Mid Century Productivity Growth in Relation to Current Trends by Alexander J. Field Department of Economics Santa Clara University Santa Clara, CA 95053 email: [email protected]
Abstract This paper explores the following working hypothesis: that the decade of the 1930s was the most technologically progressive of any in U. S. economic history. The hypothesis is meant seriously and in two ways: first that during this period a wide range of new products, technologies and practices were introduced into use, resulting in a high rate of measured multifactor productivity growth, and secondly, that the decade produced advances that replenished and expanded the larder of unexploited exploited techniques that provided the foundation for MFP improvement in subsequent periods. Progress in invention and innovation in the 1930s was important in facilitating US economic performance both before and during World War II, and in establishing foundations for the prosperity of the 1950s and 1960s.
Mid Century Productivity Growth in Relation to Current Trends
The twentieth century is now complete, but the Great Depression retains undiminished its title as the cataclysmic macroeconomic event of modern history. Real output, which had peaked at $103.1 billion in 1929, fell to $73.7 billion by 1933. Real fixed investment collapsed, from $14.9 billion in 1929 to $3.9 billion in 1933, a rate below that required to maintain the capital stock (i.e., net capital accumulation was negative) (all data are in 1929 prices). The construction boom that characterized the 1920s began with residential construction (during each of the four years 1924-1927 single family housing alone accounted for over 8 percent of GDP, a rate never equaled since) and was followed by an apartment and then a central business district boom. After its meltdown in the early 1930s, private construction did not recover until the 1950s. Residential construction revived with the mass produced Levittowns and Eichler developments of the postwar epoch. CBD building took even longer: the first post 1920s downtown construction in Boston, for example, was not initiated until 1959, and similar discontinuities can be seen in the vintages of downtown buildings in Los Angeles, Chicago, and almost any other major American city. The collapse in investment was mirrored by a dramatic downward move in big ticket consumption spending after 1929 – more than can be accounted for by the subsequent drop in income itself (Temin, 1989). More than four million passenger vehicles were produced in 1929. Because of the depression and the conversion of automobile production to military production in the first half of the 1940s, that rate, although approached in 1937, was not equaled again until 1950. Double digit unemployment,
which peaked at 25 percent in 1933, persisted for more than a decade, with rates as high as 19 percent experienced as late as 1938. Even after adjusting, as does Darby (1976), for those employed by the government in public works and other work relief projects, involuntary unemployment remained exceedingly high throughout the 1930s. Because of the Depression’s place in both the popular and academic imagination, and the repeated and justifiable emphasis on output that wasn’t produced, income that wasn’t earned, and expenditure that didn’t take place, it will seem startling to propose the following working hypothesis: the Depression years were, in the aggregate, the most technologically progressive of any comparable period in U. S. economic history. The hypothesis is intended seriously, and can be read as entailing two claims: first that during this period businesses and government contractors implemented or adopted on a more widespread basis a wide range of new technologies and practices, resulting in the highest rate of measured peacetime peak to peak multifactor productivity growth in the century, and secondly, that the decade produced advances that replenished and expanded the larder of unexploited or only partially exploited techniques, thus providing the basis for much of the labor and multifactor productivity improvement in the 1950s and 1960s. The hypothesis does not imply that all of the effects of the advances registered in the decade were immediately felt in the productivity data, nor, on the other hand, does it dismiss the significance of larder-stocking during the 1920s and earlier, upon which measured advance built. Rather, it draws our attention to the probability that progress in invention and innovation in the 1930s was significant, in ways not well appreciated, both in facilitating the remarkable U.S. economic performance before and during World War II, and in establishing foundations for the prosperity of the 1950s and 1960s.
Output Growth, Input Growth, and the Productivity Data The starting point for this exploration is macroeconomic data on real output growth, labor force growth, and the growth of the real capital stock, series that underlie our conclusions about trends in labor productivity and multifactor productivity growth. Labor productivity is most commonly measured as real output divided by labor input in hours; multifactor productivity as real output divided by the product of inputs of labor and capital services, each raised to a fractional power reflecting their share in national income (constant returns to scale and payment of marginal product to factors are typically assumed). The growth rate of labor productivity is thus the growth rate of output less that of hours, that of multifactor productivity the growth rate of output less a weighted average of the growth rates of labor and capital inputs, the weights reflecting the respective shares of these inputs in national income. Major contributors to the construction, adjustment, and interpretation of these data have included Edward Denison, John Kendrick, Dale Jorgenson, Zvi Griliches, Robert Solow, Moses Abramovitz, Paul David, and Robert J. Gordon. Of these, only the last three have attempted systematic historically informed overviews of the twentieth century as a whole. The data and productivity estimates to which their efforts give rise continue to be revised, for example by adjusting labor input to take account of changes in the age, gender, and educational characteristics of the labor force, and by adjusting capital input to take account of the rising importance of faster depreciating equipment, which has higher gross returns and contributes a higher gross service flow than do structures. Future adjustments may change some of the details of this story, but there is now an emerging consensus that, looking back over the last thirteen decades, the period between roughly
1913 and 1972 experienced exceptionally high rates of multifactor productivity growth, substantially higher than those evidenced in the decades preceding and following, when a much higher fraction of labor productivity growth is to be attributed simply to capital deepening (Abramovitz and David, 1999, 2000; Gordon 1999, 2000a, 2000b, 2000c). Within this half century there remains some lack of clarity about which sub-period evidences the very highest MFP growth. The analysis in Gordon 2000a and 2000b suggest that this distinction (measuring from peak to peak of peacetime business cycles) belongs roughly to a two decade period running from the end of the 1920s to the end of the 1940s, years characterized by capital shallowing, due to the low net capital formation during the Depression and World War 2. 1950-72 emerges as a close runner up. It was this then recent history, we may surmise, that so surprised Robert Solow in his 1957 analysis for which, in part, he received the Nobel prize. Solow’s work contributed to the development of the concept of the residual and its interpretation: since real output was growing much faster than could be explained by the growth of inputs conventionally measured, he (as did Abramovitz and others) suggested that the unexplained growth should be identified statistically with the contribution of a number of factors, the most important of which was technical change. Solow’s seminal article was, however, published more than forty years ago. Between 1972 and 1995 Solow’s residual and, indeed, the multifactor productivity growth to which it gives rise, all but vanished. It is true that labor productivity continued to grow, albeit at a markedly slower rate, but a very high percentage of this was attributable to capital deepening, which continued at a slower rate than was true in the 1950s and 1960s in part because of an upward trend in hours per worker. Still, to the
