change of definition which raised the minimum size of estab- lishments from about one employee in 1881 to four employees in 1889.

29 Additional evidence is given in my Statistical Measurement of Manufacturing Activity in Victoria, unpublished M.Comm. thesis, University of Melbourne (1964), chapter 3.

1970 THE ENIGMA OF AUSTRALIAN MANUFACTURING 87

were formidable. They were probably the first annual collections of their kind in the world"" and there were serious organizational defects which were dealt with by trial and error, and very slowly at that. One reason for this was that the Statist was preoccupied with the problems of collecting the agricultural statistics in the early years. However, certain positive improve- ments can be identified: in 1869 the Statist promised not to publish the details of individual schedules, thus encouraging manufacturers to supply information; the Local Government Act of 1874-75 reduced the size of statistical districts, and the Statist was given the power to compel manufac- turers to supply information; and from 1884 a bonus was paid to collectors according to the number of returns submitted. However, the weaknesses of the system of calling annual tenders for sub-enumerators remained through- out the 19th century, and the Local Government Act passed the respon- sibility for collection to local bodies. It was a responsibility which they resented and often discharged badly, and the Act had the unfortunate consequence of preventing the Statist from directly controlling the sub- enumerators, since he now had to work through local bodies.

All the evidence suggests that the annual data published in Statistical Registers of Victoria are inaccurate indicators of long-run growth trends of manufacturing activity in the last 40 years of the 19th century and that, contrary to current usage, the censuses provide the more reliable information.

This again affects Butlin's estimates. Admittedly, these are for Australia as a whole, but the Victorian figures loom large in Australian totals at this time, and their revision has serious implications for his analysis. For example, his figures indicate a phenomenal rate of growth for manufacturing in the 1860s. The censuses, however, suggest a growth rate of between five and six per cent a year in Victoria-less than half that shown by the Registers. Despite this scaling down, the decade of the 1860s still had the highest growth rate of any decade in the period 1860 to 1900. The most significant point, however, is that the censuses suggest that the rapid growth of manufacturing really derived from the 1850s.

Butlin's figures also show the growth of manufacturing in the 1870s as more significant than in the 1880s, both absolutely and relative to the growth of the economy. However, the censuses reverse this picture, at least in absolute terms. Manufacturing in the 1880s, instead of lagging behind the growth of the economy and even population, as the Register suggests, re- mained a dynamic element in the economy with a growth rate of about 4.3 per cent a year between the two censuses. This is considerably above the rate of 3.3 per cent a year in the 1870s and is consistent with what we know 30 See United Nations, Statistical Office, Department of Economic Affairs, Studies in

Methods: Industrial Censuses and Related Enquiries, Series F, No. 14 (New York: U.N., 1953), pp. 1-18 and Table I. I am indebted to Dr Duncan Ironmonger for this reference.

88 AUSTRALIAN ECONOMIC PAPERS J U N E

about the relative prosperity of Victoria in these two decades. However, it is possible that manufacturing did contribute more to the growth of the economy in the seventies than in the eighties. In the seventies manufacturing had absorbed almost 30 per cent of the addition to the work force in the decade, but in the eighties, with the work force expanding at almost twice the rate of the previous decade, manufacturing absorbed only 23 per cent of the increase. While the implications are obvious, we must bear in mind that manufacturing was a more significant element in the economy at the begin- ning of the 1880s than it had been a decade earlier, that it remained a dynamic element in an economy much more buoyant than it had been in the previous decade and that this greater buoyancy itself may have been partly due to a higher growth rate of manufacturing.

The census data also alter the accepted picture of Victorian industrial structure, for example, by showing that the clothing industry dominates manufacturing employment as early as 1871. W. A. Sinclair also gives some importance to this industry in Victoria’s recovery from the 1890s depres- sion,31 but the censuses suggest that this is a distortion: whereas the Register figures, which Sinclair used, show a doubling of employment in the industry between 1891 and 1901, the censuses show only a slight increase. The main reason for the discrepancy is probably the definitional changes adopted for the Register in the mid-nineties. These affected the clothing industry more than any other. However, Butlin argues that these changes were compara- tively unimportant, because “despite apparently radical reform . . . it is impossible to detect a break in the coverage (apart from that in Tasmania) through any tightening of the statistical delimitations of fa~tories”.~? The timing of the changes may, however, help to conceal any discontinuity in the figures. First, there were still inherent difficulties in enumerating clothing workers at this time, which means that changing the limits of enumeration would not have a full impact upon the overall employment figures. Secondly, the effects of changes in definition may be obscured by being spread over a number of years, partly because the collection machinery seems to have been cumbersome. This is suggested in the results of the widening of the coverage of clothing trades in 1894. The Statist estimated that the change increased the 1894 figures by 908,s but in the following year there was a large in- crease in the listing of clothing workers (2,62I)-an increase that is not paralleled in the Factory Inspector’s Report, which shows an increase of less than 1,OOO. Probably the effect of the change was spread-in the first year of the new limit the collectors were not able (or willing) to find and list all enterprises newly brought within the limit. In 1896 there were further 31 W. A. Sinclair, Economic Recovery in Victoria, 1894-1899 (A.N.U. Social Science

a2 Australian Domestic Product, etc., op. cit., p. 171. 33 Victorian Year Book, 1895-8, vol. 11, p. 930.

