He is grateful to Zorina Khan and James Lin for excellent research assistance, and to Jeremy Atack, Francesca Bray, Paul David, Lance Davis, Stanley Engerman, Albert Fishlow, Claudia Goldin, Stephen Haber, Chris Hall, Steven Lubar, Douglass North, Nathan Rosenberg, Manuel Trajtenberg, Michael Waldman, Thomas Weiss, Finis Welch, Mary Yeager, and participants in seminars at the NBER, the Smithsonian Institution, Stanford University, UC Berkeley, UC Santa Cruz, UCLA, and the University of Hong Kong for valuable comments and discussions. He was a Fellow at the Center for Advanced Study in the Behavioral Sciences during the writing of the paper, and received research support from the Center, the National Science Foundation, the California Institute of Technology, and the Institute of Industrial Relations and the Academic Senate at UCLA.Google Scholar

¹ Simon Kuznets defined inventive activity “as being concerned with technical inventions, i.e. new combinations of existing knowledge … potentially useful in economic production and resulting from a mental performance above the average.” See Kuznets, Simon, “Inventive Activity: Problems of Definition and Measurement,” in Nelson, Richard R., ed., The Rate and Direction of Inventive Activity (Princeton, 1962).Google Scholar Research on inventive activity is typically framed in terms of either the outputs produced or the inputs consumed in the process. The outputs encompass the flow of new ideas, techniques, equipment, and products that are discovered and/or developed. The inputs include the costs of labor and other resources involved in the search for inventions. There appears to be a general consensus among economists who have analyzed modern patent data that there is a more regular relationship between the number of patents and the value of inputs to inventive activity than with the value of outputs; for example, see Pakes, Arid, “On Patents, R & D, and the Stock Market Rate of Return,” Journal of Political Economy, 93 (04 1985), pp. 390–409.CrossRefGoogle ScholarThis article is motivated by a concern with the circumstances under which individual households and firms came to commit much larger amounts of resources on a per capita basis to inventive activity, and the interpretations offered tend to be cast in terms of the allocation of resources. Nevertheless, the presumption is that, on average and over the period in question, the level of the outputs produced varied roughly with the consumption of inputs, and that the substantive implications of the findings are not sensitive to the perspective adopted.Google Scholar

² Some recent examples of general histories of this genre are Cochran, Thomas C., Frontiers of Change: Early Industrialism In America (New York, 1981);Google ScholarHindle, Brooke and Lubar, Steven, Engines of Change: The American Industrial Revolution, 1790–1860 (Washington, D.C., 1986);Google ScholarHounshell, David A., From the American System to Mass Production, 1800–1932 (Baltimore, 1984);Google Scholar and Hindle, Brooke, Emulation and Invention (New York, 1981).Google Scholar

³ See Schmookler, Jacob, “The Level of Inventive Activity,” Review of Economics and Statistics, 34 (02 1954), pp. 183–90;CrossRefGoogle ScholarSchmookler, Jacob, Invention and Economic Growth (Cambridge, MA, 1966);CrossRefGoogle Scholar and Landes, David S., The Unbound Prometheus: Technological Change and Industrial Development in Western Europe From 1750 to the Present (Cambridge, 1969). In the Kuznets tradition, Schmookler pioneered the systematic analysis of patent statistics for the purpose of studying inventive activity. Although recognizing the variety of experiences across industries, time, and specific inventions, he argued that the importance of demand, or the extent of the market, in altering the direction and level of inventive activity had not been well appreciated. He modeled the decisions of firms regarding the level of investment in inventive activity as a function of the expected gains from, and costs of, making a discovery. He argued that the extent or size of the market facing the firm was a major determinant of the expected gains, and investigated the relationship between patenting and proxies for the size of the market in many contexts. His work with time-series data on capital-good patents for a number of industries showed that the late nineteenth- and early twentieth-century trends in railroad patents of various types, such as those concerning rails, passenger cars, and freight cars, oscillated synchronously with the levels of investment in those types of capital equipment. Overall, Schmookler concluded that the expected profits from invention in particular industries were the prime determinants of the levels of invention in those areas, and that movements in the demand for the relevant products exerted substantial effects on the expected profits. Landes employed similar ideas in his treatment of the onset of industrialization in Britain. Although most of his references were concerned with inventions in individual industries, he, like Schmookler, seemed to intend his framework to apply to both the direction and the level of general inventive activity.Google Scholar