degree that labor productivity has advanced in recent years, it has, with the exception of a few years at the end of the 1990s, done so the old fashioned way – through sacrifice of current consumption so that physical capital goods (mostly structures and equipment) could be piled up at a faster rate than the growth of labor hours. Thus the economic history of the last three decades of the twentieth century has recapitulated in the United States a pattern evident in the late nineteenth century, but quite markedly absent during the second quarter of the twentieth century, and more generally over the five to six decade period between World War I and the end of the 1960s. Table 1 summarizes estimates of growth from several of Gordon’s recent analyses, along with estimates from Abramovitz and David (1999), in a manner that illuminates areas of both uncertainty and agreement. With respect to which sub-period registers the highest rate of MFP growth, Gordon 1999 gave the nod to 1950-64, but Gordon 2000a and 2000b reconfirm the supremacy of 1928-50. He writes, for example, that, in comparison to Japan and Europe, in the United States “a substantial part of the great leap in the level of multifactor productivity had already occurred by the end of World War II” (Gordon, 2000b, p. 22), contrasting this view with standard views of productivity trends, which tend to emphasize postwar developments. An emphasis on the postwar period is warranted if we are examining trends in labor productivity growth, because during the 1950-72 period, high rates of MFP advance were paired with a revival of capital deepening, an unbeatable combination in terms of rapidly raising output per hour and, as a result, real living standards. But it is MFP growth, roughly identified with the impact of technical change, that is at issue here.
The Abramovitz and David (2000) estimates also show peak MFP growth occurring before 1950: “Before allowing for the vintage effect, the rate of refined TFP growth from 1948 to 1966 stands higher than that from 1929 to 1948. Allowing for the vintage effect, the reverse seems to have been true” (2000, p. 29).1 Although the authors note the contrast between total productivity growth during these two periods, they do not pursue the significance of this finding for our understanding of the interwar period or the trajectory of technological progress in the twentieth century as a whole. In general, though Abramovitz and David agree with Gordon (2000a, 2000b) in placing peak MFP growth in approximately the same two decades preceding 1950, their narrative and numbers place somewhat less emphasis on a peak in the second quarter of the twentieth century and more on a broad plateau extending from the end of the nineteenth century through the mid 1960s. All but one of these recent analyses give pride of place to the second quarter of the twentieth century, with Gordon 1999 showing the most rapid advance in the two decades immediately subsequent. What seems indisputable from the macroeconomic data is that one or the other of these two periods experienced the highest measured compound annual MFP growth in U.S. economic history. The 1928/29-1948/50 period is critical from the standpoint of the trajectory of technical change, both because of its direct effect on growth during the period and because of its lagged effect on MFP advance in the 1950s which, when coupled with renewed capital deepening, produced a golden age of labor productivity growth and living standard improvement (this is not in dispute). In either event we must ask what happened between 1928 and 1950, or between 1929 and 1948, to use the Abramovitz1
The terms multifactor productivity and total factor productivity are used interchangeably here.
David chronology, to place the American economy in the position of world dominance it enjoyed at mid-century. Partly of course this involved wartime devastation in Europe and Japan, but partly, it must have reflected the extraordinarily high rates of MFP growth over the period, and the question on this account comes down to how much of this is the direct result of wartime experience, and how much is to be attributed to prior peacetime advances, particularly those achieved during the 1930s, both in measured productivity advance and in larder stocking.
Table 1 Compound Annual Average Rates of Private Non-Farm MFP Growth, United States Gordon, 1999 Gordon, 2000a Gordon, 2000b Abramovitz/David, 1999a
.21 .39 .39
.86 1.13 1.14 1.28
1.01 1.40 1.42 1.38
1.01 1.96 1.90 1.54
1.86 1.23 1.47 1.31
1.69 1.56 .89
1.04 .67 .16 .04
.34 .01 .59
.26 .27 .79
Abramovitz/David long-swing intervals: 1890-1905, 1905-27, 1929-48, 1948-66, 1966-89. See Table 1:IVA. Capital input estimates include adjustments for vintage effect. For Abramovitz/David, output includes housing services and the input series includes the housing capital stock, in contrast to Gordon.
The decade of the 1930s remains a central, dominating feature of twentieth century economic history, but one to which little systematic attention has been given from the standpoint of economic growth. If the conclusions towards which the macroeconomic data point stand up, as I believe they will, we will need a serious, historically informed analysis of the Depression and war years from the standpoint of long term growth. Hypotheses Two broad hypotheses may be suggested with respect to the record of productivity advance reflected in the 1928/29 to 1948/50 comparison. Either the growth in MFP levels is primarily attributable to an exceptional concatenation of technical advances 8
across a broad frontier of the American economy during the 1930s, building on unexploited opportunities at the end of the 1920s, or it is principally the consequence of the production experience of World War II: a persisting benefit of the enormous cumulated output over that four year period as well perhaps of spinoffs from war related R and D. For the latter hypothesis, the explanation of how we got where we were by 1948/50, to make reference to a classic article by Arrow, is principally that the economy was one large C-47 factory, permanently reaping the gains from wartime learning by doing (Arrow, 1962; Alchian, 1963). Certainly, the war experience left us with advances in such areas as radar, microwave technology, aeronautics, and atomic energy, as well as additional experience in producing large quantities of aircraft and vehicles, synthetic rubber, aluminum, and ordnance. Most of the growth accounting studies deal with nonfarm output, but we might add that the expansion of munitions plants led to a permanent decline in the real price of fertilizer in the postwar period, benefiting agriculture (Olmstead and Rhode, 2000, p. 710). Whether these advances were, in the aggregate, more significant in accounting for the level of output and productivity achieved by 1948/50 than those already attained by 1941, or whose foundations were in place by that point, is a question that has not heretofore been asked. The alternative hypothesis is that many gains, both in the achievement of higher measured productivity levels and in the expansion of the larder, had already been reached by the outbreak of war, and indeed helped make possible its successful prosecution. This then implies that throughout the 1930s, behind the dramatic backdrop of continued high unemployment, technological innovations were occurring across the American economy,
especially but not exclusively in chemical engineering (including petrochemicals and synthetic rubber), aeronautics, electrical machinery and equipment, electric (including hydroelectric) power generation, and civil/structural engineering, that these trends have something to do with the rising real wages during this period of those who managed to stay employed, and that the sum total of these changes had, by the onset of World War II, increased the natural or potential output of the U.S. economy far beyond what contemporary observers and economists at the time believed possible. Many of these developments involved entirely new products, not just process improvements in the production of goods already in the market.
