TECHNOLOGICAL

FORECASTING

AND SOCIAL CHANGE

17, 283-311

(1980)

A Long-Wave Hypothesis of Innovation

ALAN K. GRAHAM and PETER M. SENGE

ABSTRACT The flow of basic innovations into developed economies appears to occur in waves. Brief periods of high receptivity to basic innovations occur every 40-60 years; in between, economic conditions favor less radical improvement innovations. This paper presents a theory and evidence showing how long waves can develop from overexpansion and collapse of capital-producing sectors of an economy. Each wave of capital expansion utilizes a new ensemble of technologies. The economy appears currently to be near the end of a long-wave expansion, which may explain much of the current malaise in innovation, investment, and productivity. Both national innovation policies and research management can benefit by understanding long-wave changes in the economy.

Introduction Worldwide concern has developed in recent years over slackening productivity gains and declining technological innovation. In the United States, productivity per manhour increased by an average of 3.2% per year from 1947 to 1966; the rate of increase slowed to 2.1% from 1966 to 1973 and was down to 0.8% per year from 1973 to 1979 [l]. During 1979, productivity fell steadily [2]. Declining productivity growth is often attributed to a slackening pace of major technological innovations. In the words of a National Science Foundation study in 1974, the vast majority of innovations between 1953 and 1973 or minor product or process differentiation [3, represented mundane improvements, pp. 30-311. Only a small fraction (less than 0.5%) represented basic (i.e., new marketcreating) innovations. By contrast, many studies have shown a much higher percentage of basic innovations during the 1930s and 1940s. Not only the percentage, but the number of basic innovations was greater during the 1930s and 1940s than during recent decades, despite ever-increasing efforts to innovate during the postwar period: R&D personnel in the United States rose every year from 1955 to 1976 [4]. This paper suggests that lagging productivity growth and innovation is a natural consequence of a long wave in economic behavior whose period ranges from 40 to 60 years. This long wave is characterized by buildup, overexpansion, and relative decline of capital-producing sectors. The long wave creates a shifting historical context for the implementation of new inventions. Midway into a capital expansion, opportunities for applying new inventions that require new types of capital become poor. The nation is already committed to a particular mix of technologies and the environment greatly favors improvement innovations over basic innovations. During a long-wave downturn, basic innovation opportunities gradually improve, as old capital embodying the technologies of
ALAN K. GRAHAM and PETER M. SENGE are, respectively, Research Associate and Assistant Professor in the System Dynamics Group at the Sloan School of Management, Massachusetts Institute of Technology. @ Elsevier North Holland, Inc., 1980 0040-1625/80/08028329$02.25

284

ALAN K. GRAHAM

AND PETER M. SENGF

the preceding buildup depreciates. Near the trough of the wave, there are great opportunities for creating new capital embodying radical new technologies. The old capital base is obsolescent, bureaucracies that thwarted basic innovation have weakened, many companies committed to producing old types of capital are bankrupt, and traditional methods are no longer sacrosanct. This explains why times of economic depression, such as occurred in the 183Os, 188Os, and 1930s in many developed economies, are periods of unmatched basic innovation. Long waves in economic growth and innovation have been discussed by many economists. However, most previous studies have been primarily empirical and few convincing theories to explain observed long-wave behavior have been advanced [5]. Some economists, such as Schumpeter [6] and Mensch [3], have developed theories that relate innovation to the long wave. The present work differs from past innovation theories of the long wave in emphasizing expansion of capital-producing sectors as a primary cause of the long wave. Although long waves can be explained without new innovations as an explicit causal factor, the long-wave theory has important implications for innovation. The long-wave theory just outlined derives from our work on the System Dynamics National Model, a large computer simulation model that attempts to show how major patterns of macroeconomic behavior arise out of microlevel decision making. The next section describes the model and develops the long-wave theory and its implications for innovation. The third section examines a variety of empirical data that appear to support the theory. The fourth section identifies general implications for innovation policy in three areas: macroeconomic policies to stimulate investment, corporate research strategy, and social and managerial innovation. The last section summarizes the theory and its implications. The Long-Wave Theory
THE SYSTEM DYNAMICS NATIONAL MODEL (SDNM)

The central long-wave mechanisms in the System Dynamics National Model (SDNM) come from the interactions of a capital-goods-producing sector and a consumergoods-producing sector. Each sector is represented by a detailed model of a firm, with production, factor ordering, and inventory and backlog management. (The model used here simplifies the full SDNM by omitting price setting, money flows, and financial accounting.*) Figure 1 indicates the major linkages between the two sectors. The consumer-goods sector responds to demand for consumer goods . 3 Both the consumer-goods and the capital sectors utilize labor and capital as factors of production. Labor is supplied at a wage and hiring delay that vary in response to relative demand and supply of additional workers. Capital (the aggregate of plant and equipment) is supplied by the capital sector, in accordance with the capacity of the capital sector to meet demand. An important feature of the model is the self-ordering that takes place in the capital sector. The aggregate capital sector must obtain capital from itself. Orders for capital generated by the capital sector

The theory described subsequently For a fuller description

is also described by Forrester [7].

of the model, see Forrester et al. [Xl.

Consumer demand is determined exogenously in the present model to focus on the interaction of the two production sectors. The full SDNM contains a separate household sector which generates consumer demand, The overall size of the work force is determined outside the model.