⁴ Joel Mokyr has most directly challenged Landes's perspective on invention. See Mokyr, Joel, “Demand vs. Supply in the Industrial Revolution,” this Journal, 37 (12 1977), pp. 981–1008.Google Scholar Although Nathan Rosenberg put more emphasis than Landes on supply-side factors, his studies of early nineteenth-century technology provide a balanced view when considered as a whole. Indeed, his attention to the extensive borrowing and modification of an already existing stock of British technology implies that the Americans did benefit from a substantial amount of less than fully exploited knowledge during the initial phase of their industrialization. In such a case, or when bottlenecks emerged, shifts in demand could have induced significant responses in inventive activity or patenting. See Rosenberg, Nathan, “Technological Change in the Machine Tool Industry, 1840–1910,” this Journal, 23 (12 1963), pp.414–43;Google ScholarRosenberg, Nathan, “Factors Affecting the Diffusion of Technology,” Explorations in Economic History, 10 (Fall 1972), pp. 3–34;CrossRefGoogle ScholarRosenberg, Nathan, Technology and American Economic Growth (New York, 1972);Google Scholar and other essays in Rosenberg, Nathan, Perspectives on Technology (Cambridge, 1976).CrossRefGoogle Scholar Other works that deal with related issues include David, Paul A., Technical Choice, Innovation and Economic Growth: Essays on British and American Experience in the Nineteenth Century (Cambridge, 1975);Google Scholar and Habakkuk, H. J., American and British Technology in the Nineteenth Century: The Search for Labour-Saving Inventions (Cambridge, 1962).Google Scholar

⁵ See Sokoloff, Kenneth L., “Productivity Growth in Manufacturing During Early Industrialization: Evidence from the American Northeast, 1820–1860,” in Engerman, Stanley L. and Gallman, Robert E., eds., Long-Term Factors in American Economic Growth (Chicago, 1986);Google ScholarSokoloff, Kenneth L., “Was the Transition From the Artisanal Shop to the Non-Mechanized Factory Associated With Gains In Efficiency?: Evidence From the U.S. Manufacturing Censuses,” Explorations in Economic History, 21 (10 1984), pp. 351–82;CrossRefGoogle Scholar and Rothenberg, Winifred B., “The Emergence of Farm Labor Markets and the Transformation of the Rural Economy: Massachusetts, 1750–1850,” this Journal, 48 (09 1988), pp. 537–66.Google Scholar

⁶ In contrast, those who argue for the supremacy of supply contend that the acceleration of invention stemmed from favorable developments on the cost side. In this view, technical breakthroughs in one area helped in the resolution of other technological problems, which in turn led to even further progress. If this were the major force behind the rise in invention at the outset of growth, one might expect the spurts of activity, and the associated gains in productivity, to be clustered in specific sets of industries. Inventions, whatever their source, alter the stock of knowledge, and that stock of knowledge surely affects both the conditions of supply and demand for goods, as well as the impact on further invention of shifts in demand. These complications, and the recognition that there is no general answer, make it very difficult to empirically distinguish between so-called demand and supply factors. Moreover, the circumscribed nature of the surviving evidence precludes precise estimates of the quantitative significance of the two sets of variables in more than a few idiosyncratic cases. It may be possible to obtain some notion of the orders of magnitude of their contributions during early industrialization, but such fine discrimination is not required to determine whether or not inventive activity was highly responsive to the growth of markets.Google Scholar

⁷ Igniting creative energy is not exactly the appropriate metaphor. Humans have exhibited enormous creative energy and inventiveness throughout their history, but much of it has always been directed toward pursuits that are not fully captured in our conventional measures of economic activity, such as the production of art, the wooing of companions, schemes for redistribution, and the acquisition of status. The total of all forms of inventiveness may indeed have increased substantially with industrialization, but a metaphor that also encompasses the reallocation of creative energy is more fitting. For further discussion of related issues, see Hirschman, Albert, The Passions and the Interests: Political Arguments for Capitalism Before Its Triumph (Princeton, 1977).Google Scholar

