FOR NJOSH ONLY

The Supply of Gold under the Pre-1914 Gold Standard Author(syf % D U U \ ( L F K H Q J U H H Q D Q G , D Q : 0 F / H D n Source: The Economic History Review, New Series, Vol. 47, No. 2 (May, 1994yf S S 9 Published by: Wiley on behalf of the Economic History Society Stable URL: http://www.jstor.org/stable/2598083 Accessed: 06-02-2017 18:47 UTC JSTOR is a not-for-profit service that helps scholars, researchers, and students discover, use, and build upon a wide range of content in a trusted digital archive. We use information technology and tools to increase productivity and facilitate new forms of scholarship. For more information about JSTOR, please contact [email protected]. Your use of the JSTOR archive indicates your acceptance of the Terms & Conditions of Use, available at http://about.jstor.org/terms Economic History Society, Wiley are collaborating with JSTOR to digitize, preserve and extend access to The Economic History Review This content downloaded from 142.104.160.169 on Mon, 06 Feb 2017 18:47:55 UTC All use subject to http://about.jstor.org/terms Economic History Review, XLVII, 2(I994yf S S 9 The supply of gold under the pre-igi4 gold standard' By BARRY EICHENGREEN and IAN W. McLEAN G old mining is one of the most important yet least understood extractive industries. Its nineteenth-century history has been studied in the context of two distinct and very different literatures: that on the classical gold standard and that relating to economic development in the regions of recent (Europeanyf V H W W O H P H Q W 0 L Q L Q J L V F H Q W U D O W R W K H O L W H U D W X U H R Q W K H J R O d standard because of what was allegedly that monetary standard's principal virtue: its tendency to stabilize the general level of prices over long periods. Under the gold standard the authorities stood ready to buy or sell whatever quantities of gold were supplied or demanded at the official price. Hence a fall in the economy wide price level implied a rise in the relative price of gold, and an increase in gold mining, assuming a positive price elasticity of supply. The increase in supply would moderate or reverse the decline in the price level since by the 'rules of the game' the authorities would translate increased gold reserves into increased money supplies.2 The virtues of this stabilizing mechanism have been much praised by the advocates of the gold standard and given formal expression in theoretical models.3 For so central an issue, the presumption that the flow of newly mined gold was in fact responsive to relative prices is buttressed by remarkably little evidence.4 Sceptics have argued that nineteenth-century gold supplies were dominated by a sequence of chance discoveries and by a small number of exogenous technological improvements.' In this view, fluctuations in supply reflected not movements along a stable supply curve but random shifts in capacity and cost due to forces unrelated to the monetary standard. While it is possible to argue that discovery and invention were themselves induced by changes in the profitability of mining gold, little such evidence I The research on which this article is based was undertaken principally during a visit by McLean to Harvard University and one by Eichengreen to the Australian National University. We are grateful for the hospitality of both institutions. Helpful comments by Michael Bordo, Hugh Rockoff, Mark Rush, and two anonymous referees are also acknowledged, as is the research assistance of James Giesecke, Simon Molloy, and Stephen Woodland. 2 For this argument the authorities do not need to translate mechanically an acquisition of gold into a corresponding amount of high-powered money; all that is necessary is that the two aggregates move in the same direction, i.e., that sterilization be at most partial. On different definitions of the 'rules of the game' and their implications for analysis of the classical gold standard, see Eichengreen, Gold standard. Note that the tendency for a fall in the price level to elicit an increase in gold supply is symmetrical with the tendency of a price level rise to depress gold production. The former is taken only as illustrative. 3 The views of the gold standard's advocates are surveyed by Bordo, 'Gold standard'. For a theoretical analysis of the classical gold standard which emphasizes the stabilizing influence of gold production, see Barro, 'Money'. 4 This evidence and the literature in which it is interpreted are discussed below. 5 For a review of these arguments, see Rockoff, 'Some evidence'. ?V Economic History Society I994. Published by Blackwell Publishers, io8 Cowley Road, Oxford OX4 iJF, UK and 238 Main Street, Cambridge, MA 02142, USA. This content downloaded from 142.104.160.169 on Mon, 06 Feb 2017 18:47:55 UTC All use subject to http://about.jstor.org/terms SUPPLY OF GOLD UNDER THE PRE-I9I4 GOLD STANDARD 289 has been presented. Moreover, historical studies of economic development in regions of recent settlement, epitomized by the work of Blainey, have suggested that factors other than the real price of gold provide the key to understanding variations in production.6 From this perspective, the pattern of settlement, the expansion of agricultural production, and the operation of capital and labour markets all influence the extent to which a region's gold reserve is exploited and condition the response of gold production to changes in price. These two bodies of literature thus provide radically different views of the determinants of gold supply in the nineteenth century, with strikingly different implications for the role of mineral production in the determination of commodity prices under the pre-194 gold standard. In this article we analyse the supply of gold in a framework designed to incorporate the insights of both traditions of historical writing. The results indicate that factors other than gold prices had an important bearing on gold production, and suggest that existing statistical estimates of the price elasticity of supply are likely to be biased and misleading. I The thesis that gold supply under the pre-I9i4 gold standard was responsive to relative prices has a long and distinguished lineage. The classical economists noted the tendency of gold production to respond to changes in the level of commodity prices, although they differed in their characterization of the elasticity of response, the lag, and the tendency for the gold market to be buffeted by chance discoveries. An early statement of the classical view was that of Cantillon in I755: he argued that gold supply was governed in the long run by the cost of land and labour required for its production. By implication, a decline in costs due to a falling price level would elicit a partially offsetting rise in production. Where that production occurred and how long was required for it to take place depended on the location and characteristics of the major gold and silver deposits.7 Ricardo similarly argued that the quantity of specie available to the world economy was ultimately determined by the cost of production.8 Like his contemporary Thornton, Ricardo's interest was stimulated by the Bullionist Controversy. Thornton described the dynamic process in which the issue of paper currency, by raising the price level, depressed the volume of gold output. As production costs rose relative to the market price of gold, the profitability of mining declined, causing 'those mines which have not yielded any rent, to be no longer worked; and the supply of gold . . . to be in consequence, somewhat reduced'.9 John Stuart Mill's position was generally consistent with those of his predecessors, emphasizing the tendency of gold production to vary with the 6 Blainey, 'Mineral discovery'. 7 Cantillon, Essai, bk. 2, chs. 6-7. A useful discussion of the classical school, upon which we have drawn in what follows, is provided by Bordo, 'Gold standard'. 8 Ricardo, 'Proposals'. 9 Thornton, Inquiry, p. 266. ? Economic Hiswry Society '994 This content downloaded from 142.104.160.169 on Mon, 06 Feb 2017 18:47:55 UTC All use subject to http://about.jstor.org/terms 290 BARRY EICHENGREEN and IAN W. MCLEAN relationship between the exchange value of gold and its production cost. Mill was more cautious, however, in characterizing the supply elasticity. The volume of production, he argued, was heavily influenced by the characteristics of ore deposits, leaving limited scope for price-induced changes in the volume of production. Hence any tendency for changes in the quantity of newly mined silver and gold to stabilize the overall price level was likely to operate with long lags.'0 Subsequent assessments painted an increasingly gloomy picture of the extent to which a stable yet elastic supply of gold could be relied upon to prevent price fluctuations. Wicksell argued that the response occurred only with long lags which did little to stabilize prices over the relevant horizon." Fisher, writing at the beginning of the interwar period, while adopting the classical assumption that gold production should in principle fluctuate inversely with the price level, noted that gold supply was in practice subject to unpredictable shifts, which led him to offer various alternatives to the gold standard. 