Monographs, 1956), pp. 84-95.

1970 THE ENIGMA OF AUSTRALIAN MANUFACTURING 89

changes, again affecting clothing. These changes together could have in- creased the total Register enumeration by over ten per cent between 1893 and 1896, so that the result is certainly a significant improvement. Finally, the discontinuity may not be obvious because the economy was still enmeshed in deep depression when the changes were made. This is significant because the statistics suggest that the depression affected medium and small-scale establishments more than large-scale establishments, which means that those included by lowering the limit were relatively less important than they would have been in more normal times. Moreover, the changes were made when the manufacturing sector seems to have been recovering from the nadir of depression. The break in the continuity could be in the form of an apparently more rapid recovery than was true, and a comparison of G.D.P. movements supports this conclusion. Perhaps half of the apparent increase in the manu- facturing work force between 1893 and 1896 was the result of definitional modifications in cIothing and other industries.s4 This suggests that the apparent increase in clothing employment shown by the Register during the nineties may be quite misleading, and this is confirmed by comparison with the censuses.

The revised analysis of manufacturing employment is relevant to many themes, but this article will consider just one more. The observation that the rapid growth of manufacturing in Victoria seems to have begun in the 1850s is important in interpreting Victorian history of the fifties and sixties. Both the growth of manufacturing and the successful demand for protection have been associated with the rapid decline of goldmining employment in the early sixties. The traditional explanation emphasises the competition for labour between gold and other activities and the way in which the decline of goldmining suddenly brought a flow of labour on to the general labour market. The Register supports these explanations, because it suggests that absorption of labour by manufacturing was extraordinarily high throughout the 1860s. The censuses, however, show a higher rate of absorption into manufacturing in the late fifties, when employment in goldmining was still high, than in the sixties, when there was a prolonged decline in the number of goldminers.

There is other evidence to support the view that manufacturing was grow- ing in the late fifties: the large number of soundly-based manufacturing enterprises with origins in that decade"5 and the vigorous origins of the trade union movement in the fifties seem inconsistent with the idea that a

34 For industries other than clothing, see the reference in footnote 33; also Victorian Year Book, 1902, pf. 255.

35See the list given by G . Serle, The Golden Age (Melbourne: M.U.P., 1963), p. 234. To Serle's list we may add the business of I. McIlwraith, founder of McIlwraith Industries Ltd., and the famous Phoenix Foundry at Ballarat. The wire-working firm of Greer and Ashburner took over a firm established in 1850.

90 AUSTRALIAN ECONOMIC PAPERS JUNE

manufacturing sector barely existed in 1861 .30 Moreover, the various Vic- torian censuses of the fifties show that the work force outside goldmining increased rapidly during most of the decade. It increased by perhaps 50,000 between 1851 and 1854, another 35,000 by 1857 and another 76,000 by 1861, giving a total of about 160,000. The corresponding increases in the whole ten years of the sixties was only about 62,000.37 Moreover, a year- by-year analysis of the number of goldminers, compared with numbers in other economic sectors,3s fails to show a simple competitive relationship. Labour was usually absorbed into other sectors of the economy most rapidly when the number of goldminers was increasing. Once goldmining began to decline, the rate of absorption into other sectors slackened.

This is not to say that goldmining can be ignored in the history of manu- facturing development at the time or that it is unimportant in explaining protection, but it does suggest that the relationship implicit in current treat- ments of the question is too ingenuous. The rise of manufacturing was not simply the result of a sudden flood of erstwhile goldminers on to the labour market in the early sixties, nor was protection simply required to find employ- ment for labour pouring from the g01dfields.~~ The censuses demonstrate that this labour had been flowing for some time and that other sectors of the economy had been absorbing it, but despite a reduction of labour coming on to the general labour market in the early sixties, both absolutely and relative to the economy, there was apparently much difficulty in finding e m p l ~ y m e n t . ~ ~ The decline of goldmining must have been a major cause of this recession. Those businesses which had been accustomed to servicing the gold towns would have found them changing in character as well as shrinking. These two factors would produce a sense of change and instabil- ity, an atmosphere in which protective sentiment would find receptive list- eners. The censuses suggest that a large and increasing proportion of the population had vested interests to protect; for them “protection” was not

36 Although building operatives played a conspicuous part in the trade union move- ment and in the struggle for the eight-hour day, unions in a wide range of manufac- turing activities were formed in the fifties. See Serle, op. cit., pp. 212-215.