A LONG-WAVE

HYPOTHESIS

OF INNOVATION

285

ORDERS DELIVERY OF CAPITAL

FOR

CONSUMER DEMAND

Fig. 1. Major

interactions

between capital sector and consumer goods sector in the System

Dynamics

National Model.

combine with orders generated by the consumer-goods sector to determine the total capital demand. Likewise, shipments of capital are received by both sectors.
LONG WAVES IN CAPITAL EXPANSION

Figure 2 shows the behavior generated by the model in response to small random fluctuations in consumer demand.5 The random fluctuations are superimposed on a constant rate of consumer demand. The figure shows several modes of fluctuating behavior over the 160 years of the simulation. Labor and shipments in the consumer-goods sector (Fig. 2a) primarily exhibit a short-term cycle with periods between 3 and 10 years. For example, this periodicity is clearly evident in the behavior of shipments between years 78 and 96, when there are three distinct cycles. Model-generated variables are quite similar to observed short-term business cycles in period and relative timing of fluctuations [9111. Extensive analysis of the model has shown that interactions between employment decisions and inventory and backlog management are most important in generating the short-term business-cycle behavior. A second longer-term mode of oscillatory behavior is also evident in the simulation. This long-term fluctuation is most apparent in the behavior of capital-sect& variables such as the stock of capital in the capital sector (Fig. 2b), which has distinct peaks at years 20,

SThe standard deviation of the random component is 2.5% of the underlying deterministic component. random component has an exponential autocorrelation function with a 0.25-year time constant.

The

sFor example, interactions between investment and consumption and between investment and monetary policy, often assumed to be critical to the business cycle, appear to be less critical than inventory-employment interactions. See Mass [93 and Senge [ 111.

ALAN K. GRAHAM

AND PETER M. SENGE

Ttme

CdPlTAL OOQl_ 0 .._.__.... ._. 20 40

STOCK 6d 80 Time .100. 120 140 i60

Fig. 2. MO-year simulation of coupled capital- and consumer-goods-sector consumer-goods sector and b) variables for capital-goods sector.

model. a) Variables

for

and 124. This long-wave behavior is also present in delivery delay for capital and production of capital, as well as in backlog of orders for capital in the goods sector. 7 Causes of the long wave appear to reside in basic features of the capital sector and its relation to the rest of the economy. Figure 3 shows three positive feedback loops involved in the self-ordering of capital from the capital sector. The first loop states that when
68,

The simulation also presents some evidence of an intermediate term fluctuation with a period between IS and 30 years-for example, note the capital production fluctuations between I20 and 160. This fluctuation appears to match the so-called Kuznets cycles observed in most developing countries. See Mass [9] and Senge [ 121,

A LONG-WAVE

HYPOTHESIS OF INNOVATION

287

average sales of capital rise, producers attempt to increase production to meet the higher sales demand. Desired production increases, which raises the desired stock of capital in the sector (given a desired capital-output ratio). A higher level of desired capital leads to more orders for capital. As these orders are eventually filled, sales rise, causing further increases in desired production. The second positive loop shown in Figure 3 traces a parallel reinforcing process involving inventory of finished capital (finished inventories of equipment and completed structures). A decline in inventory relative to some target inventory level boosts desired production as producers attempt to increase output to replenish inventory. The increase in desired production feeds back to boost capital orders and leads to still further inventory decline. The last positive loop in Figure 3 shows a parallel response involving the backlog of unfilled orders in the capital sector. As backlog for capital increases (representing increased unfilled demand for capital), this tends to increase desired production, orders for capital, and backlog still further. Figure 4 shows two more reinforcing positive feedback loops involving the desired mix of capital and labor. When desired production in the capital sector increases, desired labor increases as well as desired capital. The sector creates more job vacancies, which eventually increases employment. Given a limited number of workers available for employment, this tends to reduce unemployment and to increase the delay in filling job vacancies. As competition among firms for the remaining workers increases, average wage levels rise. The higher wages encourage substitution of capital for labor. Increased desired capital-labor ratio boosts desired capital stock in both sectors, thereby increasing orders for capital. This eventually feeds back to boost desired production still further through the channels shown in Figure 3. As the capital sector expands, rising delivery delays for capital further feed the expansion, through the feedback loop shown in Figure 5. Rising order backlog in the capital sector boosts delivery delay, as capital producers have increasing difficulty meeting current demand. Rising delivery delay forces many customers to order further ahead

Fig. 3. Positive feedback loops affecting desired production

/FOR ORDERS CAPITA y\\

in the capital sector.

1
DESIRED CAPITAL IN CAPITAL SECTOR + (+I /

AVERAGE SALES IN CAPITAL SECTOR

\+

INVENTORY IN CAPITAL SECTOR /

\ -

IN CAPITAL SECTOR

sThe small negative sign adjacent to the arrow linking inventory to desired production indicates an inverse effect: A decline in inventory increases desired production, and increase in inventory reduces desired production.