12 Marshall also advocated symmetallism and a tabular standard as superior to a monetary system 'dependent on the hazards of mining'.13 He too suggested that the supply of gold was insufficiently elastic to offset price level shocks, and attributed the price stability of the preceding era as much to fortuitous technological advance in the mining industry as to the systematic response to relative prices. Keynes, not surprisingly, adopted the perspective of his teacher: in his early writings he suggested that price stability in the late nineteenth century was due in large measure to the discovery of new gold deposits, attributable to chance but also to Europe's penetration of the regions of recent settlement. Since the annexation of territory had run its course, gold discoveries had become less likely."' By the end of the I920S, the international gold standard had been reconstructed and prices began to exhibit a pronounced downward trend. Pessimism about the responsiveness of the mining industry deepened. The Gold Delegation of the League of Nations in its First interim report provided a discouraging assessment of the scope for additional production, and submitted two forecasts of output in the I930s, one compiled by the staff of the delegation on the basis of official or semi-official estimates provided by the principal producing countries, the other constructed independently by Kitchin.'5 The forecast constructed from official sources projected a small increase in output from $405 million in I930 to $407 million in I93I and then a steady decline to $3I4 million in I940. Kitchin's estimate rose from $404 million in I930 to $4I0 million in I932 before declining to $370 million in I940. In fact, the volume of production increased dramatically from 20.8 million fine ounces in I930 to more than 27.4 million fine ounces 10 Mill, Principles, pp. 502-3. For a comparable assessment see Bordo, 'Gold standard', pp. 40-I, which characterizes Mill as arguing that there is only a long-run tendency at best for stabilizing gold supplies to minimize price-level fluctuations. Similar views appear in Cairnes, Essays. " Wicksell, Interest, p. 3I. 12 Fisher, Purchasing power. 13 Marshall, Money, p. 52. 14 Keynes, Tract. '5 League of Nations, Gold delegation, pp. I2, 6o. ?V Economic History Society 1994 This content downloaded from 142.104.160.169 on Mon, 06 Feb 2017 18:47:55 UTC All use subject to http://about.jstor.org/terms SUPPLY OF GOLD UNDER THE PRE-I9I4 GOLD STANDARD 29I in I934.16 Subsequent commentators concluded that the Gold Delegation had understandably failed to anticipate the rise in real gold prices associated with the Great Depression but had also underestimated the price elasticity of supply.17 The dominant impression left by this review of previous opinion is one of lack of consensus on the elasticity and stability of the supply of newly mined gold. Yet many recent studies have proceeded on the presumption that supply should be price elastic. This would appear to be the position of the majority of the US Gold Commission, whose members stated in their report that: Under the gold standard, a rise in the purchasing power of gold ultimately increased the rate of growth of the U.S. monetary gold stock by raising the rate of world gold output and inducing a shift from non-monetary to monetary use of gold. Movements in the purchasing power of gold thus preceded long-term movements in the monetary gold stock."8 A recent study by Rockoff seeks to evaluate the view that increases in gold output in the nineteenth century were due to movements along, rather than an 'accidental, if fortunate, series of shifts in, the supply curve' of newly mined gold.'9 Rockoff surveys the secondary literature on the history of gold mining, emphasizing the circumstances surrounding particular gold discoveries, the sources of productivity growth, and the influence of public policy on industry structure, conduct, and performance. He concludes that the proximate source of increases in the supply of gold in the nineteenth century was not changes in the rate of extraction of known resources but a small number of great discoveries. His assessment of the role of real gold prices in these discoveries is mixed; of the California discovery of i848, he suggests that 'it seems to have been the accidental by-product of the expansion of agriculture into the interior of the state' rather than an induced response to changing relative prices.20 If at times Rockoff is hesitant to generalize about causes, his overall assessment is favourable to the induced- response hypothesis.2' He concludes that the surges [in gold output at mid century and in the late i89os] were due primarily to the development of new gold fields and to a lesser extent to technological changes. The question is whether these events should be viewed as fortunate accidents or as changes induced by previous changes in the real price of gold. Circumstantial evidence, particularly for the second surge in 16 Report to the Congress, tab. SC2, p. i89. 17 Hardy, Enough gold?, pp. 60-4. "I Report to the Congress, I, p. II2. Congressman Henry Reuss, it should be noted, dissented from this view, insisting that 'New discoveries were a far more important source of change in the world gold stock than changes in demand' (ibid.yf 7 K H V W D I I D S S H Q G L [ W R W K H 5 H S R U W V S H F L I L H G W K H G X D O P H F K D Q L V m by which the response of the mining industry supposedly arises: increased exploitation of known reserves and induced discovery of new resources. As the authors of the appendix put it, 'Until recent decades . .. a rise in the real price of gold would lead to an increase in output and ultimately to the possibility of gold discoveries' (ibid, p. i6oyf . 19 Rockoff, 'Some evidence', p. 6I3. 20 Ibid., p. 625. 21 He goes only so far as to suggest that 'prospecting, partly in response to changes in the real value of gold, played an important role in the great discoveries' of the nineteenth century, and cites Blainey's study of Australian experience as consistent with this view (ibid., p. 626yf . (C Economic History Society 1994 This content downloaded from 142.104.160.169 on Mon, 06 Feb 2017 18:47:55 UTC All use subject to http://about.jstor.org/terms 292 BARRY EICHENGREEN and IAN W. MCLEAN production, suggests that the discoveries were induced. Tentatively, then, it seems appropriate to regard changes in supply produced by new discoveries at the end of the century as movements along a long-run curve rather than as a series of curves with arbitrary shifts between them.22 There have been two attempts to provide support for this position by estimating the price elasticity of gold supply under the pre-1914 gold standard, those of Bordo23 and of Rush.24 Rush's estimate is derived from a regression of gold production on its own lagged value and on the real price of gold. The latter variable is the reciprocal of the price level. This specification of the gold price was chosen because under the gold standard the nominal price of gold was fixed, hence the real or relative price of gold fluctuated inversely with the general price level. As reported, however, the results are difficult to interpret.25 Further, Bordo provides only limited information on the model employed in his analysis. He reports that deviations from trend in monetary gold stocks are most highly correlated with deviations from trend in the purchasing power of gold upon lagging the second variable by 25 years when world production data for the period i821-I9I4 are used, by I4 years when US production data for the period i879-1914 are used, and by i6 years when world production data for the same period are used.26 No explanation is offered for the length of the lag or for why it falls so dramatically in the late nineteenth and early twentieth centuries.27 The price-theoretic approach of these authors contrasts with the work on regions of recent settlement, exemplified by that of Blainey.28 Blainey argues that the probability of discovery had a systematic component which moved inversely with the business cycle but concludes that relative prices had at best a 'dubious' role.29 Instead of prices he stresses the relationship of discovery to the state of the regional labour market. The mechanism is the tendency for the unemployed to turn to prospecting as a form of support. Along with unemployment, Blainey suggests that low interest rates associated with depressed macroeconomic conditions aided speculative ventures.30 Besides linking alluvial gold discovery to the business cycle, Blainey posits an additional ambient pre-condition-'The principle that human settlement 22 Ibid., pp. 639-40. 23 Bordo, 'Classical gold standard'. 24 M. Rush, 'Money, output and unemployment in the United States under the gold standard', typescript, I978, cited in Barro, 'Comment'. 25 While t-statistics in excess of 3 are obtained for both explanatory variables, in his report on Rush's results Barro provides no ancillary statistics with which to evaluate the consistency of the point estimates and standard errors. 26 Bordo, 'Classical gold standard', p. i0. 27 Hirsch, 'Influences', noted that evidence supported the existence of a positive correlation between the real gold price and world gold production from igio to about i950, when it broke down because of large increases in South African output during a period when the real price of gold was declining. A study of the period i950 to i980 conducted by the staff of the US Gold Commission found that Free World gold production was negatively and significantly related to the real price of gold in the current and immediately preceding year: Report to the Congress, i, pp. i66, i8o. See also the discussion in Cooper, 'Historical facts'. 