⁸ A prominent exception to this generalization is Higgs, Robert, “American Inventiveness, 1870–1920,” Journal of Political Economy, 79 (05/06 1971), pp. 661–67.CrossRefGoogle Scholar

⁹ Since Schmookler's work, many scholars have used patent records to study inventive activity. See, for example, the papers in Griliches, Zvi, ed., R & D, Patents, and Productivity (Chicago, 1984);CrossRefGoogle Scholar and Nelson, , ed., The Rate and Direction of Inventive Activity.Google ScholarThe problems with such patent-based measures have been extensively discussed, but economists have generally come to accept them as useful indicators, especially of the inputs devoted to the inventive process. For example, see Pakes, “On Patents.”Google Scholar

¹⁰ The sample was drawn from Ellsworth, Henry L., A Digest of Patents Issued By the United States From 1790 to January 1, 1839 (Washington, D.C., 1840);Google Scholar and Burke, Edmund, List of Patents For Inventions and Designs Issued By the United States From 1790 to 1847 (Washington, D.C., 1847).Google Scholar A machine-readable record of patents from this period is also being prepared by MacMurray, Robert R.. See his “Technological Change in a Society in Transition: Work In Progress On a Unified Reference Work in Early American Patent History,” this Journal, 45 (06 1985), pp. 299–303.Google Scholar

¹¹ The real and relative value of this fee fluctuated substantially with marked swings in the price level. Nevertheless, the $35 in current dollars represented at least 30 percent of per capita income throughout the period but seldom rose over 60 percent.Google Scholar

¹² Discussions with Steven Lubar, Robert Post, and Deborah Warner of the Smithsonian Institution helped the author to make sense of the complicated history of the enforcement and administration of the patent system during the period, as did Lubar's recent paper, “Invention Becomes Enterprise: The Transformation of Patent Law in Antebellum America,” unpublished manuscript, 1988. See also Post, Robert C., “‘Liberalizers’ versus ‘Scientific Men’ in the Antebellum Patent Office,” Technology and Culture, 17 (01 1976), pp. 24–54;CrossRefGoogle ScholarBugbee, Bruce W., Genesis of American Patent and Copyright Law (Washington, D.C., 1967);Google Scholar and Jones, Stacy V., The Patent Office (New York, 1971).Google Scholar

¹³ The classification of business or cyclical activity by year is drawn from Thorp, Willard L., Business Annals (New York, 1926).Google Scholar Similar information and judgments are provided by Smith, Walter B. and Cole, Arthur H., Fluctuations in American Business, 1790–1860 (Cambridge, MA, 1935).Google ScholarTime-series regressions, estimated with dummy variables and interacted growth terms for years of recession and years of prosperity, yield statistically significant evidence of cyclicality over annual data. However, the analysis of cyclicality in this article is based not on comparisons of individual years, but of years grouped together to identify sub-periods of “generally” consistent cyclical activity (such as “prosperity” or “recession”). The principal deviations from this procedure include: the definition of the 1812–1822 sub-period, where the contraction of 1815–1822 was extended back three years because of the war being fought on U.S. territory; the classification of the 1805–1811 Embargo sub-period; and the division of 1836 at June 4th because of the change in the patent system. One is struck by the similarities in the cyclical patterns of patenting observed here and those noted by Schmookler. See his Invention and Economic Growth, chap. 6. In particular, he noted the high correlation between long-term trends in patenting and the expansion of the market, the salience of long swings in the growth of patenting, and the tendency of movements in patenting to slightly lag turning points in business cycles. His analysis relied on moving averages, as he too held reservations about the significance of short-term or year-to-year changes in patents.Google Scholar