Interpretations of Rising Labor Productivity in the 1930s It is one of the surprising features of the Depression years, given our general perception of the decade that, as Claudia Goldin notes in her recent survey of U.S. labor market development in the twentieth century, non-farm hourly labor productivity grew considerably more rapidly in the 1930s than it did in the 1920s (2000, pp. 566-67). Goldin attributes this trend to Depression era cuts in hours per worker and the probability that those unemployed were less educated and “probably less skilled” than those retained. Goldin’s interpretation, however, is emblematic of an ambiguity in the labor economics literature involving whether labor productivity is pro- or counter-cyclical. Supporters of the view that it is counter-cyclical, such as Goldin, adduce the hypothesis of selective retention, whereas supporters of the pro-cyclical view emphasize labor hoarding. My own reading of the empirical evidence suggests, on average, mild procyclicality, but the unresolved character of this debate underlines the importance in
calculating growth rates of labor productivity of making comparisons between peaks in business cycles. The selective retention mechanism Goldin identifies cannot be appealed to in explaining the growth of real output per hour between 1928/29 and 1948/1950. This is a peak to peak comparison, and the result holds up even when labor is quality adjusted. Nor can this rise in labor productivity be easily attributed to a rise in the capital to labor ratio, since the standard fixed asset data show this to be a period of capital shallowing. If the numbers are right, the rapid growth in labor productivity must be attributed, at least in an accounting sense, to high MFP growth over these years. The real business cycle (RBC) approach would seem to be a particularly inauspicious framework within which to approach a study of the Great Depression, given the apparent inability of anyone yet to identify the smoking gun on the supply side responsible for the 1929-33 downturn. Nevertheless, the approach, which aims to account for cycles using the same types of supply side explanations one uses for long term growth, does predict procyclical labor productivity movements in a way exclusively demand based explanations, or an emphasis on selective retention, do not. RBC arguments are a reminder that not all cyclical episodes have similar underpinnings, and that supply as well as demand shocks play a role in the shorter as well as medium and long term behavior of the macroeconomy. Ben Bernanke and Martin Parkinson note the sharp procyclical movements in manufacturing productivity during the 1929-39 period. But they reject ab initio the possibility that supply side shocks might have played a role in producing this result: “we believe that it is quite unlikely that the preponderance of interwar cyclical variation (at
least during the 1930s) was due to technological shocks to the production functions of individual manufacturing industries” (Bernanke and Parkinson, 1991, p. 441). In a recent paper within the RBC tradition addressing the weakness of recovery after 1933, Harold Cole and Lee Ohanian note the rapid labor and MFP growth during these years (Cole and Ohanian, 2001). Their focus, however, is on the consequences of rising real wages, whose causes, somewhat ironically, they identify as institutional. The National Industrial Recovery Act and the effective continuation of its policies after it was declared unconstitutional represent, in their view, a significant governmentally induced negative supply shock to the economy.2 Cole and Ohanian are right to emphasize that we should pay attention to developments on the aggregate supply side during this period. But they miss the likelihood that a considerable portion of the upward wage movement was in fact warranted by labor productivity improvement as a consequence of positive supply shocks.3 Perhaps there is more going on in this decade of economic depression than we are prepared at first glance to see. Goldin’s, Cole and Ohanian’s, and Bernanke and Parkinson’s characterizations of the data (as opposed to their explanations) are all consistent with a set of powerful supply shocks in the Depression years that laid the groundwork not only for the remarkable Allied victory in the war, but also the age of high mass consumption, as W. W. Rostow described it, which began almost immediately thereafter in the 1950s. Most of the corporations occupying the commanding heights of the economy, including RCA, AT&T, IBM, Dupont, Alcoa, GM, Kodak, and General 2
Its labor policies, including encouragement of the right to organize, they see as persisting through the instrumentality of the Wagner Act. Encouragement of firms to cartelize they see persisting through lax antitrust enforcement.
Electric – companies that because of a strong commitment to organized privately funded research and development activity (Mowery and Rosenberg, 1989, pp. 74-75) were contributing disproportionately to MFP advance, had returned to profitability well before the onset of World War II (Galambos, 2000, p. 947). These companies, because of their dominant market position, and because of the close integration of their R and D efforts with marketing and new product development, were less likely to suffer from the divergence between private and social return to research that has traditionally served as the theoretical foundation for governmental support of such activity. Within manufacturing, there is also evidence that the downturn differentially affected high and low productivity firms, with a shakeout taking place as the least efficient establishments were culled out (Bresnahan and Raff, 1991). This Darwinian mechanism may have reinforced the technological trends I identify in contributing to measured productivity growth between 1928/29 and 1948/1950. But Bresnahan and Raff limit the source of productivity improvement to a between firm (or plant) effect, much of which would have been reversed with the return to fuller employment before the war. I argue that the productivity outcome, measuring peak to peak, was due to more than temporarily losing the weakly performing tail of the firm or plant productivity distribution. Gavin Wright has emphasized the importance in considering the history of labor productivity of changes in labor regimes, stressing particularly the cutoff in immigration resulting from the beginning of unrestricted U-boat warfare in the First World War and consolidated with the Immigration and Naturalization Acts of 1921 and 1924. Increased
Elsewhere I argue that the main obstacles to recovery lay in the construction, not, as they suggest, the manufacturing sector (Field, 1992).