28 Blainey, 'Mineral discovery'. Although Blainey is not concerned with the operation of the gold standard per se, his investigation of the circumstances and timing of the major Australian gold discoveries is relevant to the present debate. 29 Ibid., p. 3I2. 30 Ibid., p. 308. (C Economic History Society 1994 This content downloaded from 142.104.160.169 on Mon, 06 Feb 2017 18:47:55 UTC All use subject to http://about.jstor.org/terms SUPPLY OF GOLD UNDER THE PRE-1914 GOLD STANDARD 293 hastens mineral discovery'-which he argues applies to California and the Witwatersrand as well as to Australia. In this view, the spread of rural settlement into regions containing mineral deposits not only increased the probability that their presence would be discovered but promoted the development and successful extraction of known reserves.3' Blainey notes, however, that the discovery of gold fields lagged unevenly behind a region's initial penetration by shepherds, carters, or timber-cutters, the length of the lag depending in part on business cycle conditions. Overall, Blainey paints a richer and more complex picture of the determinants of gold discovery and extraction in newly settled regions than do those who focus solely on relative prices. The list of possible non-price influences which were important to the timing of gold discovery and production in particular regions may be extended. Changes in regulations governing access to gold-bearing lands, or the taxation of gold recovered might induce output variation.32 Prior discovery of other minerals could stimulate awareness of the geological potential of an area, including the prospects of a gold find.33 The reduction in transport costs through road, port, or railway construction, where the improved communications were a favourable externality to the gold mining industry, must have reduced the costs of search and raised the profitability of gold production. The invention and diffusion of new mining technology similarly may have spurred output in regions where it was especially suited. Well-known examples include the transfer to Victoria of Californian placer mining methods in the I850s; the adoption in the US towards the end of the century of the New Zealand-invented mining dredge; and the adoption of the cyanide process for extracting gold from quartz deposits. The rate of extraction following discovery may have been higher for alluvial than for reef deposits: the rush phenomenon is most dramatic when associated with the former, as in the early years of the mid-century mining booms in California and Victoria. Finally, the possibility that regional variation in gold production was primarily the result of chance discoveries cannot be ruled out. 31 Blainey, 'Mineral discovery', p. 300. An essay by McCarty similarly argues that the timing of mineral discovery in the western US was the consequence of exploration and settlement. See McCarty, 'Suggestions'. McCarty acknowledges the earlier writings by Blainey on the history of Australian mining: ibid., p. I I-0, n. 7. 32 It is possible that the law of trespass on crown lands may have deterred prospecting for gold in Australia prior to i85i: Blainey, Rush, pp. II-2. Similarly, it has been claimed by Morrell that changes in Russian laws governing prospecting rights and the taxation of gold in i8I2, i826, and i846-54 caused fluctuations in Siberian gold production: Gold rushes, pp. 45, 47, 59. 33 This is likely to have occurred in Australia following the discovery of large copper deposits in South Australia in the mid and late i840S. The link was stronger in South Africa. Although minor gold discoveries had previously been made, the diamond discoveries of the i870s (the most notable was that at Kimberleyyf Z H U H L P S R U W D Q W W R W K H W L P L Q J R I W K H V X E V H T X H Q W J R O G G L V F R Y H U L H V I X U W K H U Q R U W K H V S H F L D O O y of the Rand field in i886. The success of the diamond industry gave an impetus to the search for gold that was already under way, and when the major discoveries were made, the enterprise and capital required for the exploitation of the gold-bearing reefs were drawn from those whose fortunes were made in the diamond mines. See Morrell, Gold rushes, pp. 3I3, 337-44. (? Economic History Society i994 This content downloaded from 142.104.160.169 on Mon, 06 Feb 2017 18:47:55 UTC All use subject to http://about.jstor.org/terms 294 BARRY EICHENGREEN and IAN W. MCLEAN II World gold production in the nineteenth century was dominated by two major phases of expansion starting in i848 and I89o.34 Extant data suggest that during the first half of the century annual average world output was trendless, fluctuating between i0 and 20 tonnes. Brazil, Colombia, Russia, and Mexico were the major producers.35 In the i850s output rose dramatically to over 200 tonnes annually as a result of the Californian and Australian discoveries in i848 and i 85 i. There followed a gradual decline in world production to approximately 150 tonnes in the i88os. The second major surge in the i89os reached 450 tonnes per annum by the end of the decade and levelled out at approximately 700 tonnes before the outbreak of the First World War. This expansion reflected both the revival of activity in Australia and the US and the emergence of major new producers (see table iyf . Table i. Major gold producers, 1850-i9I4 (average annual output, tonnesyf i850-9 i860-9 i870-9 i880-9 i890-9 I900-I4 US 83.4 70.6 6i.5 49.6 70.2 I32.3 Australia (76.8yf L R 6 Russia 25.4 25.I 37.I 34.8 38.4 39.9 New Zealand (0-3yf , , , , 4 Mexico (i.6yf , R I South Africa (4.5yf , , 2 Canada (2.0yf 8 Rhodesia (I.2yf , 6 Total i8i.o I7I.7 i58.0 I30.6 254.6 5I3.9 Note: figures in parentheses based on data for portions of the period indicated; Russian estimates before i870 based on 5-year average production estimates Source: Schmitz, Metal production, pp. 80-5. Closer analysis reveals a great diversity of national and regional experience. Although production in the US, Australia, and New Zealand exhibits the two-peaked pattern discernible in world output (see figures ia and ibyf W K e mid-century rush in New Zealand lagged behind the American and Australian gold rushes by more than a decade; and during the second surge, production in Australia and New Zealand peaked in the first years of the twentieth century before declining, while US production reached a plateau only after I905 and failed to turn down decisively before the First World War. Evidence for Russia suggests a different pattern. There were no major increases in output in the i850s, but some expansion during the i870s associated in part with the opening of fields in the area of the Amur River. No increase in activity occurred during the i89os, but following i906 there 34 The best introduction to the history of the world gold industry is Morrell, Gold rushes. An account of the mid-century experience in California may be found in Paul, California gold. Writings on the Australian rushes of the i850s include Maddock and McLean, 'Supply-side shocks', and Portus, 'Gold discoveries'. Useful accounts of nineteenth-century mining in these two countries which adopt a somewhat broader focus are Paul, Mining frontiers, and Blainey, Rush. 35 World production estimates cited in this paragraph are drawn from Report to the Congress, i, pp. i88-90, and Schmitz, Metal production, pp. 79-85. ?V Economic History Society i994 This content downloaded from 142.104.160.169 on Mon, 06 Feb 2017 18:47:55 UTC All use subject to http://about.jstor.org/terms SUPPLY OF GOLD UNDER THE PRE-I9I4 GOLD STANDARD 295 Tonnes 300- 250 - 200- South Africa 1501 100 .4, ~~~~Australia United Australia ~ ??0 -r States 0 I 1830 1840 1850 1860 1870 1880 1 890 1900 1910 Year Figure Ia. Gold production: Australia, South Africa, and the United States, i834-I9IS Source: Schmitz, Metal production, pp. 80-5 began a steady rise in annual production to a peak just before the First World War.36 In the i89os and i9oos these established producers were joined by such new suppliers as Canada, Mexico, Southern Rhodesia, and South Africa. After an initial coincidence of production experience, subsequent output patterns diverge. Mexican output peaks around i9io and subsequently declines; Canadian output peaks relatively early in i9oo, then declines by more than half before reviving strongly after i9io. Apart from a sharp fall associated with the Boer War of i899-I902 the output of South African and Southern Rhodesian mines continues to grow steadily throughout the First World War. Thus, even at the national level, gold industries do not seem to follow closely the pattern of expansion and slowdown observed when world output is viewed in aggregate. But even national trends may mislead as to behaviour at the regional level. For example, the four gold-producing Australian colonies rarely exhibit consistent supply behaviour (see figure 2yf 7 K H P L G - century gold rushes were concentrated in the south-eastern corner of the Australian continent. Although the initial gold discovery was made in New South Wales not far west of Sydney in early i85i, most gold during the i85os and i86os was found several hundred miles south in Victoria. By contrast, later discoveries were extremely widely scattered. Queensland and 36 Schmitz, Metal production, pp. 8o, -84; Morrell, Gold rushes, ch. 3. (C Economic History Society i994 This content downloaded from 142.104.160.169 on Mon, 06 Feb 2017 18:47:55 UTC All use subject to http://about.jstor.org/terms 296 BARRY EICHENGREEN and IAN W. MCLEAN Tonnes 60 - 50 - 40 ~~~~~~~~~~~~Russia 'I 2~~ ~~~~~ ~~~~~ ~~~ d I 4 IMexico 30 - I Canadai 20 - I~~~~~~~~~~~~~~~~~~~~~~~~~ FigureIb. Gld prductin: Caada, NexiowN Zealand , usa n Southern Rhodesia, I834-I9I5~~ ~Sother Fine ounces 3.0 - ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ Rodsi 10 I 1830 1840 1850 1860 1870 1880 1890 1900 1910 Year Figure i b. Gold production: Canada, Mexico, New Zealand, Russia, and Southern Rhodesia, W834-I9rn Source: Schmitz, Metal production, pp. 80-5 Fine ounces 3.0 - (millionyf 2.5 - 2.0 Victoria 1.50 Western iAustralia 1.0I 1850 1860 1870 1880 1890 1900 1910 1920 Year Figure 2. Gold .production: Australian colonies/states, i8Si-iqI4 Source: Kalix, Fraser, and Rawson, Australian mineral industry, pp. I76-7 ?D Economic History Society 1994 This content downloaded from 142.104.160.169 on Mon, 06 Feb 2017 18:47:55 UTC All use subject to http://about.jstor.org/terms SUPPLY OF GOLD UNDER THE PRE-I9I4 GOLD STANDARD 297 $(m.yf 8 26 24 22- colorado 20- 18 \! . 