¹⁴ The sharp increase in the number of patents granted between 1805 and 1811 appears at first to contradict the claim that patenting was positively related to cyclical activity. In this sub-period, foreign trade was drastically reduced by a series of circumstances including the Embargo of 1807, and three of the seven years are classified as ones of recession or depression. Although the spurt in patents defies conventional cyclicality, it actually helps establish the linkage between patenting and the state of the market. The suspension of foreign trade boosted demand for domestically produced manufactured goods and other items competing with imports. Hence, the stimulus to invention provided by this exogenous political development might be understood as arising from the vast proportional increase in demand for non-agricultural commodities and the associated advances in the specialization of producers and the extent of markets. These developments might reasonably have led to the growth of effective demand for producers of goods in areas like the Northeast, which enjoyed a domestic comparative advantage in non-agricultural commodities. With the limited market for domestic manufactures beforehand, the Embargo period might have begun with considerable unexploited knowledge and potential for learning-by-doing. From this perspective, it is perhaps not surprising that the sudden shifts in resources and concern with production for the market precipitated a spurt in patenting. See, for example, the discussions of the impact of the Embargo in Cochran, Thomas C. and Miller, William, The Age of Enterprise: A Social History of Industrial America (New York, 1961);Google ScholarSpivak, Burton, Jefferson's English Crisis, 1803–1809: Commerce, Embargo, and the Republican Revolution (Charlottesville, 1979);Google Scholar and Goldin, Claudia D. and Lewis, Frank D., “The Role of Exports in American Economic Growth During the Napoleonic Wars, 1793 to 1807,” Explorations in Economic History, 17 (01 1980), pp. 6–25.CrossRefGoogle ScholarFor evidence of increases in agricultural productivity and the market orientation of farmers in Southern New England during these years, see Rothenberg, “The Emergence of Farm Labor Markets”;Google ScholarRothenberg, Winifred B., “The Market and Massachusetts Farmers, 1750–1855,” this Journal, 41 (06 1981), pp. 283–314;Google Scholar and Rothenberg, Winifred B., “The Emergence of a Capital Market in Rural Massachusetts, 1730–1838,” this Journal, 45 (12 1985), pp. 781–808.Google Scholar See Lindstrom, Diane, Economic Development in the Philadelphia Region, 1810–1850 (New York, 1978), for a detailed discussion of developments in eastern Pennsylvania. All of these fragments of the story lend support to the idea that an expansion of intra-regional specialization and trade within the Northeast, where all of the rise in per capita patents occurred, could have raised per capita income there as well as increased the size of markets facing potential inventors.Google Scholar

¹⁵ See the presentation of the model, and related discussion in Schmookler, Invention and Economic Growth, pp. 116–25.Google Scholar

¹⁶ For discussion of the greater severity in Southern New England and New York of the contraction of 1815–1822, see Cochran and Miller, The Age of Enterprise.Google Scholar For discussions of the many transportation investment projects undertaken during that contraction see the treatments in Meyer, Balthasar Henry, MacGill, Caroline E. et al. , History of Transportation in the United States Before 1860 (Washington, D.C., 1917);Google ScholarGoodrich, Carter, Rubin, Julius, Cranmer, Jerome, and Segal, Harvey, Canals and American Economic Development (New York, 1961);Google Scholar and Taylor, George R., The Transportation Revolution, 1815–1860 (New York, 1962).Google Scholar Another possible explanation of the appearance of dampened cycles in regions with lower rates of patenting per capita, such as the Southern Middle Atlantic and the Other U.S. region, is based on there being two components of patenting. One reflects a modest background level of inventive activity which was largely exogenous to the existence of a market, and the other represents a response to integration into a wider market. This structure implies that the higher the level of patenting within a geographic unit, the greater the proportion of inventive activity that was sensitive to national business conditions, and the more cyclically volatile the record should appear.

¹⁷ This feature of the record is not transparent because of the noise resulting from the small numbers of sample patents for the first two sub-periods, 1791–1798 and 1799–1804. After allowing for this problem, the stability of the national patent shares is striking. For example, in 1805–1811 and in 1843–1846, the proportion of patents classified as agricultural is the same, 19.2 percent; the proportion of manufacturing patents rises only slightly, from 44.7 to 47.1 percent. The surface stability in the national distribution of patents after 1805–1811 appears to be largely the result of averaging across regions that were in different phases of a common transition and were growing at different rates. Although the signal in the data is weak, regions seem to have experienced an increase in the proportion of their patents that were agricultural when they underwent their first major spurt in inventive activity; as they developed further, the agricultural share declined. For example, when Southern New England realized its acceleration in patenting during 1805–1811, the agricultural and manufacturing shares amounted to 16.7 and 42.6 percent respectively; by the middle of the 1830s, they had shifted to 7.9 and 56.3 percent. This process is visible in New York as well, but is screened in the national record by the relative growth of the Other U.S. region and the later development of the rest of the Northeast. The stability is surprising, but is at least partially accounted for by patents having been classified on the basis of final use (where the latter was apparent). Accordingly, manufactured goods such as seed drills and ploughs are included in the agricultural category. More generally, agricultural invention might have flourished during the earliest stages of development in a region, because many farmers would have benefited from expansion in their markets and increasingly specialized in producing for that market, and because certain agricultural products were important inputs in manufacturing processes.Google Scholar