labor supply drawn from the American South compensated for this cutoff of supply, but only in part. Wright’s argument, therefore, suggests that the rapid growth of manufacturing productivity during the 1920s – over 5 percent per year as measured by Kendrick, was not simply due to the availability of new technologies but also reflected a rise in the capital-labor ratio in response to a reduction in the elasticity of labor supply. Combining this with Paul David’s emphasis on the delayed impact of electrification on both factory layout and the American home, one has a plausible narrative applicable to the 1920s (David and Wright, 1999). There is evidence that this regime persisted, in the sense that we emerged from the Second World War with a high wage, relatively egalitarian labor regime that continued through the early 1970s. But it is questionable how much purchase these mechanisms give us in understanding productivity trends in the 1930s and 1940s, particularly those affecting multifactor productivity. First, a rise in the capital-labor ratio in response to a new labor regime, although it will almost certainly increase labor productivity, will have no effect in and of itself on the calculated growth rate of multifactor productivity. Secondly, the 1930s and 1940s were characterized by the highly unusual circumstance of capital shallowing. Although labor hours grew slowly due to immigration restriction and continuing declines in fertility, capital services grew more slowly still, due to the collapse of gross and net investment in the 1930s and the limited investment in the private sector during the Second World War. Kendrick’s Data The argument advanced here appears to conflict with John Kendrick’s conclusion that although private domestic economy MFP grew at a rate of 2 percent per year
between 1919 and 1929, it did so at only 1.6 percent per year between 1929 and 1937 (Kendrick, 1961, p. 72). Given the conventional emphasis on the boom of the 1920s and its contrast with the disastrous macroeconomic performance in the 1930s, we might be inclined to accept these differentials and move on to more interesting matters. But Kendrick compares a fully employed economy in 1929 (3.2 percent unemployment) with a 1937 economy in which 14.3 percent of the labor force (9.2 percent according to Darby) was still out of work. Although large firms such as those mentioned above were doing well, thousands of medium and smaller ones were not. A more reasonable peak to peak comparison is to choose as an end point 1941, when unemployment, although still averaging 9.9 percent (6 percent according to Darby), was closer to what it was in 1929, but before war spending had seriously distorted the economy. The choice of 1941 warrants further discussion. Admittedly, a military buildup in anticipation of the Second World War had already begun by 1941, a year in which federal military spending totaled $6.3 billion. This was more than three times the military spending of $1.8 million in 1940, and represented about 5 percent of 1941 U.S. GNP. This increased government spending undoubtedly contributed to higher employment and output levels before the war, through standard multiplier mechanisms (Vernon, 1994). But cumulated military procurement was still relatively minor compared to what would follow. Total federal military spending reached $22.9 billion in 1942, $63.4 billion in 1943, $76.0 billion in 1944, and $80.5 billion in 1945 (U.S. Bureau of the Census, 1975). In other words, by the end of 1941, only 2.5 percent of the $249.1 billion total military spending occurring between 1941 and 1945 inclusive had already been undertaken.
On the supply side, whatever productivity improvements may have been achieved producing destroyers and munitions in 1941 would have had little effect on Kendrick’s calculated private sector MFP. Since government production is valued in the national income and product accounts at cost of production, productivity in government production is by definition equal to 1. There would have had to have been extremely rapid spillover effects from public to private production for the war build up to have affected private sector productivity by this date. It seems difficult, therefore, to credit achieved productivity levels in 1941 to the effect of new management techniques learned, or new technologies discovered, as the result of cumulated war production. Would 1942 be more appropriate for a peacetime peak to peak comparison? By 1942 the economy was moving towards a full scale war footing, so we would need to worry more about spillovers. On the other hand, it makes little sense to argue that 1940 is a better candidate, since unemployment (14.6 percent) was actually higher than it had been in 1937. In contrast to the work of other students of productivity change, Kendrick’s 1961 book includes detailed appendices providing annual measures, in levels, of inputs, outputs, and productivity indexes.4 It is thus particularly useful in addressing the issues of timing raised in this paper. Using his data (Commerce version), I calculate a compound annual average growth rate of private domestic economy MFP of 2.27 percent per year between 1929 and 1941. In contrast, MFP grows at 1.99 percent per year between 1941 and 1950 (Kendrick, 1961, Table A-22, pp. 334-5).
Annual BLS analysis of multifactor productivity is available beginning only in 1948; see www.bls.gov.
Table 2: Compound Annual Average Growth Rates of Private Domestic Economy MFP, United States, 1919-1950 1919-1929 1929-1941 1941-1950
MFP growth (Commerce) 2.00 2.27 1.99
MFP growth (Commerce) 1919-1928 1.74 1928-1941 3.61 1941-1948 1.50 Source: Kendrick, 1961, Table A-22, pp. 334-35.
Note: Kendrick data include agricultural sector.
Suppose we use instead the 1928 end of 1920s peak favored by Gordon and the 1948 end of 1940s peak favored by Abramovitz and David. Unemployment in 1928 was 4.2 percent, vs. 6.0 percent (Darby measure) in 1941. Unemployment in 1948 was 3.8 percent, vs. 5.2 percent in 1950, making 1948 arguably a better endpoint than 1950 for a peacetime peak to peak comparison. 5 The results using these beginning and endpoints are even more dramatic in terms of the rapidity of productivity growth they suggest for the Depression years, showing a compound annual average growth rate of private domestic economy MFP of 3.61 percent between 1928 and 1941, more than twice as rapid as that registered between 1941 and 1948. Sectoral and Microeconomic Approaches The macroeconomic evidence that the fastest rate of multifactor productivity growth over the last century and a half, and probably two centuries, took place in the 1928-1948 period, and within that period, in the years before the war, is corroborated by a variety of evidence at the microeconomic level. Alfred Kleinknecht’s 1969 study of product and process innovations from 1850 to 1969 (Figure 1) provides a data set on fundamental innovations, divided into product, process, instrumentation, and other. The peak for the total and two of the four components is in the 1930s, and is particularly marked for product innovations. Jacob Schmookler’s 1966 enumeration of basic and
Indeed , since the economy peaks in 1948:Q4, Gordon’s use of 1950 as a reference point is puzzling.