16- 14 I California 12 -' /Alaska IV INevada 10 1 8 - 6- 4- ./ S. Dakota 1880 1885 1890 1895' 1900 1905' 1910 Year Figure 3a. Value of gold production: US states and territories: major producers, i88i-i9io Source: US Department of Interior, Mineral resources of the United States $(m.yf 0 9 8- 7 6 5- 4- 3- 1 ~~~~~ ~. ... ....... Oregon 0 , .. ,,,,, ., ..* ' ' ''Utah 1880 1885 1890 1895 1900 1905 1910 Year Figure 3b. Value of gold production: US states and territories: minor producers, i88i-i9io Source: US Department of Interior, Mineral resources of the United States C Economic History Society 1994 This content downloaded from 142.104.160.169 on Mon, 06 Feb 2017 18:47:55 UTC All use subject to http://about.jstor.org/terms 298 BARRY EICHENGREEN and IAN W. MCLEAN Western Australia were added to the list of gold-producing colonies during the I87os and i88os. The dispersed gold fields of Queensland were responsible for two surges in the i870s and i88os but made no additional contribution to national gold production in the i89os when the rush to the west occurred. The fields of Western Australia were the major source of gold at the turn of the century. Victoria's contribution to the revival in activity after i890 was small, and only modest increases were reported in New South Wales.37 American data on gold production by state and territory also indicate diversity. Prior to the Californian discoveries the principal gold-producing states were in the South: Alabama, Georgia, North Carolina, South Carolina, Tennessee, and Virginia.38 None participated to any significant degree in either the mid-century expansion of output or the second increase in production after i890. California, of course, dominated the American gold industry at mid-century, although with the discovery of the Comstock Lode in i859 mining spread to Nevada. Production in California peaked in the early or mid I 85s.39 By the late i870s production in Nevada exceeded that in California, and no other producers approached their level of output. In the early i89os the output of California rose only moderately while that of Colorado increased dramatically (see figures 3a and 3byf 7 K H H T X D O O \ G U D P D W L c increase in Alaskan production lagged behind that of Colorado by nearly IO years, while the revival in Nevada occurred only after I905. By that time, the mines of Colorado were already producing markedly lower levels of gold. Varying supply behaviour is also exhibited by the minor gold producers. This diversity of output behaviour at both national and regional levels casts doubt on the notion that changes in the average price of commodities (which was proportional to the reciprocal of the relative price of gold, since the nominal price of gold was fixedyf V X I I L F H W R D F F R X Q W I R U I O X F W X D W L R Q V L n gold output. For the world as a whole there may appear to have been a broadly consistent relationship, as the two expansions in gold production in the i85os and i89os followed with a lag periods in which the relative price of gold rose (commodity prices declinedyf + R Z H Y H U D V Z H K D Y H V K R Z Q W K e fluctuations in world production levels do not coincide with fluctuations in gold production in all countries and regions." Is this because national and regional commodity price movements varied from those observed in international prices, but in a manner which maintains consistency with the trends in national and regional gold production? Or did region-specific non- price factors contribute significantly to the diverse pattern of national and regional gold production? For the price-theoretic explanation of fluctuations in gold supply to be sustained, the postulated (positiveyf U H O D W L R Q V K L S E H W Z H H n the real price of gold and gold output should be observed at the national 37 See Blainey, Rush, chs. 2-6, and Serle, Golden age, chs. I, 3, and 8. 38 Martin, 'U.S. gold'. 39 See the discussion of alternative estimates in Berry, Early California, ch. 5. 40 With respect to South African experience, Frankel, Gold, points out that the rates of return to investment to gold mining had to be competitive with those elsewhere in the economy. We do not here explore this aspect of the history of fluctuations in the gold mining industry. (C Economic History Society I994 This content downloaded from 142.104.160.169 on Mon, 06 Feb 2017 18:47:55 UTC All use subject to http://about.jstor.org/terms SUPPLY OF GOLD UNDER THE PRE-I9I4 GOLD STANDARD 299 and regional levels as well as for the world as a whole once one controls for these non-price factors. III For the US, figure 4 depicts the supply of gold and its relative price prior to I914. While casual observation might suggest the existence of a positive (laggedyf U H O D W L R Q V K L S E H W Z H H Q W K H W Z R Y D U L D E O H V L W L V X Q F O H D U Z K H W K H r the hypothesis that the supply of gold was a positive function of its relative price would survive formal statistical tests. Both Barro and Bordo have implied that the hypothesis of no relationship is rejected by the data. By contrast, in appendix note i we show that time series analysis of annual US gold production and prices between i844 and I9I3 indicates the importance of omitted non-price factors. The problem is that it is not obvious a prior which variables need to be incorporated. For guidance we turn in section IV to an analysis of the history of the major gold producing regions in the nineteenth century. But first, we examine the relationship between price and supply at a less aggregate level, namely, across I4 gold-producing American states at the end of the nineteenth century. A B 90 5- 80 4 reduction 50 1 _ Price 4 0 - _ 1 - - - - I - I 1830 1840 1850 1860 1870 1880 1890 1900 1910 Figure 4. US gold production and real price of gold, i835-i915 Source: Report to the Congress, i, tabs. SC-5, pp. I93-4, SC-i6, pp.2I9-25 It is not possible to replicate at the regional level the analysis of the influence of price on supply just reported for the US as a whole. Annual state-level estimates of gold production are available, though only from the late i870s. To obtain the real price of gold we need an index of commodity prices for each state with which to deflate the nominal price of gold (which C Economic History Society I994 This content downloaded from 142.104.160.169 on Mon, 06 Feb 2017 18:47:55 UTC All use subject to http://about.jstor.org/terms 300 BARRY EICHENGREEN and IAN W. MCLEAN was fixed and uniform across statesyf + R Z H Y H U Q H L W K H U Z K R O H V D O H Q R U U H W D L l price indexes are available for the gold-producing states during this period.4' Fortunately, there exists an alternative approach to capturing relative price effects. Rather than appealing to the general price level as an indirect measure of the (opportunityyf F R V W R I S U R G X F L Q J J R O G Z H X W L O L ] H D G L U H F t measure of the major components of costs: the cost of labour, in the form of the daily wage rate of miners (Wyf L Q W K H J R O G D Q G V L O Y H U L Q G X V W U \ R I W K e state (or territoryyf 7 K H W K H R U \ R I W K H I L U P V X J J H V W V W K D W J R O G S U R G X F W L R n (Gyf V K R X O G K D Y H E H H Q D Q L Q F U H D V L Q J I X Q F W L R Q R I W K H U D W L R R I S U R G X F W S U L F H s to variable cost.42 Where the real producer price of gold (the nominal price relative to the cost of productionyf Z D V K L J K V R V K R X O G K D Y H E H H Q W K H O H Y H l of gold production. We therefore use this measure of the real producer price to capture any price effects responsible for inter-temporal and inter- regional variations in production. Thus, we expected an inverse relationship between regional gold production and the daily wage rate of miners. In other words, the higher the regional cost level as reflected in miners' wages, the lower will be the real price of gold in the region (where real price is the fixed nominal price of gold deflated by the general price levelyf D Q d hence the lower the anticipated level of regional gold production.43 The expected negative relationship between regional gold production and the cost of labour (the miners' daily wage rateyf Z D V Q R W V X S S R U W H G E y evidence drawn from the mining censuses of i889 and I902, and for each of the I4 gold-producing states for which suitable data were obtained (see appendix note 2yf 7 K H G L U H F W S D U W L D O \f correlation between the two variables was positive rather than negative (table 2yf 7 K H W K U H H J R O G S U R G X F L Q J V W D W H s with the lowest costs (miners' wages below $i.50 per dayyf Z H U H D O V R W K e lowest volume producers at both the i889 and I902 mining censuses (Georgia and the Carolinasyf 7 K H U H P D L Q L Q J V W D W H V D O O H [ K L E L W H G P X F K K L J K H U F R V t structures (miners' daily wage between $2.50 and $4yf W K R X J K D Z L G e variation is observed in their levels of gold production. Further, the relationship between the change in costs and that in output between the two mining censuses was not consistent with the view that relative prices regulated the supply response. Theory would predict that states with the smallest increases (or largest declinesyf L Q Y D U L D E O H O D E R X U \f costs would show the largest increases in gold production. Those states experiencing the largest proportional increase in gold production over the period J889-I902 were not necessarily those in which costs decreased or 41 See Coelho and Shepherd, 'Differences'; M. R. Haines, 'A state and local consumer price index for the United States in i890', National Bureau of Economic Research, working paper on historical factors in long run growth, no. 2, May i989. 