Hostname: page-component-78c5997874-fbnjt Total loading time: 0 Render date: 2024-11-18T17:41:36.553Z Has data issue: false hasContentIssue false
  • English
  • Français

International Migration and Economic Growth: Australia, 1865–1935*

Published online by Cambridge University Press:  03 February 2011

Allen C. Kelley
Allen C. Kelley
Affiliation:
University of Wisconsin
Get access Rights & Permissions [Opens in a new window]

Extract

The nature of migration into Australia in the nineteenth and twentieth centuries provides much valuable material relating to the mechanism of interregional population transfers during one of the most important periods of demographic redistribution in recorded history. Examining this country's experience during the seventy years, 1865–1935, two problems are treated which not only contribute to the understanding of Australian growth but also provide insight into important aspects of American economic development.

Type
Articles
Information
The Journal of Economic History , Volume 25 , Issue 3 , September 1965 , pp. 333 - 354
Copyright
Copyright © The Economic History Association 1965

Access options

Get access to the full version of this content by using one of the access options below. (Log in options will check for institutional or personal access. Content may require purchase if you do not have access.)

References

1 Compare: Thomas, Brinley, Migration and Economic Growth (Cambridge, [Engl.]: The University Press, 1954), ch. viiGoogle Scholar; Kuznets, Simon, “Long Secular Swings in the Growth of Population and in Related Economic Variables,” Proceedings of the American Philosophical Society, CII (1958), pp. 25–52Google Scholar; and Abramovitz, Moses, “The Nature and Significance of Kuznets Cycles,” Economic Development and Cultural Change, IX (Apr. 1961), pp. 225–48CrossRefGoogle Scholar; Richard Easterlin, “Influences in European Overseas Emigration Before World War I,” ibid., pp. 331–51; Fleisher, Belton, “Some Economic Aspects of Puerto Rican Migration to the United States,” The Review of Economics and Statistics, XLV (Aug. 1963), pp. 245–53CrossRefGoogle Scholar; Jerome, Harry, Migration and Business Cycles (New York: National Bureau of Economic Research, 1926), chs. iii-ivGoogle Scholar; Reder, Melvin W., “The Economic Consequences of Increased Immigration,” The Review of Economics and Statistics, XLV (Aug. 1963), pp. 221–30.CrossRefGoogle Scholar

2 Phillips, P. D. and Wood, G. L., The Peopling of Australia (Melbourne: Macmillan, 1930), p. 54.Google Scholar

3 The migration statistics are recorded net overseas arrivals into Australia and include the expeditionary forces during the war. This period, although graphed in Chart I, is excluded from all statistical computations.

4 Over the entire period, approximately 25 per cent of total population increase was the result of net overseas additions. During long-swing expansions, however, this share increased to almost 40 per cent.

5 The long waves in natural population increase are plausibly associated with those of migration. The connection, however, may be more complex than the simple expected relationship. Namely, the large influx of migrants in the 1850's resulted not only in a skewed age distribution of the population but also provided the basis of subsequent echo-like waves in household formation and birth rates. Whether this influence was reinforced by, or whether it indirectly initiated, long waves in net migration is an unsettled issue. See Hall, A. R., “Some Long Period Effects of the Kinked Age Distribution of the Population of Australia, 1861–1961,” The Economic Record, XLIX (Mar. 1963), pp. 43–52.CrossRefGoogle Scholar

6 A related question arises whether the results obtained below derive mainly from short-run variations in the time series (i.e., random, or business-cycle influences) or from major movements which might individually be identified and denoted as long swings. The model of migration developed in Section III does not provide the guide lines for separating these various fluctuating elements, and thus such an exercise is not formally undertaken here. Some preliminary impressions regarding the significance of long swings may be formed, however, by visually filtering out short-run influences with the assistance of an arbitrary statistical measure, a five-year moving average. To this end, Chart I presents both the annual and the smoothed data of Australian net migration. Examination of these series clearly suggests that much of the correlation between Australian net migration and appropriate explanatory variables is to be found in the very wide swings in the series.