288

ALAN K. GRAHAM

AND PETER M. SENGE

-+ ORDERS
+ :%yR CAPTAL\+

DESIRED CAPITAL IN CAPITAL SECTOR + / / DESIRED CAPITALLABOR RATIO

DESIRED PRODUCTION IN CAPITAL SECTOR

+
(+)

t+

LABOR IN CAPITAL SECTOR

\
WAGES +\,,,,Y ,N PLOhENT

FILLING

VACANCIES

Fig. 4. Positive feedback loops affecting desired capital-labor

ratio in both sectors.

and to place orders with multiple suppliers in an attempt to ensure reasonable lead times in capital acquisition.g Ordering further ahead and placing multiple orders results in increased desired capital on order, increased orders for capital, and further increase in order backlog in the capital sector. Capital price, although constant in the present model, potentially creates still further reinforcing pressures. Rising delivery delays for capital tend to push up the price of capital. As shown in Figure 6, the resulting increase in return on investment encourages orders for capital through two channels. High return directly encourages producers to invest in order to take advantage of profit-making opportunities. High return is also a key index of credit worthiness and thus encourages financial investors to support capital expansion. Both responses to high return on investment increase orders for capital and backlog of unfilled orders in the capital sector, and further drive up capital delivery delays.
Fig. 5. Positive feedback loop affecting capital delivery delay and capital ordering.
+ ORDERS FOR CAPITAL

/ DESIRED ChPlTAL EQUIPMENT ON ORDER

+ (+I BACKLOG OFORDERS IN CAPITAL SECTOR

DELIVERY DELAY CAPITAL SECTOR

IN

9Mitchell [ 131 gives a detailed account of the illusory

demand

created by rising delivery delays

A LONG-WAVE

HYPOTHESIS

OF INNOVATION

289

ORDERS +

SUPPORTABLE FINANCING IN CAPITAL SECTOR +( ~;;~Fzf;

RETURN ON INVESTMENT IN CAPITAL SECTOR

CAPITAL /

SECTOR

PRICE OF CAPITAL EOUIPMENT

Fig. 6. Positive feedback loops involving return on investment

in capital sector.

Figure 7 combines the separate loops previously described into a composite picture of the reinforcing pressures that underlie expansion of the capital sector. The figure provides a useful backdrop for describing long-wave dynamics. During a long-wave expansion, rising orders for capital increase desired production in the capital sector and lead to still more orders for capital. Imagine an expansion initially stimulated by an increase in consumer demand. The consumer-goods sector must acquire more capital. In responding to this increased demand, the capital sector in turn must expand its own capital stock, which adds still further to the demand. Initially, expansion in capital production relies heavily on expansion in employment. After a period, available labor is reduced and wages begin to be bid up. As producers gradually perceive a persisting increase in labor costs relative to capital costs, desired capital intensity begins to rise, adding further to capital ordering. At the same time, rising order backlogs and delivery delay force many producers to order capital further ahead, whereas rising capital price is creating a high return on investment and thereby further supporting expansion. The reinforcing processes created by the positive feedback loops in Figure 7 cannot continue to generate growth forever. Eventually the capital sector expands well beyond the point required to supply the capital demands of the consumer-goods sector and to maintain its own capital stock. Once this occurs, the reinforcing processes that created rapid growth reverse. Falling capital orders reduce desired capital production and further reduce capital orders. Falling desired production lowers labor demand, increases unemployment, and brings down wages. Low wages encourage substitution of labor for capital in both sectors, further lowering orders. Falling capital orders also lower backlog in the capital sector and cause delivery delay to turn down. Desired capital on order declines along with capital price and return on investment, further lowering orders. This rapid reversal of forces is evident in the precipitous decline in capital production accompanying each long-wave downturn in Figure 2b. The sharp cutback in capital production caused by a long-wave downturn eventually leaves the system with inadequate capital-production capability. The stage is then set for another long-wave expansion. More importantly now, the stage is set for creating new types of capital embodying new technologies. As the overexpanded capital stock from the preceding expansion depreciates, unparalleled opportunities emerge to replace the old capital with radically new types of capital.

,SUPPORTABLE FINANCING IN CAPITAL SECTOR

DESIRED CAPITAL IN GOODS SECTOR (+) / / / DESIRED CAPITAL IN CAPITAL SECTOR DESIRED PRODUCTION IN CAPITAL SECTOR

RETURN bN INVESTMENT IN CAPITAL SECTOR 4

FoR\Y
(+I
LABOR IN CAPITAL SECTOR / UNEMPLOYMENT / DELAY IN VACANCIES

BACKLOG 0F ORDERS IN CAPITAL SECTOR

I/
DESIRED CAPITAL LABOR RATIO k

PRICE OF CAPITAL EOUIPMENT

Fig. 7. Positive feedback loops affecting capital ordering.