42 Note that we have reasoned from the static theory of the firm rather than from inter-temporal aspects of the firm's optimization problem such as those considered by Hotelling. Introducing the inter- temporal aspects of the problem would greatly complicate the analysis without changing the essential point. Hotelling's Rule, like our simple formulation, suggests that at a point in time the production of gold will be highest where costs are lowest. See Levhari and Pindyck, 'Pricing'. 43 One possibly important element in mining costs is the cost of capital. Existing studies indicate that the west as a whole recorded higher interest rates (on short- or long-term deposits and mortgages on houses or farmsyf W K D Q W K H U H V W R I W K H H F R Q R P \ D W W K H H Q G R I W K H Q L Q H W H H Q W K F H Q W X U \ ' D Y L V , 'Investment market', and Snowdon, 'Mortgage rates'. It is unclear by how much the cost of capital to miners and mining companies varied by state within the west; an examination of the capital market in (C Economic History Society I994 This content downloaded from 142.104.160.169 on Mon, 06 Feb 2017 18:47:55 UTC All use subject to http://about.jstor.org/terms SUPPLY OF GOLD UNDER THE PRE-I9I4 GOLD STANDARD 30I Table 2. Gold output and related variables. I4 US states and territories: correlation coefficients General Rural Average annual gold production (varying periodsyf Miner's unemployment settlement wage rate rate ratio (Wyf 8 5 \f (S:yf * L R \f G (5yf * \f G (oyf UR o.633a S -0.928b -o0536b G (ioyf D , 5 G (5yf , R b G (3yf , E R b G (oyf R , 4 4 E R E R b G (fyf D " D R E R E , E b Notes: a and b indicate significance at the Ioyb D Q G , \b levels respectively. W is the average wage of miners as recorded in the mining censuses of i889 and i9o2; UR is the ratio of those recorded as unemployed (for 7 to I2 months of the year preceding the censuses of i890 and i902yf W R W K H W R W D O J D L Q I X O O \ H P S O R \ H G ; S is the acres in farm land I0 years prior to the census as a proportion of the maximum acres of farm land recorded at any census through I970; and G represents gold production; G (Ioyf * \f, and G (3yf U H I H U W R W K H D Y H U D J H D Q Q X D O S U R G X F W L R n over the IO, 5, and 3 years respectively following the year of the mining census. G (oyf U H I H U V W R S U R G X F W L R Q L Q W K H F H Q V X s year. G (fyf U H I H U V W R I X W X U H J R O G S U R G X F W L R Q G H I L Q H G W R F R Y H U \ H D U V W R , \f. The production data for the I4 states and territories were pooled across the two mining censuses of i889 and I902, yielding 28 observations. For a full explanation of G. W, UR, and S. see appendix, n. 2 grew most slowly. Instead, three states experienced declines in both costs and production, while six states saw increases in both." We conclude that no close or systematic relationship is evident either in time series for the US as a whole before 19I4, or in the gold industries of I4 producing states at the end of the nineteenth century.45 Influences other than price must have played an important role in accounting for variations in the output of gold, both across time and between regions. To such influences we now turn. IV Gold-producing countries and regions tended to share certain broad characteristics. Nearly all the important gold discoveries in the century the three Pacific Coast states at this time suggests the convergence of interest rates on those of San Francisco: see Odell, 'Integration'. 4It is pertinent to note that, according to information in the mining censuses, the average daily wage of miners exhibited considerable variation across mining states and territories. In i889 the lowest recorded was $i.02 (North Carolinayf W K H K L J K H V W 1 H Y D G D \f, and the US average $3.I2. In I902 the comparable figures were $i.ii (North Carolinayf L 0 R Q W D Q D \f, and $3.11 for the national average. These inter-state variations may reflect not only differences in the cost of labour in the industry, but also in the capital intensity of mining (alluvial mining generally being less capital intensive than either hydraulic or quartz mining operationsyf D Q G D O V R L Q W K H U L F K Q H V V R I W K H R U H G H S R V L W % R W K W K H V H I H D W X U H V R I J R O G P L Q L Q J F R X O d differ markedly even between mining areas within a state. Analysis of these additional influences on gold mining activity at the local level is beyond the scope of the present inquiry. 4 The possibility arises that the link between the price and supply of gold is attenuated by the increased use of substitutes for (more accurately, supplements toyf J R O G D V L Q W H U Q D W L R Q D O P R Q H W D U y reserves. It is clear from Lindert, 'Key currencies' (esp. fig. I, p. 24yf W K D W V W H U O L Q J D Q G R W K H U I R U H L J n exchange reserves were comparatively unimportant relative to official gold reserves until after i900. Even then, their growing significance would affect the (presumably weakeryf O L Q N U X Q Q L Q J I U R P J R O G W o the price level, rather than having any clear effect on the link running from the price level to gold production or stocks, and it is with the latter causal influence that the present analysis is concerned. The same point can be made with respect to domestic gold 'substitutes' (such as bank depositsyf . (C Economic History Society I994 This content downloaded from 142.104.160.169 on Mon, 06 Feb 2017 18:47:55 UTC All use subject to http://about.jstor.org/terms 302 BARRY EICHENGREEN and IAN W. MCLEAN before I9I4 occurred in the regions of recent European settlement. In an era before the search for minerals was systematically based on the science of geology, the existence of gold deposits could be detected only from surface or near-surface evidence. Thus the probability -of a discovery was higher in areas undergoing initial exploration and agricultural settlement. 'The more men walk over ground that is rich in minerals the higher the chance that they will find and recognize the minerals', as Blainey puts it. 'The nearer they are to settlements and towns the higher the chance that someone will explore and mine those deposits that come to the surface of the earth. All this is obvious and therefore rarely perceived.'46 Sometimes gold discoveries occurred during the exploration of a region but before its settlement. The gold and silver deposits discovered at Comstock Lode in western Nevada in i859 reflected the eastward extension of exploration from California as mining prospects declined in the golden state. Similarly, the discoveries that produced the rushes to Kalgoorlie in Western Australia and to Alaska in the i89os were in territory in the process of exploration and before significant European occupation. Other major gold finds followed shortly after initial occupation of the region. The fur trade was important to the opening up of areas of Siberia and Canada subsequently found to contain gold. In northern California there had occurred a tentative occupation of some areas of land by a small and scattered population of ranchers before the gold discovery at Sutter's Mill in i848. Similarly, in New South Wales and Victoria in i85i and later in Otago in New Zealand in I 865, gold was found in areas where sheep had been grazed on an open- range basis with low livestock densities for a number of years-fewer than I0 years in the case of Otago, perhaps I0 to 20 years in Victoria, and rather more in New South Wales. The subsequent northward spread of the thin pastoral occupation of the semi-arid inland regions of eastern Australia was accompanied by a series of discoveries in Queensland from the late i850s to the i88os. Likewise in the mountain areas between California and the Great Plains, the mining frontier moved sometimes ahead of, sometimes simultaneously with, the extension of pioneer ranchers. By contrast, few major discoveries were made in areas that had experienced close population settlement and intensive agricultural occupation over a long period. Even within the gold-producing countries, the areas which had moved beyond the frontier stage and become more densely populated as a result of the mid-century surge in gold production (California and south- eastern Australiayf F R Q W U L E X W H G O L W W O H W R W K H U H Y L Y D O L Q P L Q L Q J D F W L Y L W \ \ H D U s later. The timing of regional settlement was not the only cause of differences in regional gold production histories before I9I4. Blainey has argued on the basis of a detailed examination of Australian mineral discoveries that local business conditions were of importance as well, though his dating has been challenged.47 Australian colonies in the late nineteenth century did 46Blainey, Rush, p. 5. 