7 Census of New South Wales (Sydney, 1891), pp. 183–87Google Scholar; Official Yearbook of the Commonwealth of Australia (Melbourne: McCarron, Bird & Company, 1917 and 1934), pp. 108 and 783 respectively.Google Scholar

8 Ferenczi, Imre and Willcox, W., International Migration (New York: National Bureau of Economic Research, 1929), pp. 628–29Google Scholar; Britain, Great, Parliamentary Papers, XXVI, 1933–34, p. 9; Thomas, p. 305.Google Scholar

9 T. Brassey, M. P., has vividly delineated two of these issues in 1872, noting: “New South Wales alone contains 375,000 square miles; and a large proportion of this unoccupied territory possesses every natural advantage for agricultural development. But so long as the price of land in our Australian colonies remains at £ 1 an acre, when 160 acres can be obtained in America for nothing, it is not likely that an English tenant farmer, with only a small capital at his command, will undertake a much longer and more expensive voyage to Australia, in preference to the shorter and infinitely cheaper passage across the Atlantic to America.” Brassey, T., M. P., , Work and Wages (London, 1872), p. 208; quoted in Thomas, p. 204.Google Scholar

10 Ibid., p. 205.

11 This refers to limitation of “white” settlers only; Australia's strong opposition to and restriction of nonwhite migrants is well known.

12 In several instances, the immigrant issue provided the main theme of political encounters. In 1877, for example, the New South Wales Government voted a large sum for migrant assistance, an event which drew strong opposition from labor groups (Trades and Labour Council, Working Men's Defense Association, etc.). This was enough to dissolve Parliament, topple the Government, and bring about a reduction in the size of the immigrant fund. See Coghlan, T. A., Labour and Industry in Australia (London: Oxford University Press, 1918), p. 1285.Google Scholar

13 Commonwealth Bureau of Census and Statistics, Demography Bulletin, XLIV (Melbourne: Government Printer, 1926), p. 726; Thomas, p. 305.Google Scholar

14 In at least one instance (N.S.W.), religious preferences also entered indirectly as a criterion for state assistance. See Coghlan, p. 1280.

15 On the flexibility of Government action in this area, see Coghlan's interpretation of the immigration policies of Victoria and Queensland, pp. 907–21, 1281, 1285.

16 At one point in Victorian experience, 25 per cent of the land revenue was set aside for encouraging immigration, as provided in the 1862 Land Act. “The automatic provision of funds for immigration purposes was opposed by many persons for it took no account of the labour market in Melbourne, whose condition, whether of prosperity or depression, was no wise indicated by land revenue.” Ibid., p. 921.

17 Victorian Year Book (Melbourne: Albert G. Mallet, 1912–13), p. 188.Google Scholar

18 Willcox, W. F., International Migrations (New York: National Bureau of Economic Research, 1931), Vol. II, ch. vi.Google Scholar

19 This basic model was proposed to me by Moses Abramovitz. For an analysis based on a similar approach, see Fleisher (cited in n. 1).

20 In the case of Australia, passage fares also fluctuated in the short run. But as argued above, these rate changes are related to the nature of disequilibrium in the labor market, the manifestation of which is implicitly examined by the analysis of unemployment rates.

21 There is ample evidence to support this position. Migration societies, formed to pool funds for the purpose of supporting individuals desiring to move, were a common phenomenon in Britain throughout the period. Similar groups were organized in receiving countries. See, for example, Coghlan, pp. 599–600, on the activities of the Family Loan and Colonization Society, Sidney Herbert's Female Emigration Society, and the Highland and Island Emigration Society.

22 Harry Jerome places considerable emphasis on this point to explain the sensitivity of migration to American business conditions. He estimates that prior to 1890, approximately one third of the immigrants had their passage paid by relatives. Jerome (cited in n. 1), p. 77.

23 One implication of this line of reasoning is that our basic model of migration takes on increasing relevance, other things equal, the smaller the costs of movement. This observation applies, not only to an explanation of total British emigration to all receiving areas, but also to the understanding of the differential elasticity of response both to alternative destinations and to a given country as costs change through time. Tests supporting these propositions would reveal that, ceteris paribus, the degree of positive association of British emigration to the domestic unemployment rate is higher (1) for countries with relatively low transport costs and (2) in a given country as the costs of passage decrease through time.

24 Because this study focuses on fluctuations in migration, as distinct from long-run trend considerations, the series on net migration should be trend adjusted. Several postulated relationships, however, failed to yield evidence of a statistically significant trend. We thus can conclude that the fluctuating characteristic of this series is indeed the main temporal feature of Australian migration experience.