A LONG-WAVE A LONG-WAVE

HYPOTHESIS HYPOTHESIS

OF INNOVATION OF INNOVATION

291

Innovation activity has not yet been incorporated explicitly into the System Dynamics National Model. Nevertheless, the long-wave behavior already examined produces a sequence of economic circumstances which suggest a tentative hypothesis for how innovation might vary over the wave. lo Relating the long-wave theory to innovation requires first distinguishing between invention and innovation. Following accepted practice, we refer to invention as the discovery of a new idea or technical process, and to innovation as the first practical application of the invention on a significant scale. We also must distinguish basic innovations from improvement innovations. Mensch [3, p. 471 defines basic innovations as
innovations which produce new markets and industrial branches cultural sphere, in public administration, and in social services. human activity.

or open new realms of activity in the Basic innovations create a new type of

In contrast, improvement innovations can be thought of as incremental improvements on an existing technology that do not alter its fundamental nature. In order to examine the implications of the long wave for innovation, consider first the conditions present at a trough in the long wave. The economic climate is highly uncertain and very little investment is made in new capital. There are virtually no market forces that would impel inventions through the development process and out into the market. Moreover, managers and investors will tend to avoid the uncertainty of new innovations. But as a new expansion wave begins, investment picks up and the future looks brighter. There are greater incentives to take risks, for demand is growing. As new technologies are tried and succeed, incentives for risk taking improve still further. It is the period of late depression and early upswing following the Great Depression of the 1930s that produced nylon, jet airplanes, helicopters, electronic computation, radar, television, catalytic petroleum refining, and many other important innovations. As the long-wave upswing continues, the nature of capital investment and innovation gradually changes. As physical capital is put into place, it embodies the earlier basic innovations. For automotive technology, a whole system of paved roads, factories, service stations, and car dealers was gradually put into place. Homes and businesses began to locate in suburbs to take advantage of the automobile. For television, transmitters and home sets slowly spread. The physical capital gradually becomes committed to the earlier technology. There is an increasing social and managerial commitment as well, as society learns to deal with the innovations. People begin to make careers out of auto repair. Dealerships in parts, crude oil, and gasoline are organized. Financial institutions adapt to the specific needs of the technology: finance companies emerge that do virtually nothing but auto loans. Engineers learn how to make cars very efficiently. The earlier basic innovations are followed by numerous improvement innovations. As the economy develops a physical and managerial infrastructure committed to a particular mix of technologies, incentives shift away from further basic innovations toward improvement innovation. An improvement innovation has a ready-made market, so that even minor improvements may be worthwhile. In contrast, a radically different basic innovation faces a hostile reception. There is no demand for it, people may not know how to market or finance it, customers may not understand it, there are imperfections, one may not be able to use current manufacturing technology to make it, and development enMany of the ideas in this section are also developed by Forrester

[ 141

292

ALAN K. GRAHAM

AND PETER M. SENGE

gineers probably do not understand it. In this light, it is not surprising that attempts to introduce radically new transportation technologies or to alter private automobiles substantially face enormous resistance if these attempts occur when the automobile economy is healthy. By the midpoint of a long-wave expansion, innovation has become institution. What little headway is made by determined innovators on a new product or process usually falls far short of mass acceptance. Thus, as the upswing progresses, basic innovations become less frequent. This is not to say that the scientific progress on basic inventions slows down. Rather, the basic inventions do not become commercialized and available. A backlog of untapped technology builds up, awaiting the next upturn. Most of the innovations of the 1930s and 1940s rest solidly on scientific achievements of earlier decades. At the end of a long-wave upturn, a number of cross-currents confuse the economic and technical scene. Despite a well-developed body of science, opportunities for investments in new technologies are typically scarce. Despite decades of progress, growth in investment and productivity falters. A number of crosscurrents can arise from the long wave of technology reaching its limits to growth. The economically minded might call this a case of decreasing returns to the capital factor: every time one increases the capital stock per person, there is a smaller increase in production. The politically minded identify government regulation as the limitation on growth. Since the 193Os, the use of autos, airplanes, power tools, powerful chemicals, and pharmaceuticals have all increased steadily and dramatically. It should come as no surprise that these increases have created problems that in turn have created regulations on emission standards, airport noise, occupational health and safety, and licensing restrictions on new drugs. Finally, the ecologically minded point to the limits to growth imposed by declining natural resources. As a particular type of capital expands, the finite recources it requires become more difficult and costly to acquire. At the long-wave peak, then, the technologies of that wave have spread throughout the economy. It is to be expected that they would encounter limits to growth along a number of dimensions, just as appears to be happening today. As the long-wave peak unfolds into a depression, however, the climate for basic innovation gradually shifts. Capital embodying the technology of the preceding long wave is not replaced. Companies go out of business. Bureaucracies that resisted radically new ways of doing things weaken. During such times, determined innovators find the forces opposed to change far less formidable. Once a large enough portion of the workers previously involved in automobile-related tasks are unemployed, there would probably be much greater willingness for employment in constructing alternative transportation systems. Hence, the long-wave theory suggests a shifting environment for innovation. The theory suggests sharp clustering of basic innovations near a long-wave trough. It also suggests that, once the pattern of technologies characterizing a particular long wave has been established, society finds it much easier to accommodate improvement innovations that do not threaten basic technological mixes but which offer marginal improvements in efficiency. Once under way, a long wave may give rise to several generations of improved automobiles, televisions, airplanes, or computers, rather than to a radical technology that supplants an established technology. Empirical Evidence The preceding section has attempted to show that the structural assumptions underlying the long-wave theory plausibly describe real economic interactions. This section

A LONG-WAVE

HYPOTHESIS

OF INNOVATION

293

considers evidence for the long-wave behavior produced by these assumptions. The following section briefly discusses evidence for the long wave in economic aggregates such as commercial construction and unemployment, and then examines evidence of long-wave behavior in innovation.
LONG WAVES IN MACROECONOMIC BEHAVIOR