47 Blainey, 'Mineral discovery'; Morrissey and Burt, 'A note'; Blainey, 'Rejoinder'. See also Davies, 'Blainey revisited'. (Cyf ( F R Q R P L F + L V W R U \ 6 R F L H W \ , 4 This content downloaded from 142.104.160.169 on Mon, 06 Feb 2017 18:47:55 UTC All use subject to http://about.jstor.org/terms SUPPLY OF GOLD UNDER THE PRE-I9I4 GOLD STANDARD 303 not always experience economic fluctuations simultaneously; thus it could be misleading to treat the country as a single economy. Blainey argues that local unemployment, in addition to lowering wage costs, encouraged those who were out of normal employment for substantial periods to turn to prospecting.48 The preceding discussion suggested that the probability that gold would be discovered in a region was higher at early stages of settlement and development. One measure of settlement is the extent of rural occupation of a territory. This appears appropriate given the emphasis in the literature on the importance of visual observation for indications of the presence of gold. Again using regional evidence for the US at the turn of the century, the level of settlement (Syf Z D V F D O F X O D W H G I U R P L Q I R U P D W L R Q F R O O H F W H G D W W K e decennial censusyf D V W K H U D W L R R I O D Q G L Q I D U P V L Q H D F K V W D W H W R W K H P D [ L P X m area in farms recorded for that state through I970, and lagged (io yearsyf to accommodate the lapsed time between settlement, discovery, and production.49 In general, the US regional evidence supports the existence of a negative relationship between gold output and prior settlement. Most gold produced in the US in the late nineteenth and early twentieth centuries came from the newly settled western regions. The contribution from states east of the Mississippi was negligible. On one level, this feature of gold mining in the US simply illustrates the uneven regional endowment of gold deposits. It also reflects a circumstance of discovery characteristic of the period: it was necessary that there be evidence at or near the surface, and hence major new finds were less likely in the longer-settled regions. Secondly, within the gold-producing states, the three long-settled producers (Georgia and the Carolinasyf U H F R U G H G W K H O R Z H V W O H Y H O V R I R X W S X W D Q G R f output expansion. Three-quarters of the growth in American gold production between i890 and i9io came from just three locations: Colorado, Alaska, and Nevada, all of which were in relatively early stages of settlement. California, the fourth most important producer, was by this time more 'settled' than these regions, and it recorded a growth in output only half that of the other three. The direct correlation between the change in gold production and the prior level of settlement is negative, though not significant at the io per cent level. Similarly, we observe a consistently negative (though weakyf F R U U H O D W L R Q E H W Z H H Q W K H O H Y H O R I J R O G S U R G X F W L R Q Y D U L R X V O \ D Y H U D J H G - and the prior level of settlement in the I4 states and territories for which we have the required information (see table 2yf 7 K X V W K H U H D S S H D U V W R E H a systematic (though not tightyf U H O D W L R Q V K L S E H W Z H H Q J R O G S U R G X F W L R Q D Q G S U L R r levels of settlement, as suggested by writers such as McCarty.50 We attempt to capture the influence of local business conditions with the proportion of the labour force in each state which was recorded at the censuses 48 Blainey, 'Mineral discovery', pp. 308-9. 49 In i870, for example, the settlement proxy (acres of land in farms in each gold producing state as a percentage of the maximum recordedyf Z D V I R U * H R U J L D I R U 1 R U W K & D U R O L Q D D Q G I R r South Carolina. California recorded 30.2, Oregon 11.2, and the other nine western states and territories all recorded less than 3.5. 50 McCarty, 'Suggestions'. (C Economic History Society I994 This content downloaded from 142.104.160.169 on Mon, 06 Feb 2017 18:47:55 UTC All use subject to http://about.jstor.org/terms 304 BARRY EICHENGREEN and IAN W. MCLEAN of I890 and I900 as unable to find work in their principal occupation for more than six months at any time during the census year. For this period and sample of states, there is some evidence indicating a positive relationship between recorded unemployment (so definedyf D Q G J R O G S U R G X F W L R Q D V V X J J H V W H d by Blainey.51 Although there is considerable variation in outcomes, no large producer of gold (more than $5 millionyf K D G D Q X Q H P S O R \ P H Q W U D W H E H O R w 2 per cent. The average (unweightedyf X Q H P S O R \ P H Q W U D W H D F U R V V W K H , V W D W H s was I.75 and 2.9 in i890 and i900 respectively. Further, the correlation between regional gold production and unemployment is positive whether gold production is taken as that in the year of the census, or averaged over several years thereafter to reflect a lagged supply response to the depressed local economic conditions (table 2yf . This series of bivariate analyses of price and non-price determinants of supply behaviour in the gold industry is appropriate to an initial evaluation of the competing claims in the literature. Two features of the regional data on the gold industry in the US at this time bedevilled more formal attempts to choose between the alternative hypotheses. First, there existed a very high and negative correlation (-0.93yf E H W Z H H Q W K H O H Y H O R I V H W W O H P H Q W 6 \f and the proxy measure of the regional price level (the nominal wage rate received by miners, Wyf 7 K L V L V X Q V X U S U L V L Q J $ I U R Q W L H U U H J L R Q L Q W K H H D U O y stages of settlement and development, could be expected to have relatively high costs (including wagesyf D Q G W K X V K L J K F R P P R G L W \ S U L F H V 6 H F R Q G O \ , the dominant influence on current levels of regional gold production was levels of production in the recent past, as is clear from the time series in figure 3: gold production is not generally subject to wide fluctuations from year to year. The direct correlation between the average level of gold output for the five years before and the five years after the two mining censuses for each of the I4 states is o.98. Further, current output was closely and positively related to future output (see table 2yf 7 K D W L V R X W S X W O H Y H O V Y D U L H d across regions as if producers knew the size of their gold reserves: higher levels of annual output are thus observed where the deposit (future productionyf L V O D U J H U 7 K H V L ] H D Q G R W K H U F K D U D F W H U L V W L F V R I D U H J L R Q V J R O d deposits appear to have a major influence on the level of regional output. This is not an industry exhibiting dramatic supply responses to short-run changes in general economic conditions."2 51 Blainey, 'Mineral discovery'. 52 We report a typical result from our attempts to employ multiple regression analysis in accounting for regional variations in gold supply behaviour. The average level of gold production in each region over the IO years after a census year is the dependent variable (Gioyf 7 K H L Q G H S H Q G H Q W Y D U L D E O H V D U H , for each region, the average daily wage paid to miners (Wyf V H U Y L Q J D V D S U R [ \ I R U W K H U H D O \f price of gold; the long-term 'unemployment' rate at the census (URyf W K H O H Y H O R I V H W W O H P H Q W , 2 \ H D U V S U L R U W o the census (Syf W K H O D J J H G G H S H Q G H Q W Y D U L D E O H * L \f; and finally, a measure of the operating profit ratio for mines in each region (OPRyf G H V L J Q H G W R F D S W X U H R W K H U O R F D W L R Q V S H F L I L F L Q I O X H Q F H V R Q W K H U H W X U Q V W o gold mining, for example, the yield of an ore body. This model is estimated on information for the I4 gold-producing states and territories, pooling data from the censuses of i890 and i900 to give 28 observations. Gio = 3.I7I - I.920W + o.467UR - 0.037S + 2.3430PR + I.287G-I (0-53yf , , \f (I.03yf R \f (i.89yf , 2 \f (C Economic History Society I994 This content downloaded from 142.104.160.169 on Mon, 06 Feb 2017 18:47:55 UTC All use subject to http://about.jstor.org/terms SUPPLY OF GOLD UNDER THE PRE-I9I4 GOLD STANDARD 305 V The discovery and production of gold in the nineteenth century was of major importance both to the rapidly expanding international economy which depended on it for monetary reserves and to the regions of recent European settlement in which the gold was mined. To the historian of the world economy after i85o and the student of economic development in the US, Australia, New Zealand, Canada, or Southern Africa, the determinants of gold production are of considerable interest. These determinants are equally relevant to discussions of a commodity-based international monetary