Product Process Instrument Other Total
20 15 10 5 18 50 -5 9 18 60 -6 9 18 70 -7 9 18 80 -8 9 18 90 -9 9 19 00 -0 9 19 10 -1 9 19 20 -2 9 19 30 -3 9 19 40 -4 9 19 50 -5 9 19 60 -6 9
Source: Kleinknecht (1969) improvement innovations shows a similar peak in the 1930s, particularly its second half, as does the chronology provided by Gerhard Mensch (1979, p. 132). These studies have often been ignored in macroeconomic inquiry, in part because their results seem so at variance with our impressions of the economic “success” of the depression decade. A second type of evidence arises from David Mowery’s analysis of National Research Council data on the establishment of R and D labs in manufacturing. Mowery tabulated lab establishment by subperiod, and found that within manufacturing the peak period was 1919-1928 (660 founded). The runner up period in his analysis, however, was 1929-36 (590) – a remarkable statistic, given the depression years involved. Since the latter period is of eight years duration, there were actually more labs founded per year in the worst years of the depression (73), than in the decade preceding (66) (see Mowery and Rosenberg, 1989, Table 4.1, pp. 62-63). This pattern is consistent with Mowery’s data on R and D expenditures. In the macroeconomically disastrous decade of the 1930s, industry R and D expenditures more than doubled in real terms, with acceleration in the last years of the decade (Mowery and
Rosenberg, 1989, p. 69). Mowery also reports that employment of research scientists and engineers grew 72.9 percent between 1929-33 while employment totals in other occupational categories collapsed. In the Second World War, in contrast, research and development employment growth slowed as employment in other categories skyrocketed. Federal spending for non-defense R and D also fell substantially during World War II (Mowery and Rosenberg, 2000, p. 819). Table 3 R and D Employment in Manufacturing, 1921-1946
1921 1927 1933 1940 1946
R and D Employees 2,775 6,274 10,918 27,777 45,941
13.6% 3,499 9.2% 4,644 13.3% 16,859 8.4% 18,164 Source: Mowery and Rosenberg, 2000, p. 814.
As Table 3 indicates, manufacturing R and D employment in 1940 was 4.4 times what it had been in 1927, and the compound annual growth of such employment was significantly slower during the war years (1940-46) than in the seven years previous. What happened after the war? Although federal spending added dollars and manpower to R and D activity, it devoted major portions of it toward a limited number of sectors, including (outside of medicine and health) aerospace, missiles, computers, and instrumentation. The increasing federalization and persisting influence of military priorities on R and D may be partly responsible for the narrower (in comparison with the 1930s) range of industries experiencing dramatic MFP growth at the close of the century (Gordon, 2000c, Oliner and Sichel, 2000).
Public and Private Capital Formation in the Second Quarter of the Century Much of the productivity improvement in the 1930s took place in manufacturing, but the transportation network and more broadly infrastructure was another important locus. The 1920s were notable for the rapid diffusion in the United States of the automobile, the full exploitation of the greater flexibility offered by electricity in the organization of manufacturing, and a rash of new electric powered appliances that began to revolutionize the American home and kitchen (Mowery and Rosenberg, 2000). But Americans initially had little idea how to build urban, suburban, and interurban infrastructures for an automobile age, and the increments to housing services output were disappointing, given the quantities of real investment undertaken (Field, 1992). The 1930s witnessed the development of improved techniques for utilizing concrete in conjunction with steel in bridge, tunnel, dam, and highway design. Perhaps of even greater importance, the decade saw the working out of a paradigm for building infrastructure suitable for an age of automobiles and trucks, with implications for the spatial configuration as well as design and construction of roads, highways, bridges, wholesale and retail distribution facilities, and residential subdivisions. Due to network effects, the design improvements in conjunction with infrastructural investment generated a boost in output in housing and wholesale and retail distribution beyond what can be swept back to the physical capital formation itself. In other words, infrastructural investment contributed to economy wide multifactor productivity, as well as labor productivity growth rates. Housing services today comprise over 10 percent of GDP, over 14 percent of private nonfarm production, and an even larger share of private consumption. The sector
is often neglected by economic historians, perhaps because of its unusual production function (inputs are effectively limited to capital services), and because so much of its contribution to final output is imputed. Gordon excludes the housing sector in his MFP calculations. Current BLS practice is to include the rental portion but not the owner occupied portion of the stock. Abramovitz and David (2000) include the output of the housing sector and its capital service inputs, as did Denison (1974). The productivity of housing investment was far higher in the 1950s than it had been in the earlier boom of the 1920s (Field, 1992). The massive investment in construction during the 1920s, particularly residential construction, was simply not very effective at translating bricks and mortar into machines for producing housing services, in part because designers had not yet figured out very well how to build subdivisions for the automobile. Much of the development work on these principles was done during the 1930s under the aegis of the newly formed Federal Housing Authority and diffused throughout the United States after the Second World War through the efforts of this agency in conjunction with local zoning authorities. A large portion of the infrastructure required for economically successful postwar housing construction was, however, put in place during the 1930s, as the consequence of the use of public funds to improve the road transport system. During the 1920s, infrastructure, particularly streets and highways, did not keep up with the burgeoning sales of private vehicles. Public expenditures in the 1930s substantially remedied this, in a manner that impacted the productivity of the housing sector as well as that of the economy as a whole.6 Denison’s data show a surge in the real value of the service flow
The argument is that these publicly provided capital services are complementary to many types of private production: housing especially, but distribution.