One of the first economic historians to identify economic fluctuations with a 40-60year period was the Russian economist Nikolai Kondratieff [5]. l1 From his examination of long-term time series, especially price levels, covering the nineteenth century and the first part of the present century, Kondratieff concluded, There is, indeed, reason to assume the existence of long waves of an average length of about 50 years in the capitalistic economy. . . . Because of Kondratieffs pioneering work, the long wave is often called the Kondratieff wave. Similarly, Simon Kuznets, perhaps the most respected ecdnomic historian of the past quarter century, built on the earlier work of Schumpeter to hypothesize the following dating of the long wave, beginning at the end of the eighteenth century [ 171:
Prosperity 1787-1800 1843-1857 1898-1911 Recession 1801-1813 1858- 1869 1912-1924-5 Depression 1814-1827 1870- 1884-5 1925.6- 1939 Revival 1828- 1842 1886-1897

Kuznets noted, as had others before him, that each long-wave cycle was associated with a particular mix of technologies-cotton textiles, iron, and steam power for the industrial revolution wave from 1787 to 1842, wood-powered railroads for the bourgeois wave from 1843 to 1897, and coal-powered railroads, electricity, and automobiles for the neomercantilist long wave starting in 1898. More recently, the Dutch economist J. J. van Duijn has extensively reviewed the long-wave evidence and literature and proposed a revised long-wave dating which extends up to the present day. His dating is as follows [ 15, p. 5631:
Prosperity 1783-1803 1847- 1866 1893-1913 1949-1967 Recession 1815-1826 1866-1875 1921-1929 1967-1975 Depression 1826- 1837 1875-1884 1929-1938 Recovery 1837-1847 1844- 1893 1938-1949

Datings such as those of van Duijn and Kuznets are based on diverse statistical and nonstatistical data drawn from different countries and do not distinguish behavior of capital-producing sectors from overall economic activity. l2 Hence, they cannot be related directly to long-wave behavior generated by the System Dynamics National Model. Un-

Van Duijn Mandel [ 161.

[ 151 points out that several earlier economists had also identified long-wave behavior. See also

Kuznetss estimates of long-wave behavior before 1900 are based primarily on data for Great Britain; his estimates since 1900 are based on data for many countries, including the United States. Van Duijn attempted to consider the scanty U.S. data from about 1870.

294

ALAN

K. GRAHAM

AND PETER M. SENGE

fortunately, the desired long-term data for the capital-producing sectors of a hingle economy appear to be unavailable. Figure 8 plots data for capital and labor inputs to overall production in the United States. Both upper and lower graphs trace a distinct pattern of economic evolution. The graphs suggest a four-stage development pattern. For example. the upper graph shows that
Fig. 8. Data for capital-labor allocation in the United States. a) For 1889-1938 data (Source: Kendrick [18]); b) for 1947-1979 data (Source: Bureau of Economic Analysis [191). ,,O,;;;;;;;;;;=;=; ii==z;i=zzC :I::,:iii iiiii:iiii==zzz,zi ====== ==== =======,==== z;i:z Cz;=z ~ ====

100:

.i

,....,....

:...:
1918

1929:
~

.. . . ..~.......~.~__.......~.....~.

30

40 50 Capital Stock

60
Index

70
(1929=100)

a0

90

100

110

1969

:I973

16

Capital

Stock

(bIllIons

of 1972 dollars)

A LONG-WAVE

HYPOTHESIS

OF INNOVATION

295

labor and capital both expanded from 1889 to 1918. This corresponds to the early stage of a long-wave expansion in which all factors are expanding. From 1918 to 1929 labor growth slackened while capital continued to expand. This corresponds to the period in which capital is being substituted for labor due to the relatively low price and delivery delay for capital. Labor then declined from 1929 to 1932, while capital held more or less constant. Once the long-wave peak in capital production has occurred, reducing production requires massive cutbacks in employment, particularly in the capital-producing industries. Lastly, the period from 1932 to 1939 was characterized by rising labor and declining capital stock. This corresponds to the later stages of the long-wave decline in the model in which capital-labor ratios decline due to the relatively low cost and high availability of labor. The lower graph in Figure 8 suggests a similar pattern in the United States since World War II.13 The data show that labor and capital in U.S. manufacturing industries were on a general growth path until 1969, although the post-Korean War period showed relatively little employment growth. From 1969 to 1979, there was relatively little expansion in total manufacturing employment, while capital continued to expand. l4 This same pattern has been observed in several other industrialized countries. I5 For comparison, Figure 9 presents similar data taken from the model simulation shown in Figure 2. Along the horizontal axis in Figure 9 are values for the total capital stock in the two production sectors; along the vertical axis are values for total labor in the two sectors. The data cover the long wave which begins just after year 40 in Figure 2. In comparing the model-generated and actual data, one must keep in mind that the present simplified model generates long-wave behavior about a nongrowing underlying trend in economic activity. By contrast, long-wave behavior in reality occurs superimposed on an underlying growth trend. Consequently, the model tends to underestimate long-wave expansions and to overstate long-wave contractions. Nonetheless, Figure 9 shows that each of the four stages of long-wave development previously described are present in the model behavior. l6 Figures 10 and 11 present data for two physical time series for which long-term U.S. data are available: unemployment rate and total construction per capita [22]. The series show major economic downturns at about 40-year intervals. Figure 10 shows two clear peaks in unemployment at 1894 and 1933. The unemployment rate in 1975, 8.5%) was the highest since the Great Depression. Similarly in Figure 11, if one ignores the declines in construction caused by the two world wars, the major declines in construction occurred between 1892 and 1896 (34%) and between 1926 and 1933 (77%). The third largest nonwar decline occurred between 1973 and 1975-a 25% drop. l7