system, since the supply characteristics of the reserve commodity will influence the behaviour of money supplies and general price levels. It is not surprising, therefore, to find that the supply of gold is a consideration in debates over the operation of the international gold standard. What is surprising is the dearth of useful information on this crucial question. In attempting to correlate writings on international monetary economics and economic development in areas of recent settlement, we have examined the role of relative prices and other economic factors in influencing the growth of gold production. With respect to world gold output, we demonstrated that variations over time in national production in many of the principal gold producers did not follow the pattern suggested by the long swings in the real price of gold (and reflected to some extent in aggregate world gold outputyf ) X U W K H U W K H U H J L R Q D O S U R G X F W L R Q W U H Q G V Z L W K L n two of the major producing countries, Australia and the US, also frequently departed from that which would be anticipated had the relative price been the dominant influence on gold production. In this article we have also considered the determinants of gold supply in the US at the end of the nineteenth century and beginning of the twentieth. In an analysis of regional trends among I4 gold-producing states we find varying degrees of support for the influence of real gold prices, the prior level of regional settlement, and local business conditions. Particularly striking was the positive association between, on the one hand, the degree to which a state had been occupied for agricultural purposes, and on the other the extent to which that state participated in the expansion of US gold production at the end of the century. Clearly, an explanation of the determinants of gold supply at this time which emphasizes one set of factors at the expense of others is overly simplistic and likely to mislead. In particular, an account which focused exclusively on the influence of the real price of gold would obscure as much as it revealed about the determinants of nineteenth-century gold production. University of California at Berkeley University of Adelaide where R2 = o.84; t-statistics are reported in parentheses. All independent variables have the expected sign, but most are not statistically significant; previous output levels dominate 'current' regional gold production. This general result persisted in.a variety of specifications of the 'model', and where different averages or transformations of the data were employed. (Cyf ( F R Q R P L F + L V W R U \ 6 R F L H W \ , 4 This content downloaded from 142.104.160.169 on Mon, 06 Feb 2017 18:47:55 UTC All use subject to http://about.jstor.org/terms 306 BARRY EICHENGREEN and IAN W. MCLEAN APPENDIX: NOTES ON STATISTICAL SOURCES AND METHODS I Time series evidence on United States gold production In this appendix we use time series regressions to analyse the relationship between annual aggregate US gold production and the real price of gold (the price of gold per ounce in US dollars divided by the wholesale price indexyf 7 K H E D V L F V S H F L I L F D W L R Q I R O O R Z V 5 X V K Z K R V e results are invoked by Barro to substantiate the existence of a significant well-behaved positive price elasticity of gold supply. Our own analysis points strongly to misspecification: in particular, to the omission of important non-price determinants of supply. Our implementation of Rush's model departs from this previous work in two respects. First, we estimate the relationship over the entire pre-World War I period for which price and production data are provided in the appendix to the Report to the Congress of the Gold Commission, rather than the i897-I913 subperiod on which Rush concentrates. (Estimates for a variety of other subperiods produced results consistent with those reported below.yf Second, we relate US gold prices to US gold production rather than, as does Rush, to world production. US output estimates are more reliable. Moreover, the appropriate explanatory variable for world gold output is world gold prices, not US gold prices, which will differ due to costs of transportation, insurance, etc. The basic specification produced the following result: log (output Jyf R R O R J S U L F H \f + 0.95 log (output t - 0 (i.98yf \f (41.95yf R2 = o.96 DW = 1.33 with t-statistics in parentheses. Durbin's h provides a strong indication of positive serial correlation. To test whether this serial correlation reflects omitted variables, we followed two procedures suggested by Maddala.A' The first is to use the RESET test of Ramsey, regressing the estimated residuals on various powers of the fitted dependent variable and testing whether or not the estimated coefficients are significant." (We used log (outputyf W R W K H V H F R Q G W o fifth powers, inclusive.yf $ O O I R X U F R H I I L F L H Q W V G L I I H U H G V L J Q L I L F D Q W O \ I U R P ] H U R D W W K H S H r cent level, which in conjunction with the significant Durbin's h provides evidence that the serial correlation reflects omitted variables. The second procedure, due to White, involves regressing the estimated residuals squared on all the explanatory variables and their cross products and testing the significance of the coefficients.56 Two of the three independent variables differed significantly from zero at the 95 per cent level, in conjunction with Durbin's h again providing evidence that the serial correlation reflects omitted variables. We take these results to indicate the importance of including nonprice factors in the analysis of nineteenth-century gold production. 2 Variables used in state-level supply analysis The variables used in the state-level analysis of the determinants of gold supply are described in the following notes and in table Ai. Gold production by state Our estimates of gold production by state and territory are those provided by the Director of the Mint, as reported in Mineral resources of the United States (Department of the Interioryf D Q G L Q W K H 6 W D W L V W L F D O D E V W U D F W R I W K H 8 Q L W H G 6 W D W H V 7 K H V W D W H s are Arizona, California, Colorado, Georgia, Idaho, Montana, Nevada, New Mexico, North Carolina, Oregon, South Carolina, South Dakota, Utah, and Washington. Differences between the Mint estimates and the mines returns of the US Geological Survey, which arise mainly from the difference in the stage of production at which output was measured, are typically small in any one year and negligible when cumulated over several years (see the 53 See above, n. 24. 54 Maddala, Introduction, pp. 208-9. 55 Ramsey, 'Estimation'.. 56 White, 'Heteroskedasticity'. (C Economic History Society i994 This content downloaded from 142.104.160.169 on Mon, 06 Feb 2017 18:47:55 UTC All use subject to http://about.jstor.org/terms SUPPLY OF GOLD UNDER THE PRE-I9I4 GOLD STANDARD 307 Table Ai. Gold production and related indicators in 14 US states and territories Unemployment Gold production Miner's WOage Settlement rate ($'oooyf S H U G D \ \f (yb \f (yb \f i889/1i890 Arizona goo 3.I7 0.3 I.9 California I3,000 2.74 43.9 2.3 Colorado 4,000 3.o8 3.0 i.6 Georgia I07 I .05 96.6 0.5 Idaho 2,000 3.59 2. I 2.5 Montana 3,500 3.48 o.6 2.2 Nevada 3,000 3.60 4.9 3.8 New Mexico I,000 3.I5 I .3 I.6 North Carolina I45 I.02 94.I 0.7 Oregon I,200 3.i6 i9.8 I.9 South Carolina 45 1.I5 83.o 0.5 South Dakota 2,900 3.49 6. i I.2 Utah 500 3.03 5.I I.9 Washington I75 3-43 7.4 I.9 I900/I902 Arizona 4,o83 3.28 3. I 3.8 California i6,89i 2.88 56.7 3.7 Colorado 27,693 3.I7 I I.9 3.2 Georgia I24 i.i8 93.5 I. I Idaho I,869 3.42 8.5 4.2 Montana 4,744 3.6I 3.0 3.4 Nevada 2,964 3.4I I 5.2 6.7 New Mexico 688 2.58 i.6 2.2 North Carolina 56 I.II 95.3 0.3 Oregon i,8i8 2.90 32.5 2.9 South Carolina 47 I.30 8I.3 0.9 South Dakota 6,479 3.57 25.0 2.0 Utah 3,690 2.88 I0.3 3.6 Washington 58o 3.35 2I.9 2.6 Notes and sources: see appendix, n. 2 discussion in Mineral resources, I9I0, pt. I, pp. I26-30yf / D U J H U G L V F U H S D Q F L H V H [ L V W E H W Z H H n the gold production estimates of the mining censuses and those of the Geological Survey, which persisted despite the two agencies' attempts at cooperation. For discussion, see Thirteenth census of the United States, p. I4. Miners' wages The average wage for miners in gold and silver production for I889 was obtained directly from Eleventh census of the United States, p. 59; and for I902 was calculated from information provided in the Twelfth census of the United States, pp. 580-i. It might be objected that miners' wage rates do not accurately capture production costs; however, census estimates indicate that wages were a major component of operating expenditures (68 and 75 per cent in I889 and I902 respectivelyyf D Q G W K D W P L Q H U V Z H U H L Q W X U Q D V L J Q L I L F D Q W S D U W R f total employment in the industry (52 and 50 per cent respectivelyyf $ Q R W K H U S R W H Q W L D l objection is that miners' wages are themselves influenced by changes in industry output, giving rise to simultaneity bias. If higher levels of output boosted labour demand, which in turn inflated wages, a positive coefficient might result, obscuring any negative relationship running from costs to levels of production. However, the gold-mining industry was only a small sector of any one state economy at this time. Employment in gold and silver mining as a proportion of total employment