¹⁸ Again, the estimates of the distributions of the patents filed during 1791–1798 and 1799–1804 should be viewed with caution, because of the relatively few observations from those years.Google Scholar

¹⁹ Although some of the patterns observed may be due to variation over time or place in the inclination to patent, it seems unlikely that such hypotheses could provide a complete explanation. The most plausible of such arguments is that the inclination to register an idea with a patent might have been higher in big cities. However, the application procedure was relatively simple, and commonly described in broad-circulation periodicals of the time, until 1836. Moreover, it is not obvious as to how one might plausibly account for suitable cycles in inclination or culture. A related possibility is that the cycles in patenting were driven by movements in the real value of the patent fee, but the series on prices and patents do not conform well with it. For example, the number of patents rose sharply during the 1820s, while the price level plunged.Google Scholar

²⁰ The interruptions in trade also had the effect of diverting the resources of urban merchants from foreign to domestic commerce. For further discussion of the relationship between patenting and urban centers, see Pred, Allan R., Urban Growth and the Circulation of Information: The United States System of Cities, 1790–1840 (Cambridge, MA, 1973);CrossRefGoogle Scholar and Higgs, “American Inventiveness.”Google Scholar

²¹ Another example is Pennsylvania, where the rural and urban rates rose from 11.3 and 8.7 respectively to 38.1 and 31.4 between 1812–1822 and 1830–1836. Even with so many diverse counties grouped together, the early advances in patenting are dominated by jumps, rather than growth along smooth continuous paths. This pattern is even more striking when one examines the records of individual counties separately. All counties in Northern New England are lumped together in the text here, because there were no metropolitan centers during this period, and generally no statistically significant differences between urban and rural counties in patenting.Google Scholar

²² The figures generally vary between 50 and 85 percent, depending on the time period and the regions compared. The results are strongest for Southern New England, where the metropolitan counties contain only a tiny fraction of the regional population.Google Scholar

²³ The geographic extension of markets increases both competition and the size of markets for local producers; each effect could stimulate more invention. For discussions of the impact of the Erie Canal on household manufactures and production patterns in New York, see Tryon, Rolla Milton, Household Manufactures in the United States, 1640–1860 (New York, 1917);Google Scholar and Cole, Arthur H., The American Wool Manufacture (Cambridge, MA, 1926), vol. 1.Google Scholar For treatment of the changes in the agricultural sector of Northern New England, see Bidwell, Percy W. and Falconer, John I., History of Agriculture in the Northern United States, 1620–1860 (Washington, D.C., 1941);Google Scholar and Bidwell, Percy W., “Rural Economy in New England at the Beginning of the Nineteenth Century,” Transactions of the Connecticut Academy of Arts and Sciences, 20 (04 1916), pp. 241–399.Google Scholar

²⁴ As indicated in the regressions reported below, there is no significant relationship across counties between bordering on the ocean and levels of patents per capita. The same types of regressions, when estimated with county-level literacy rates as an additional independent variable, indicate a strong positive relationship between literacy in 1840, the first year for which such rates are available, and patenting during earlier sub-periods. The line of causation is not clear, however, since literacy in 1840 is only weakly correlated with patenting in 1836–1842 and 1843–1846. Further investigation is underway. The low levels of patenting in counties near Pittsburgh, and on major rivers in western Pennsylvania are somewhat puzzling. The latter pattern might stem from the indirectness of the water route to eastern cities.Google Scholar

²⁵ Although the most straightforward explanation of this finding would probably focus on supply-side factors such as culture, ethnicity, occupational mix, and access to information or capital, some of these disparities in patenting across geographic units could plausibly be due to differences in the density, extent, or structure of markets not already captured by other independent variables.Google Scholar