13Capital stockestimatesafter 1947 are based on assumed constant service lives, as estimated by Coen [20]. Initial (1947) value of capital stock is based on that of Wasson [21].
I4 Total U.S. employment continued to expand in the 1970s due to the high growth in employment in services and government. Thus, the long-wave dynamics are most evident in manufacturing, as suggested by the theory in the second section. I5 Mensch [3] chapter 9 shows similar data for Great Britain, West Germany, sRespectively, years 41-51, 51-63, 63-73, and 73-83. Japan, and Holland.

It should be noted that the time series do not match Kuznets and van Duijns datings of the long-wave downturn in the late nineteenth century. The United States data suggest a long-wave trough in the mid- 1890s;

296

ALAN K. GRAHAM

AND PETER M. SENGE

it(

,44.47 -45

.@=j:
1, .., : ..,

;
I I

1.

i .,.,

9001

I.

1:
),

; . . . .
: : : :

:76. ._,...:...,

._

. . . . . . . .

**o;. i.
: \

86&..1.

;_

: ..,.,_,...,...,...,... :

Y+ v:
I.

70: I

: I : I : :

,_.,...,

..I..

.I.

i..:::
:-:.i :

p,..

1
:

i
/

.73 :

79:

840

;:~iiiii;iiii:i 600

: :_____ _ __.____ __________.___ __I__ ____:______.:_______:___....:_______: ~:_____I_:_______I.__~--.:-..-.-.~--.___i:._ __:____ __.____ .______ ______. __.____ 650 700 750 800 850 900 Tot01 Copitol Stock allocation.

Fig. 9. Model-generated

data for capital-labor

The data presented in Figures 10 and 11, as well as the data presented in Figure 8, suggest that the United States may once again be moving into a period of capital excess and a potential major economic downturn. Further examination of post-World War II data reinforces this supposition. Figure 12 shows that the total private investment as a fraction of GNP has been on a downward trend since the mid-1960s. The ratio of investment to GNP peaked at 8.5% in 1966. Since then, a lower peak value has been reached at each successive business cycle. Investment equaled less than 7.4% of GNP in the first quarter
Kuznets and van Duijn date the long-wave trough in the late 1870s and 1880s. This difference may be explained by noting the different countries involved. Kuznetss dating scheme is based on British datauntil about 1900. Van Duijn considered U.S. data from about 1870, but his dating is mostly influenced by the superior European data. The United States experienced a substantial downturn in the mid-1870s, as evidenced by the minimal growth in the construction estimates from 1869 through the mid-1870s. However, the limited data prevent us from determining clearly whether the 1870s or the 1890s should be viewed as the long-wave trough in the United States. The absence of closely coordinated major depressions in the latter half of the nineteenth century may be explained by the more limited coupling through trade in the 1870s~1880s than in the 1930s. Some scholars have argued that the rapid westward expansion in the United States bouyed the economy sufficiently during the latter half of the nineteenth century that a representative long-wave downturn never occurred. Whichever, the limited data available do suggest long-wave forces sufficient to produce declines in capital expansion and employment during the 1890s which are much more severe than typical business cycle declines.

A LONG-WAVE HYPOTHESIS OF INNOVATION

291

Fig. 10. Historical data for unemployment

rate.

of 1979, the apparent peak for the present business cycle. Figure 13 shows a similar downward trend for capacity utilization since the mid- 1960s. Figures 14 and 15 plot data concerning the relative incentives to shift the capitallabor mix in the economy. The two variables computed-indices of labor return to cost and capital return cost-attempt to measure the profitability of expanding output by adding one unit of labor and one unit of capital, respectively.18 For example, rising labor

Fig. 11. Historical data for total construction five-year averages).

per capita in 1929 dollars (data prior to 1889 based on

sRetums to cost are defined as the marginal revenue product divided by the marginal cost for productive factor. Marginal products am computed under the assumption of a Cobb-Douglas production nology, an assumption consistent with data for aggregate industrial groupings, as shown by Jorgenson Indices for return to cost are plotted because marginal revenue products and marginal costs are based on price indices, rather than absolute price levels.

each tech[23]. both

298

ALAN K. GRAHAM AND PETER M. SENGE

1947

51

55

59

63 Years

67

71

75

79 of Com-

Fig. 12. New plant and equipment

expenditures

merce, Bureau of Economic Analysis, new plant and equipment expenditures,

as a fraction of GNP-U.S. Department all industries.

return to cost indicates increasing marginal profit from adding a unit of labor. In the model, capital return to cost begins to decline about halfway up a long-wave expansion, when incentives to substitute capital for labor stop increasing. Conversely, labor return to cost gradually begins to rise as a long-wave expansion progresses and pressures to increase capital intensity give way to pressures to increase labor intensity. Figures 14 and 15 show similar long-term trends in the economy. Figure 14 shows a long-term increase in incentives for labor intensive production as measured by a rising index of labor

Fig. 13. Capacity utilization in total manufacturing-Governors, utilization rate, all manufacturing industries.