25 The relative importance of Ua and Uuk is ascertained by reference to the following relationships:

26 A purely statistical matter may also result in a bias toward an apparent lagged relationship. The data on both net overseas arrivals and employment conditions are recorded annually in Australia. There is a time lag, however, of two to three months during which migrants make the voyage. Even if the migrant's response were contemporaneous with Australian conditions, a slight lag of recorded overseas arrivals would be expected.

27 A test of this hypothesis is basically different from most of the other considerations developed here. With respect to institutional delays, the general form Mt = f(Ut-n), where n represents an approximation to the lag, is appropriate. The object is to identify, both theoretically and statistically, the size of n.

On the other hand, a distributed lag response is interpreted here as a behavioral reaction pattern, and calls for a variant of the following formulation:

.

Assumptions on the nature of the weights are crucial to the model. Employing a moving average as the explanatory variable, for example, explicitly establishes fixed weights through the averaging procedure. Compare Working, E. J., The Demand for Meat (Chicago: Institute of Meat Packing, 1954)Google Scholar and Koyck, L. M., Distributed Lags and Investment Analysis (Amsterdam: North Holland Publishing Co., 1954)Google Scholar. See also Nerlove, M., Distributed Lags and Demand Analysis for Agricultural and Other Commodities, Agricultural Handbook No. 141 (Washington: Government Printing Office, 1958)Google Scholar. Both Nerlove and Koyck assume the weights decline geometrically through time. A good discussion of the nature of these models is found in Fisher's, F. M.A Priori Information and Time Series Analysis (Amsterdam: North Holland Publishing Co., 1962), ch. ii.Google Scholar

28 A two-year lag resulted in an r 2of .27.

29 Of the two alternations of (1) incorporated in (2)—the one-year lag versus the reciprocal transformation of the unemployment rate—the latter appears to be the more important influence in improving the explained variance. Lagging only the unemployment rate results in an r 2 of .29. On the other hand, employing the reciprocal without a lag increases this measure to .35. The combined effect of these two changes is represented in (2).

30 This general result is additionally found with respect to the remainder of the hypothesis tested in the present section. The empirical findings presented below thus implicitly reflect a rejection of the hypothesis of significant British influence on the postulated variables under consideration.

31 A similar observation applies to the Australian Government's decisions on assistance schemes.

32 The Durbin-Watson statistics are .62 and .84 respectively. These models are not rejected, however, since our theoretical explanation clearly would admit autocorrelation of residuals. An example of one such systematic element might be the effect of American (or other) long swings on the flow of migrants to Australia.

The selection of (3c) for more intensive study is thus somewhat arbitrary. This procedure is partially justified in noting that the general conclusions of this study are largely invariant to the specific choice of the distributed lag formulation explored above.

33 This equation was fitted by individually employing 1/Uat-n, n = 0, ‥, 2. A one-period lag was best statistically, a result which is consistent with both theoretical and previous empirical findings.

34 Equation (3c) resulted in a Durbin-Watson statistic of 1.67. The long-run coefficient of 1/Uat-1 is .36/(1.00-.63) or .97.

35 It must be recalled that this is the average speed of response over approximately seventy years. This model admits the possibility of the decision function (or reaction pattern) changing through time, but investigating the significance and nature of this possibility is outside the present objectives.

36 The output estimates are those of Butlin, N. G., Australian Domestic Product, Investment, and Foreign Borrowing, 1861–1938/39 (Cambridge [Engl.]: The University Press, 1963), pp. 460–61Google Scholar; Feinstein, C. H., “Income and Investment in the United Kingdom, 1856–1914,” Economic Journal, LXXI (June 1961), pp. 367–85CrossRefGoogle Scholar; and Prest, A. R., “National Income of the United Kingdom, 1870–1948,” Economic Journal, LVIII (Mar. 1948), pp. 31–62.CrossRefGoogle Scholar

This measure is similar to that which Simon Kuznets has used to explain American immigration. He notes that the “flow of goods to consumers, total and per capita, is perhaps the best aggregate measure of changes in economic well-being—which would affect … particularly, migration from abroad.” In the final analysis, changes in the flow of goods per capita are postulated as the main stimulus to migration. He found that peaks and troughs of long swings in migration lagged those in the independent variable by several years. See Kuznets (cited in n. 1), pp. 26–27.