in each gold-producing state ranged in I889/90 from 0.04 per cent to I4.6 per cent, with an unweighted average of 4.8 per cent; and in I900/2 from 0.03 per cent to 6.I per cent; with an unweighted average of 2.4 per cent. On both (? Economic History Society 1994 This content downloaded from 142.104.160.169 on Mon, 06 Feb 2017 18:47:55 UTC All use subject to http://about.jstor.org/terms 308 BARRY EICHENGREEN and IAN W. MCLEAN occasions, South Carolina recorded the lowest figure, Nevada the highest. (The numerator in these calculations was derived from the mining censuses of i889 and I902; the denominator from Historical statistics, Series D 26.yf 'Unemployment' ratio The censuses of i890 and i900 did not record the number of people out of work on the day of the census but sought evidence that unemployment was experienced at some time during the preceding year along with an indication of duration: I to 3 months, 4 to 6 months, or 7 to I2 months. Unemployment was defined as 'time unemployed with regard to the principal occupation in which persons so reported were usually engaged and upon which they depended chiefly for a livelihood'. When the total number who were unemployed by this definition at some time during the preceding year is compared with the number of labour force participants defined as 'total persons IO years of age and over engaged in gainful occupations', the resulting 'unemployment rate' is I5.I per cent in i890 and 22.3 per cent in i900. Clearly this measure partly reflects labour market turnover and is not comparable with present-day definitions of the unemployment rate. Our goal was to construct a measure of the state of the labour market indicative of the opportunity cost of prospecting; therefore we thought it appropriate to disregard short spells of presumably frictional unemployment and to concentrate on the long-term unemployed. We used those 'unemployed' for 7 to I2 months out of the last year as a proportion of all persons gainfully occupied. For i890, see Eleventh census of the United States, tables 77, ioi; for i900, see Twelfth census of the United States, tables xxII-xxiv and 40. Settlement The 'settlement' proxy is based on census data reporting the acres of land in farms in each state at each census from i870 to I900. In order to convert these to standardized measures of the 'extent of settlement', acres of farm land were divided by the maximum acres of farm land recorded in the state at any census through I970. These data are conveniently summarized in Historical statistics, Series KI7-8i, P. 460. Footnote references Barro, R. J., 'Money and the price level under the gold standard', Econ. J., 89 (I979yf S S , . Berry, T. S., Early California: gold, prices, trade (Richmond, Va., i984yf . Blainey, G., 'A theory of mineral discovery: Australia in the nineteenth century', Econ. Hist. Rev., 2nd ser., XXIII (I970yf S S , . Blainey, G., 'A theory of mineral discovery: a rejoinder', Econ. Hist. Rev., 2nd ser., xxvi (I973yf , pp. 506-9. Blainey, G., The rush that never ended: a history of Australian mining (Melbourne, i963yf . Bordo, M. D., 'The gold standard: the traditional approach', in M. D. Bordo and A. J. Schwartz, eds., A retrospective on the classical gold standard, i82I-I93I (Chicago, i984yf S S , , . Bordo, M. D., 'The classical gold standard: some lessons for today', Fed. Res. Bank of St Louis Rev., 63 (May, i98iyf S S , . Bordo, M. D. and Schwartz, A. J., eds., A retrospective on the classical gold standard, i82I-I93I (Chicago, I984yf . Cairnes, J. E., Essays in political economy, theoretical and applied (i873yf . Cantillon, R., Essai sur la nature du commerce en general (New York, I93I; ist edn. I755yf . Coelho, P. and Shepherd, J., 'Differences in regional prices: the United States, i85i-i880', J. Econ. Hist., XXXIV (I974yf S S , , . Cooper, R. N., 'The gold standard: historical facts and future prospects', Brookings Pap. Econ. Activity, I (i982yf S S , . Davies, M., 'Blainey revisited: mineral discovery and the business cycle in South Australia', Aust. Econ. Hist. Rev., 25 (i985yf S S , , . Davis, L. E., 'The investment market, i870-I914: the evolution of a national market', J. Econ. Hist., xxv (i965yf S S . Eichengreen, B., ed., The gold standard in theory and history (i985yf . Fisher, I., The purchasing power of money (New York, i965; ist edn. I922yf . Frankel, S. H., 'Gold and international equity investment', Institute of Econ. Affairs, Hobart Pap., 45 (I969yf . Hardy, C. 0., Is there enough gold? (Washington, D.C., I936yf . Hirsch, F., 'Influences on gold production', I.M.F. Staff Pap., I5 (i968yf S S . Kalix, Z., Fraser, L. M., and Rawson, R. I., Australian mineral industry: production and trade, i842- I964 (Canberra, i966yf . Keynes, J. M., A tract on monetary reform: vol. 4 of the collected writings of John Maynard Keynes (I97I; ist edn. I923yf . (C Economic History Society i994 This content downloaded from 142.104.160.169 on Mon, 06 Feb 2017 18:47:55 UTC All use subject to http://about.jstor.org/terms SUPPLY OF GOLD UNDER THE PRE-I9I4 GOLD STANDARD 309 Levhari, D. and Pindyck, R. S., 'The pricing of durable exhaustible resources', Qu. J. Econ., 96 (i98iyf S S . Lindert, P. H., 'Key currencies and gold, i9oo-I9I3', Princeton Stud. Internat. Finance, 24 (i969yf . Maddala, G. S., Introduction to econometrics (New York, i988yf . Maddock, R. and McLean, I., 'Supply-side shocks: the case of Australian gold', J. Econ. Hist., XLIV (I984yf S S , . Marshall, A., Money, credit and commerce (I923yf . Martin, D. A., 'U.S. gold production prior to the California gold rush', Exp. Econ. Hist., I3 (I976yf , PP. 437-49. McCarty, J., 'Suggestions for an economic history of North American mining in the nineteenth century', in N. Harper, ed., Pacific Circle: Proceedings of the Second Biennial Conference of the Australian and New Zealand American Studies Association (St Lucia, Queensland, i968yf S S J R , , , . Mill, J. S., Principles of political economy (New York, i96i; ist edn. i865yf . Morrell, W. P., The gold rushes (2nd edn. i968yf . Morrissey, M. J. and Burt, R., 'A theory of mineral discovery: a note', Econ. Hist. Rev., 2nd ser., XXVI (I973yf S S . Odell, K. A., 'The integration of regional and interregional capital markets: evidence from the Pacific coast, i883-I9I3', J. Econ. Hist., XLIX (i989yf S S , R . Paul, R. W., California gold: the beginning of mining in the far west (Lincoln, Neb., i965; ist edn. I947yf . Paul, R. W., Mining frontiers of the far west, I848-I880 (New York, i963yf . Portus, G. V., 'The gold discoveries, i850-i86o', in The Cambridge history of the British empire, 7, Pt. I (Cambridge, I933yf S S , . Ramsey, J. B., 'Estimation of residuals', Proceedings of the Fourth Berkeley Symposium on Mathematical Statistics and Probability (Berkeley, Calif., i96iyf S S , . Ricardo, D., 'Proposals for an economical and secure currency; with observations on the profits of the Bank of England as they regard the public and the proprietors of bank stock', in P. Sraffa, ed., The works and correspondence of David Ricardo (Cambridge, I95I; Ist edn. i8i6yf . Rockoff, H., 'Some evidence on the real price of gold, its costs of production, and commodity prices', in M. D. Bordo and A. J. Schwartz, eds., A retrospective on the classical gold standard, I82I-I93I (Chicago, I984yf S S , . Schmitz, C. J., World non-ferrous metal production and prices, I700-I976 (I979yf . Serle, G., The golden age: a history of the colony of Victoria, i85i-i86i (Melbourne, i963yf . Snowdon, K. A., 'Mortgage lending and American urbanisation, I880-I890', J. Econ. Hist., XLVIII (i988yf S S . Thornton, H., An inquiry into the nature and effects of the paper credit of Great Britain (Fairfield N.J., I979; Ist edn. I802yf . White, H., 'A heteroskedasticity consistent covariance matrix estimator and direct test of heteroskedasti- city', Econometrica, 48 (ig80yf S S , . Wicksell, K., Interest and prices (New York, i965; Ist edn. I898yf . Official publications League of Nations, First interim report of the Financial Committee of the Gold Delegation (Geneva, I930yf . Report to the Congress of the Commission of the Role of Gold in the Domestic and International Monetary Systems (Washington, D.C., i982yf . United States Bureau of the Census, Eleventh census of the United States, i890: report on the population of the United States i890, I, Pt. 2 (Washington, D.C., i897yf . United States Bureau of the Census, Eleventh census of the United States, i890, 7: report on mineral industries of the United States (Washington, D.C., i892yf . United States Bureau of the Census, Historical statistics of the United States: colonial times to I970 (Washington, D.C., I975yf . United States Bureau of the Census, Statistical abstract of the United States (Washington, D.C., annual publicationyf . United States Bureau of the Census, Thirteenth census of the United States, i910, ii: mines and quarries, i909 (Washington, D.C., I9I3yf . United States Bureau of the Census, Twelfth census of the United States, i9oo: special report: mines and quarries, I902 (Washington, D.C., I905yf . United States Bureau of the Census, Twelfth census of the United States, i9oo: special reports: occupations (Washington, D.C., I904yf . United States Geological Survey, Mineral resources of the United States, i910, pt. I: metals (Washington, D.C., i9iIyf . ( Economic History Society i994 This content downloaded from 142.104.160.169 on Mon, 06 Feb 2017 18:47:55 UTC All use subject to http://about.jstor.org/terms