Federal Reserve Board, capacity

:
70; 1948 I 52.. 56

?
60 64 Years 68 72 76

A LONG-WAVE HYPOTHESIS OF INNOVATION

299

085
:

0.7c

0.5:

1953 1955 1957 1959 1961 1963 1965 1967 1969 1971 1973 Years
Fig. 14. Empirical estimates of labor return to cost.

I 5

return to cost until 1975. Figure 15 shows that incentives to substitute capital for labor have generally abated since the mid-1960s. Much has been said in recent years about the impact of higher energy prices on capital intensity. These figures clearly show that the current pressures toward more labor-intensive production have historical antecedents which far predate the energy crisis.
LONG WAVES IN INNOVATION

Gerhard Mensch has presented a variety of data on long-term trends in innovation which are consistent with the long-wave hypothesis. Figure 16 shows the frequency of basic innovations in Western countries in 22 IO-year periods from 1740 to 1960. Figure

Fig. 1s. fimpirical 2 I i.-~----~~ ------~ : :

estimates of capital return to cost.

: : : : ------: :-+

.-~-------~-----~-----------------.----~.------------: :

,2i

::::,,_

,,,.,,.(,
1957 1959 1961 1963 1965 1967 1969

:,,,
1971 1973 197!

1953

1955

Years

300

ALAN K. GRAHAM

16161412lo6642o,, 1740 50 n

._ - --

. - I

1
w

1800

1850

1900

Calendar

Year
1740-1960

e
Trend i -Mean -I

AND PETER M. SENGE

1950

Fig. 16. Frequency of basic innovations,

[3,

p. 1301.

16 shows that there are distinct periods in history that uniquely favor basic innovations: in the 176Os, the 1820s and 183Os, the 1880s and the 1930s. These periods of intense innovation correspond to troughs in the long wave. Thus, the innovation frequency data are consistent with the hypothesis that the long wave creates unique periods of opportunity for implementing new inventions. Only when a significant portion of the capital from a previous long-wave buildup has depreciated is the door open for creating new types of capital embodying fundamentally new technologies. The correlation of capital expansion waves with basic innovation can also be seen in Figure 17, which plots the percentage of total energy use contributed by three different energy sources: wood, coal, and petroleum and natural gas. Shifts in the dominant energy source are a clear indication of shifts in the type of capital in use. The figure shows three

Fig. 17. Long waves in dominant energy source and basic innovations PETROLEUM 8

[24].

CALENDAR

YEAR

A LONG-WAVE

HYPOTHESIS

OF INNOVATION

301

distinct waves in the dominant energy source-a wood-burning wave, a coal-burning wave, and a petroleum and natural gas wave. Menschs basic innovation data are superimposed on the energy data in the figure. The basic innovation surges clearly correlate with the emergence of each new wave in energy usage. ls One alternative explanation for the long-wave behavior in Figures 16 and 17 is the echo hypothesis: that surges in basic innovations and shifts in energy technology simply arise from prior surges in basic invention~.~~ The data in Figure 18 disprove the echo hypothesis. The figure compares the frequency of basic inventions and basic innovations underlying the innovation surges in the mid-1820s, mid- 188Os, and mid 1930s. Figure 18 shows that the two processes differ sharply in their degree of continuity. 21 The flow of basic inventions is much smoother than the flow of basic innovations. Consider an extreme case of the echo hypothesis, in which the conversion of basic inventions into basic innovations could be characterized as a pure pipeline process: All basic inventions take the same number of years N to become basic innovations. If so, each set of frequency curves for inventions and innovations in Figure 18 would have identical shapes. If the length of time for application of basic inventions were different for different inventions, the dispersion in time for basic innovations would be greater than the dispersion for basic inventions. Figure 18 shows the opposite characteristic: The dispersion in basic innovations is distinctly less than the dispersion in basic inventions. Such behavior cannot be explained by any simple echo hypothesis relating invention to innovation. However, the relationship between invention and innovation shown in Figure 18 is exactly as predicted by the long-wave theory. The theory states the long-wave behavior alters the climate for innovation, tending to bunch innovations in long-wave troughs and the early periods of a new wave of capital expansion. The theory as articulated in the preceding section suggests little about the pattern of basic invention. Rather, it simply says that the opportunities for a new long-wave expansion will lead society to draw down quickly the backlog of unexploited inventions, whatever the pattern of past invention activity.23

I9 Similar conclusions [253.

can be drawn from the data on successive waves in energy use presented by Marchetti is the distinction between

*a noted earlier, the distinction between basic inventions and basic innovations AS scientific discovery and practical application.

*rNote that Mensch divides innovations in the second half of the nineteenth century into two groups, electrotechnical and chemotechnical, but combines innovations into one plot for the other two long waves.
Mensch uses an analogous statistical argument to reject the echo hypothesis. He shows that there is a distinct probability (approximately 0.3) that the apparent clusters in basic inventions in Figure 18 are random. The probability that the clusters in basic innovation are random is very small. He rejects the echo hypothesis on the basis that a random phenomenon cannot cause a nonrandom phenomenon. 23The theory could be readily extended to include invention activity, insofar as such activity is the result of organized research effort. Research expenditures tend to be greatest during periods of the greatest economic prosperity-that is, during long-wave expansions. Hence, one would expect a higher frequency of basic inventions during long-wave expansions, which is born out in the Figure 18. This hypothesis is consistent with the high correlation between inventions (patents) and economic activity shown in Schmookler [26]. According to this view, the decline in the dispersion of basic inventions for each successive cycle may reflect the increasing organization of the research effort.