from the housing sector beginning in 1940 (1974, Figure 3.1, p. 20). This may be partially attributable to a reduction in the vacancy rate, as Denison suggests, but I believe it also has to do with the almost doubling of the streets and highways capital stock during the Depression (see below). After the war the increase in the real service flow from the housing stock continued apace, reflecting not only new investment, but, if I am correct, the full exploitation of new blueprints for organizing residential subdivisions and associated infrastructure tailored to the automobile. Consideration of the impact of investment in streets and highways is part of a larger issue in the twentieth century involving the decline of the relative importance of privately owned infrastructural capital in the railroad sector and the rise of publicly owned capital in the form of road, bridge, tunnel, and airport facilities (see Gordon, 1969, p. 237). The 1930s witnessed qualitative and quantitative changes in the nation’s infrastructure, as well as its sources of funding. If publicly owned capital contributed to private sector output, some of the measured private sector MFP growth in the 1928 to 1941 period should more properly be attributed to the effect on private output of public sector infrastructural investment. The component of the public capital stock with the strongest claim to having directly influenced private sector output in the interwar period is the stock of streets and highways. Ever since E. Cary Brown’s 1956 article it has been commonplace to downplay the significance of public investment during the Depression as too small in relation to GDP to have “made much of a difference” in returning us to natural output.7 What this point of view obscures is the likelihood that although insufficient in terms of
Thus the famous quote: “Fiscal policy … seems to have been an unsuccessful recovery device in the ‘thirties—not because it did not work, but because it was not tried (Brown, 1956, pp. 863-866).
its influence on aggregate demand to compensate for the drop in autonomous private spending, public investment nevertheless had significant impacts on the supply side.
Table 4: Compound Annual Average Growth Rates of Net Stock of Streets and Highways, United States, 1928-1973
1929-1941 1941-1950 1950-1973 1973-2000 Source:
Street/Highway Street/Highway Growth Growth 4.32% 1928-1941 4.59% 0.67% 1941-1948 0.08% 4.27% 1948-1973 4.15% 1.63% Bureau of Economic Analysis, Fixed Asset Table 7.2 (http://www.bea.doc.gov).
As Table 4 shows, the real net stock of streets and highway capital in the United States almost doubled between 1928 and 1950, with virtually all of this increase occurring before 1941. The beneficial effects on output of this investment, however, go beyond what is reflected in the increased physical capital input measured by the costs of concrete, steel, and asphalt. New principles of highway design, urban and suburban layout, and residential subdivision influenced both the construction of roads, and their layout. Some of the MFP growth in the 1930s therefore reflected what was essentially disembodied technical change in the transport system. While spending on publicly owned capital has figured heavily in debates about the significance of declining infrastructure investment in the fall off in productivity growth in the last quarter of the century (Aschauer, 1989; Gramlich 1994), virtually no attention has been paid to its role on the supply side in the earlier period.
These data are available at the URL indicated in the source notes to Tables 4 and 5. In 1969 Gordon noted the absence of much demand for, or supply of government investment and capital stock data. Both of these lacuna have, to a considerable degree, now been remedied, but the implications of these numbers for twentieth century economic history have hardly been explored.
Table 5 gives an indication of the magnitude of both level and growth of this component of the nation’s capital stock, as well as the governmentally owned stock of electric and gas, transit, and airport structures. Streets and highways capital surged in value from 6.4 percent of the net private fixed capital stock in 1929 to 10.7 percent in 1941, roughly the same percentage recorded in 1973. Government owned structures associated with the generation and transmission of electricity and gas (e.g., Hoover Dam, TVA, Columbia River), along with publicly owned transit capital and airport structures also grew quite substantially in relation to the private capital stock during these years. Table 5: Net Street and Highway Capital and EGTA Structures as Percents of the Net Fixed Private Capital Stock Private Street/ % of Electric/Gas, % of Fixed Highway Private Transit, Airport Private Capital Capital Capital Structures Capital 1928 $249,786 $16,319 6.54 $1,924 .77 1941 $289,487 $30,861 10.66 $5,149 1.77 1948 $582,248 $47,892 8.22 $7,494 1.28 1973 $2,698,194 $290,389 10.76 $46,149 1.71 2000 $21,464,786 $1,423,833 6.63 $362,860 1.69 Note: Figures are net stock in millions of current dollars. Source: Bureau of Economic Analysis, Fixed Asset Tables 2.1, 7.1 (http://www.bea.doc.gov). BEA data on “other government structures” reflect principally oil and gas structures, fixed structural capital in transit systems, and airport structures. See Table 7.1, line 15, fn. 4.
Beginning in 1947, BEA data permit us to compare the magnitude of street and highway capital with railroad fixed capital in the private sector. By that year the value of streets and highway capital had grown to over two thirds (68.7 percent) of the value of fixed capital in railroads (Bureau of Economic Analysis, Fixed Asset Tables, Table 3.1ES, Table 7.1). Public investment in streets and highways contributed to the growth of labor productivity in the 1930s, both through the effective deepening of capital available
to the private sector and, as a consequence of design and network externalities, through its impact on multifactor productivity growth.9 A second empirical question involves the significance of public investment in government owned and privately operated (GOPO) capital during World War II. Wartime government involvement in the private economy was extensive (Galambos, 2000, p. 948). Washington funded the construction of large plants for the atomic bomb project (government owned, government operated), and a number of other industries critical to the war effort, including synthetic rubber, airframes and engines, aviation fuel refining, and aluminum production. These plants were owned by the government, operated by private firms during the war, and sold off to the private sector in its aftermath. GOPO capital improvements funded both the construction of plant and the acquisition of equipment for corporations engaged in war production. Over $10 billion of improvements were provided in this fashion. These expenditures were not uniformly distributed across sectors, even within manufacturing, which received the lion’s share. For example, our synthetic rubber capability was dramatically expanded after the Japanese seized control of Malaysia, and sources of the natural raw material. $700 million was spent to build 51 plants in the United States, all sold to private industry by the 1950s (see Mowery and Rosenberg, 2000, p. 857; Morton, 1982, pp. 231, 235; these numbers differ slightly from those reported in Gordon, 1969, p. 229). In aluminum, the Reynolds Metal Company owed its existence to the war: it was created through sales to
It is difficult to argue in this historical context that the public investment was simply responsive to output growth, as some have maintained in the debate about the role of such investment in the last half of the twentieth century (Gramlich, 1994, p. 1188).