37 We have additionally estimated the parameters of the above distributed lag model, employing levels of long-run per capita output in place of the formulation of (4). The general conclusions relating to the importance of these proxies of expected long-run income potentials in explaining fluctuations in migration are basically the same as was found using Δ Yat and Δ Yukt.

38 An example of a variable positively correlated with time is presented by Belton Fleisher. He argues that, with respect to Puerto Rican immigration into the United States, a major influence is to be found in the active encouragement and assistance of the resident Puerto Ricans in the U. S. Given the relatively large number of the new Puerto Rican arrivals as compared with the existing population in the U. S., one might expect this influence to display a significant positive trend. This hypothesis was found to be consistent with the facts. See Fleisher, p. 251.

A corollary of the argument is that the importance of these influences should diminish as the share of migrants to the total resident-migrant population in the country of destination increases. In the Australian case, we would thus expect this factor to be relatively insignificant in explaining variations in emigration from Britain.

39 We have arbitrarily chosen years in which the deviations from the predicted relationship exceed 10,000 to identify “exceptional” circumstances.

40 See above, p. 337.

41 From the business recovery in the early 1890's, Australia experienced in mid decade the longest drought of its history. During this period, unemployment was high and net migration predictably low or negative. Even though 1900–01 witnessed a decisive recovery in labor market conditions, potential migrants were hesitant to commit themselves without further confirmation that conditions had definitely been reversed; those years found net overseas arrivals low with reference to the long-term relationship of (3c). The years 1902–03 evidenced one of the most severe droughts in Australia's history—the migrant's previous caution was justified. Even though the labor market had begun to improve by 1906, it was not until 1908–09 that the trend was sufficiently established to again bring the rate of migration into line with our model.

42 Borrie, W. D., “The Peopling of Australia,” in The Australian Economy, ed. Arndt, H. W. and Corden, W. M. (Melbourne: Specialty Press, 1963), p. 104.Google Scholar

43 Thomas, ch. vii. For a review of Australian net capital imports which is formulated within a broad analytical framework similar to that employed in this paper, see Hall, A. R., The London Capital Market and Australia, 1870–1914 (Canberra: The University Press, 1963).Google Scholar

44 Abramovitz (cited in n. 1).

45 America obtained the bulk of British migrants throughout the period, but this feature is related largely to trend considerations; our observations and analysis are concerned with long fluctuations in this flow and as a result are relevant to the longswing problem.

46 The literature of American Kuznets cycles is an example of current research into this set of relationships.

47 According to the official view of the Ministry of Labour in 1926, from 1881–1925 “the general percentages provided a valuable guide to the direction of the changes in unemployment and a rough indication of the comparative state of employment at different periods, although they cannot be relied upon as an absolute measure of the total amount of unemployment in all industries at any particular date.” British Historical Statistics (Cambridge [Engl.]: The University Press, 1962), p. 57.Google Scholar

48 Commonwealth of Australia, Official Year Book (Canberra: Commonwealth Government Printer, 1936), pp. 543, 565–67Google Scholar; Butlin, N. G., “An Index of Engineering Unemployment, 1852–1943,” Economic Record, XXII (Dec. 1946), pp. 241–60.CrossRefGoogle Scholar

49 The same general comments made above on the characteristics of British unemployment rates thus apply to the Australian series. For a critical discussion of the Commonwealth Series, see Labour and Industrial Branch Report, No. 2 (Melbourne: McCarron, Bird, and Co., 1913), pp. 16–19Google Scholar. Commenting on the bias toward short-run employment fluctuations, this report notes: “It is not unlikely, however, that particulars of unemployment are, on the whole, more generally available for those trades in which liability to unemployment is above the average of skilled occupations” (p. 18).

50 In adding to the sample the three remaining years when the official index was calculated—1901, 1896, and 1891—the high degree of correlation remains at .984.

51 For example, Brian Fitzpatrick proposed that “1860–1890 was a generation during which children grew to middle age without personal experience of economic depression.…” The British Empire in Australia (Melbourne: Melbourne University Press, 1949), p. 272.Google Scholar