²⁶ The sizes of markets varied with population, per capita income, the dispersion of technology, the degree of factor specialization, as well as with business cycles and improvements in the transportation network. All of these factors may have spurred the growth of markets and inventive activity. Moreover, the apparent association between market size and patenting, the likelihood that the restriction of manufactures from abroad augmented the market for domestic producers, and the evidence that patenting soared during the relevant years suggest a significant role for the 1807 Embargo and other obstacles to foreign trade in igniting a sustained acceleration in invention and in productivity growth in the American Northeast. This view is consistent with recent findings concerning the record of productivity growth during the early nineteenth century. See Rothenberg, “The Emergence of Farm Labor Markets,” for evidence on labor productivity growth in agriculture during the early nineteenth century. See Sokoloff, “Productivity Growth in Manufacturing,” for estimates of rapid progress in manufacturing between 1820 and 1860. The findings on patenting and productivity buttress each other in suggesting that many improvements in technique, accounting for substantial gains in productivity, were realized long before the broad wave of mechanization that occurred in the 1830s and 1840s, and in relating these changes to the expansion of markets. Unfortunately, the evidence on manufacturing does not cover the Embargo sub-period, but the correspondence between increases in productivity and patenting between 1820 and the 1840s, is consistent with the view that invention did augment productivity, and that variation in patenting does reflect variation in such activity.Google Scholar

²⁷ An excellent example of previously outlying zones whose households altered their behavior in response to their becoming exposed to the forces of wider markets, is Northern New England during the 1820s and 1830s. As improvements to rivers, such as the Connecticut, Kennebec, and Penobscot, were made over these years, patents per capita quadrupled. The populations of Maine, New Hampshire, and Vermont grew slowly over the interval, and it seems unlikely that a jump of this magnitude could be explained fully by the inflow of invention-prone migrants.Google Scholar

²⁸ Fishlow, Albert, American Railroads and the Transformation of the Ante-Bellum Economy (Cambridge, MA, 1965), chap. 4.Google Scholar

²⁹ See Goodrich, Carter, Government Promotion of American Canals and Railroads, 1800–1900 (New York, 1960), chaps. 1 and 2;Google Scholar and Meyer et al., History of Transportation, chap. 6.Google Scholar

³⁰ For example, notice that the levels of patents per capita and the relative size of the manufacturing sector (as measured by the ratio of the manufacturing labor force to the agricultural labor force) are modest in the counties along the western half of the Erie Canal prior to its completion in 1825. Yet, major advances are apparent by the early 1830s in patenting activity. It is also interesting that substantial increases in patents per capita were realized without large changes in the sectoral allocation of labor force. Some counties have no estimates for the earlier sub-periods, because they had not yet been carved out of the land areas of the others.Google Scholar

³¹ In The American Wool Manufacture Cole came to the same conclusion about the direction of causation in the association between access to wide markets and the decline of household manufacture of woolens. As the network of navigable inland waterways was extended, more and more households came to alter their behavior in order to take advantage of the changed market environment: … the home manufacture was ‘becoming daily more exceptional.’ And this could hardly be otherwise with the growth in the factory production, the cheapening and improvement of its output, and with the advent of better distributive and more adequate transportation facilities. The operation of these forces was so slow and inconspicuous that a picture of the movement cannot often be caught. However, the figures of household production in New York State, do give us an intimation of what must have been the experience of the country as a whole…. It is evident that in 1820 the counties nearer the City of New York and those on the line of the Hudson River had already been affected by the domestic factory production or had been influenced by the importation of woolen goods; or contrariwise, the counties with the highest household manufacture were those located at some distance from the distributive centers…. By 1845, however, a curious split had occurred. The counties of heaviest per capita production were now in the northern, less well-developed section of the state, or along the southern border. To one at all familiar with the history of transportation the reason is plain. In the interval between 1820 and 1845 the Erie Canal had been completed, and some portions of railroads reaching out westward from Albany, both traversing counties which had now ceased to have heavy per capita production. The introduction of transportation facilities occasioned a decay in the household industry (pp. 280–82).Google Scholar

³² It is interesting to note that Boston, New York City, and Albany, which were the metropolitan centers in the regions with the most extensive networks of navigable inland waterways, generally exhibited much higher levels of patenting per capita than did their counterparts in other regions. One might ask whether invention in such urban centers was somehow related to the volume of intra-regional trade.Google Scholar