the private sector by the government of plants constructed during the conflict. The explicit intent was to establish a competitor to Alcoa, which had previously monopolized the industry (a bit of competition policy undone on May 3, 2000, when Alcoa acquired Reynolds). In aerospace a number of companies significantly expanded their capacity as the result of similar transfers. In spite of pioneering efforts by Gordon, we still know relatively little about the program and the plants, their role in the war effort, their subsequent disposition, and their impact on the postwar U.S. economy. The facilities were constructed via contracts from the Defense Plant Corporation, a subsidiary of the Reconstruction Finance Corporation, a depression era creation that took on a new role during the 1940s in war planning and production. Each project required a sponsor, such as the Army, the Navy or the Office of Rubber Reserve (another Reconstruction Finance Corporation subsidiary responsible for building up U.S. synthetic rubber capability). Firms paid a nominal rent of $1 per year with an option to buy at the end of the war. Many of these plants were sold rapidly to private enterprises after the war, although the synthetic rubber facilities remained in the government’s hands until 1955. The valuation of privatized GOPO plants in the capital stock figures could affect whether 1928-50 or 1950-72 registers the highest rate of MFP growth. In his 1969 AER article, Robert Gordon valued such investment at $16 billion dollars, helping to explain how real manufacturing output could rise 95 percent between 1940 and 1944, with a private investment in the sector of only $11.4 billion on new plant and equipment – far above Depression levels but still only a 16 percent net addition to the manufacturing capital stock of 1940 (Gordon, 1969, p. 225).
A valuation dispute centers around whether the prices at which the capital was sold off represented sweetheart deals to private industry or whether the prices were fair. Defenders of the transactions argue that substantial retrofitting was necessary to convert facilities designed for wartime production to peacetime use, whereas critics claim that the sale prices vastly undervalued the true extent of the transfer. To the degree that such capital is carried on the books (and in estimates of aggregate capital stocks) at artificially low prices, it will tend to favor the 1928-48 period in MFP calculations. There is, however, another implication of this possible adjustment. To the degree that the 1948 private domestic capital stock is undervalued, we will have overestimated the true level of MFP in that year because the capital stock (and its attendant services) figure in the denominator of an MFP calculation. Thus we will have calculated an excessively high rate of MFP growth for 1941-48, a rate that is already less than half that registered for the 1928-41 period. Using similar logic, adjusting upward the value of the 1948 capital stock because of an undervaluation of transferred GOPO capital would make the 1948-73 period look somewhat better from the standpoint of MFP growth. But it would also make even more dramatic the contrast between the rate of MFP growth between 1929 and 1941, as compared with that between 1941 and 1948 (see Table 2). Whatever kind of adjustment we might make for undervaluation of transferred GOPO capital in the private capital stock, the challenge of understanding the remarkable technological progressivity of the Depression years will remain.
Conclusion In spite of his many contributions to our understanding of productivity trends, Kendrick led us astray by choosing 1937 as a reference peak in calculating productivity growth in the 1930s. If we use instead 1941, his data reveal that the bulk of the very rapid growth of MFP between 1928 and 1948 took place before U.S. entry into the war. Once this is recognized, the conclusion drawn from the aggregate data finds strong support in microeconomic studies of the timing of key innovations, in sectoral studies of productivity change, and in data on private sector employment and expenditures on research and development. This finding will require rethinking of our understanding of the broad contours and determinants of U. S. economic growth in the twentieth century.
Abramovitz, Moses and Paul David. 1999. “American Macroeconomic Growth in the Era of Knowledge-Based Progress: The Long Run Perspective.” SIEPR Discussion Paper No. 99-3, Stanford, California (unpublished). Abramovitz, Moses and Paul David. 2000. “American Macroeconomic Growth in the Era of Knowledge-Based Progress: The Long Run Perspective.” In Stanley Engerman and Robert Gallman, eds., The Cambridge Economic History of the United States, volume III. Cambridge: Cambridge University Press, pp. 1-92. Accordino, John J. 2000. Captives of the Cold War Economy Westport: Praeger. Armen Alchian. 1963. “Reliability of Progress Curves in Airframe Production.” Econometrica 31 (October): 679-693. Aschauer, David A. 1989. “Is Public Expenditure Productive?” Journal of Monetary Economics (March): 177-200. Arrow, Kenneth. 1962 “The Economic Implications of Learning by Doing.” The Review of Economic Studies 29 (June): 155-173. Bernanke, Ben and Martin Parkinson. 1991. “Procyclical Labor Productivity and Competing Theories of the Business Cycle: Some Evidence from Interwar U.S. Manufacturing Industries.” Journal of Political Economy 99: 439-459. Bernstein, Michael. 1987. The Great Depression: Delayed Recovery and Economic Change in America, 1929-1939. New York: Cambridge University Press. Brown, E. Cary. 1956 “Fiscal Policy in the ‘Thirties: A Reappraisal.” American Economic Review 46 (December): 857-879.