37For the number of goldminers see Fitzpatrick, op. cit., p. 110. Census figures give even more striking results: the inter-censal increases in the work force except in gold- mining were 64,000, 74,000 and 42,000. In the sixties the increase was only 42,000. The discrepancy between the annual figures and census returns of goldminers is probably due to the problem of classifying part-time and occasional miners.

38 This is based on official estimates of intercensal population movements and assumes that changes in the proportion of work force to population between censuses took place by equal yearly steps.

39For an alternative view of the relation between goldmining and other sectors, see G. Blainey, The Rush that Never Ended (Melbourne: M.U.P., 1963), pp. 60 er. seq. Blainey supports the view that the rise of goldmining stimulated manufacturing in the fifties.

40Serle, op. cit., p. 246.

1970 THE ENIGMA OF AUSTRALIAN MANUFACTURING 91

simply an exercise in creating employment for erstwhile goldminers but a means of protecting existing enterprises and an assurance that they would participate in the growth of the Victorian economy.

APPENDIX

The estimate of general labourers engaged in manufacturing is of some importance. The calculations of the total manufacturing work force are influenced by conclusions about the number of labourers in manufacturing. No solution is perfectly satisfactory. The method used is admittedly crude: it is based on the assumption that the proportion of ‘‘labourers’’ engaged in manufacturing to all labourers is the same as the proportion of manufacturing work force to the whole work force.41 It is worth remembering that the estimate of factory employment based upon the Factory Inspector’s Report is not influenced by the allocation of labourers, so that in calculating the deficiency of Butlin’s production estimates in 189 1 the allocation is important mainly for estimating employment in small-scale enterprises. The estimate of general labourers then only affects the extent of the revision of Butlin’s results, not the proposition that a sub- stantial revision is necessary (even if no general labourers are included in the manu- facturing work force, the Victorian estimate needs to be increased by about 46 per cent).

Also, in 1891 the census excluded “ancillary” (non-operative) manufacturing workers. Thus if it is admitted that at least some general labourers were engaged in manufac- turing, my estimate of total manufacturing work force in 1891 is more likely to under- state than to overstate the aggregate. This is confirmed by comparison with 1901, when the Factory Inspector’s Report suggested that factory employment was about the same or even slightly lower than in 1891: yet the census, which included “ancillary” workers and some labourers in the manufacturing work force, had an aggregate of some 10,400 more than the 1891 censuses. In addition, not all general labourers were allocated in 1901-15,489 are listed and some of these were probably engaged in manufacturing.

For long-run purposes, the allocation of labourers in the different censuses will affect estimates of growth rates of the manufacturing work force from 1861 to 1901. We could ignore general labourers entirely on the assumption that they were a fixed pro- portion of the manufacturing labour force over the 40 years. This would not alter the results significantly except for 1891-1901: the difference of approach to the census of the latter year will obviously distort the result.

Alternatively, we could apply the method used in 1891 to earlier returns, though not to the 1901 census. In fact, when it is applied to the four censuses of 1861, 1871, 1881 and 1891, the resulting growth rates for manufacturing are not very different from those where general labourers are ignored entirely-the calculations show general labourers as a fairly constant proportion of the manufacturing labour force (varying only between 10.7 per cent and 11.3 per cent of estimated total manufacturing labour force). For 1901, it is impossible to make an estimate of the total number of labourers. Hence the method is simply to allocate undefined labourers between industries in the proportions that were found for 1891. It is not likely that alternative assumptions would alter growth rates for the nineties significantly.

The principal objection to this method is that on the surface it does not make suffi- cient allowance for variations in the building work force (building being the other major employer of non-agricultural labourers). Variations in the work force other than in manufacturing influence the allocation, but perhaps more specific emphasis should be given to variations in building and construction industries.

41 The 1947 census was the first to cross-classify accurately by occupation and industry. Manufacturing then accounted for about 26 per cent of the work force, and about 26 per cent of all labourers were engaged in manufacturing.

92 AUSTRALIAN ECONOMIC PAPERS J U N E

The results, however, are consistent with variations in the construction industry. We must first deduct from the number of general labourers an estimate of the number working on the waterfront.42 The remainder were probably mostly secondary workers -at least the statisticians of the 19th and 20th centuries treated them as such (agri- cultural labourers were listed separately). Deducting our estimate of general labourers in manufacturing leaves the estimated number of general labourers in the construction industries. Including labourers already enumerated in the construction industries and those now allocated, labourers accounted for 42.9 per cent of all employed construc- tion workers in 1871; 42.1 per cent in 1881; 39.9 per cent in 1891; and 40.0 per cent in 1901. Data for 1861 are not available.

42 The estimate is based on tonnages entered and cleared in all Victorian ports. These are compared with Sydney tonnages. It was claimed that 3,000 men were engaged on the Sydney waterfront in 1901: Sydney Morning Hercild, May 29, 1901.