Cole, Harold L. and Lee E. Ohanian. 2001 “New Deal Policies and the Persistence of the Great Depression: A General Equilibrium Analysis.” Federal Reserve Bank of Minneapolis, Working Paper 597 (May). Chapman, Gary and Joel Yudken, 1992. Briefing Book on the Military Industrial Complex. Washington D.C.: Council for a Livable World. Darby, Michael. 1976. “3 Million Employed Have Been Mislaid.” Journal of Political Economy 84: 1-16. David, Paul and Gavin Wright. 1999. “General Purpose Technologies and Surges in Productivity: Historical Reflections on the Future of the ICT Revolution.” Working Paper, All Souls College, Oxford. David, Paul, and Gavin Wright. 1999. “Early Twentieth Century Productivity Growth Dynamics: An Inquiry into the Economic History of ‘Our Ignorance’.” Working Paper. All Soul’s College, Oxford University. Denison, Edward F. 1974. Accounting for United States Economic Growth, 1929-1969. Washington: The Brookings Institution. Dumas, Lloyd. 1995. “Finding the Future: The Role of Economic Conversion in Shaping the Twenty First Century” in Lloyd Dumas, ed., The Socio Economics of Conversion from War to Peace Armonk: Sharpe, pp. 3-20. Dumenil, Gerard and Dominique Levy. 1994. “Stylized Facts about Economic Progress since the Civil War: A Vintage Model.” Structural Change and Economic Dynamics 5: 1-23. Edelstein, Michael. 2000. “War and the American Economy in the Twentieth Century.” in Stanley Engerman and Robert Gallman, eds., The Cambridge Economic History
of the United States, volume III, Cambridge: Cambridge University Press pp. 329-406. Field, Alexander J. 1992. “Uncontrolled Land Development and the Duration of the Depression in the United States.” Journal of Economic History 52 (December): 785-805. Field, Alexander J. 2001. “Not What it Used to Be: The Cambridge Economic History of the United States, Volumes II and III.” Journal of Economic History 61 (September): 806-818.
Field, Alexander J. 2003. “Communications.” Historical Statistics of the United States – Millennium edition. Cambridge: Cambridge University Press (forthcoming). Galambos, Louis. 2000. The U.S. Corporate Economy in the Twentieth Century.” in Stanley Engerman and Robert Gallman, eds., The Cambridge Economic History of the United States, volume III. Cambridge: Cambridge University Press, pp. 927968. Gelertner, David. 1939: The Lost World of the Fair. New York: The Free Press. Goldin, Claudia. 2000. “Labor Markets in the Twentieth Century.” in Stanley Engerman and Robert Gallman, eds., The Cambridge Economic History of the United States, volume III. Cambridge: Cambridge University Press, pp. 549-623. Gordon, Robert J. 1969. “$45 Billion of U.S. Private Investment Has Been Mislaid.” American Economic Review 59 (June): 221-238. Gordon, Robert J. 1999. “U.S. Economic Growth Since 1870: One Big Wave?” American Economic Review 89: 123-128. Gordon, Robert J. 2000a. Macroeconomics, 8th ed. Reading: Addison-Wesley Longman.
Gordon, Robert J. 2000b. “Interpreting the “One Big Wave” in U.S. Long Term Productivity Growth,” in Bart van Ark, Simon Kuipers, and Gerard Kuper, eds., Productivity, Technology, and Economic Growth. Boston: Kluwer, pp. 19-66. Gordon, Robert J. 2000c. “Does the “New Economy” Measure up to the Great Inventions of the Past?” Journal of Economic Perspectives 14 (Fall): 49-74. Gramlich, Edward M. 1994. “Infrastructure Investment: A Review Essay.” Journal of Economic Literature 32 (September): 1176-1196. Handy, Bruce. 2000. “Tomorrowland Never Dies.” Vanity Fair. March: 114-126. Kealey, Terence. 1996. The Economic Laws of Scientific Research. New York: St. Martins Press. Kendrick, John. 1961. Productivity Trends in the United States. Princeton: Princeton University Press. Kleinknecht, Alfred. 1987. Innovation Patterns in Crisis and Prosperity: Schumpeter’s Long Cycle Reconsidered. New York: St. Martin’s Press. Mensch, Gerhard. 1979. Stalemate In Technology : Innovations Overcome the Depression. Cambridge: Ballinger. Morton, M. 1982. “History of Synthetic Rubber.” In Raymond B. Seymour, ed., History of Polymer Science and Technology. New York: M. Dekker. Mowery, David and Nathan Rosenberg. 1989. Technology and the Pursuit of Economic Growth. Cambridge: Cambridge University Press. Mowery, David and Nathan Rosenberg. 2000. “Twentieth Century Technological Change.” In Stanley Engerman and Robert Gallman, eds., The Cambridge
Economic History of the United States, volume III. Cambridge: Cambridge University Press, pp. 803-926. Nelson, Richard. 1959. “The Simple Economics of Basic Scientific Research.” Journal of Political Economy” : 297-306. Nordhaus, William. 1982. “Economic Policy in the Face of Declining Productivity Growth.” European Economic Review (May-June): 131-158. Oliner, Steven D. and Daniel E. Sichel. 2000. “The Resurgence of Growth in the Late 1990s: Is Information Technology the Story?” Journal of Economic Perspectives 14 (Fall): 3-22. Olmstead, Alan and Paul Rhode. 2000. “The Transformation of Northern Agriculture, 1910-1990.” in Stanley Engerman and Robert Gallman, eds., The Cambridge Economic History of the United States, volume III. Cambridge: Cambridge University Press, pp. 693-742. Rostow, W.W. 1960. The Stages of Economic Growth: A Non-Communist Manifesto. Cambridge: Cambridge University Press. Solow, Robert J. 1957. “Technical Change and the Aggregate Production Function.” Review of Economics and Statistics. 39 (August): 312-20. Temin, Peter. 1989. Lessons from the Great Depression. Cambridge: MIT Press. U.S. Bureau of the Census. 1975. Historical Statistics of the United States, Colonial Times to 1970. Washington: Government Printing Office. U. S. Reconstruction Finance Corporation. 1946. “Projects Approved by the RFC – Office of Defense Plants as of June 30, 1946.” Mimeo, 113 pages.
U.S. War Production Board, 1945. Wartime Production Achievements and the Reconversion Outlook. Report of the Chairman, War Production Board, October 9, 1945. Washington, D.C.: Government Printing Office. Vernon, J. R. 1994. “World War II Fiscal Policies and the End of the Great Depression.” Journal of Economic History 54 (December): 850-68.