302

ALAN K. GRAHAM

AND PETER M. SENGE

10

1 Frequency
n-c Basic-lnnovati,..,

8
6 (a) Basic-Inventions h 27 1700 i 20 1 ?

lh
I
I

I
40

I
60

I
80

I
1800

I
20

11
40

I
60

Basic-Innovations

Basic-inventions (b)

Electrotechnical

Cc)

Chemotechnical

13 11

Frequency

/I
1 Basic-Inventions .

Basic-Innovations

97-

Cd)
5-

Fig. 18. Frequency of basic inventions and basic innovations corresponding to the three innovation surges in the nineteenth and twentieth centuries [3, pp. 140-1461. a) First half of the nineteenth century, b), c) second half of the nineteenth century, and d) first half of the twentieth century.

A LONG-WAVE

HYPOTHESIS

OF INNOVATION

303

The data in Figure 19 show this sudden drawing down of the invention backlog during long-wave troughs. 24 Each point on the figure indicates the date of a basic invention and the lead time that elapsed before the basic invention became a basic innovation. The data are grouped according to which of the four innovation surges in Figure 16 the respective inventions belonged. For example, the inventions in the group Dl all became innovations in the surge of the 1760s. The curves fitted through each set of data show a shortening of the innovation lead time around long-wave troughs. This phenomenon becomes more evident in each successive long wave. For example, the curve D4 traces through the inventions that become innovations in the wave beginning in the 1930s. The oldest invention exploited during this period was almost 100 years old. The downward slope of the curve shows that as one approaches 1930 and 1940, the innovation lead time shortens dramatically. Several inventions in the 1920s and 1930s found their way into application within about 10 years. Mensch makes several other observations that corroborate the long-wave theory. For example, he describes how strongly the speed with which different basic inventions are applied depends on the compatibility of the invention with the existing technology and capital [3, p. 1691:
There can be striking differences in the innovation time required for a given technology, depending upon whether it develops within or outside of existing industries. These differences clearly refute the popular notion that the knowledge transfer proceeds more quickly today that it did in the past. For example, the basic patents for both the transistor and holography were issued in 1948. Thus, both of these technologies began simultaneously-but it is astounding how different the course of development in the individual fields has been!

Implications of the Long-Wave Hypothesis for Innovation Policy The theory and evidence presented in previous sections suggest that the current slumps in innovation, productivity, and investment arise from powerful forces that have been building up for decades. These forces are: general adequacy of physical capital, overexpansion by physical capital producers, and physical and managerial commitment to the pervasive older technologies. This economic and technological environment is the context within which innovation policies operate. The circumstances generated by the long wave will have considerable influence over the success or failure of a given policy. These policies fall into three main categories: macroeconomic policies, corporate innovation strategies, and opportunities for social innovation.
MACROECONOMIC POLICIES

Throughout the 1950s and 196Os, standards of living rose in many countries, resulting in equal parts from increasing capital investment per worker and from innovations that increased productivity. When such conditions persisted for decades, conventional wisdom came to regard policies that stimulate investment as necessary and constructive. There are a variety of such policies: lower corporate income taxes, low-cost government financing, investment tax credit, and rapid depreciation accounting that usually amounts to a tax reduction. However, the economic climate that inspired the conventional wisdom appears to have changed. In many sectors of the economy, there is evidence of excess physical

Inventions are clustered according to whether they became innovations in the second half of the seventeenth century (Dl), the first half of the nineteenth century (D2), the second half of the nineteenth century (D3), or the first half of the twentieth century (D4).

304

ALAN K. GRAHAM

AND PETER M. SENGE

ol,,,,,,,,,,,,,,,,,,,,,,,,,,
1700 20 40 60 80 1800 20 40 60 80 1900 20 40

Date

of Basic

Invention

Fig. 19. Lead time from basic invention to basic innovation for four successive long waves [3, p. 1681.

capacity. If the economy is near a long-wave peak, more capital investment is not universally necessary. Although increased investment might be desiruble, it does not seem to be universally needed. In such an environment, stimulative policies can have only very limited impact. Indeed, a recent Brookings Institution study found that in recent years the various investment incentives in the United States have little impact on investment activity

WI.
Thus, general economic incentives seem unlikely to elicit significant outpourings of investment. Even if investment can be stimulated, such response seems unlikely to engender basic innovations that can have substantial effect on standard of living. Near the peak of a long wave, innovations are more likely to be incremental improvements rather than fundamental breakthroughs. The world is ready for gasoline-efficient autos; the world is not ready for massive conversion to automotive power sources other than petroleum. Incremental improvement innovations have a vast and ready market and an infrastructure that supports them, and basic innovations do not. Another factor that increases the difficulty of introducing basic innovations at the peak of a long wave is that they often utilize technology related to other basic innovations. Thus, if one basic innovation is delayed for economic reasons, other basic innovations that depend on that technology will also be impeded. Investment stimulus is one commonly suggested area for improving the standard of living through innovation. The other major area might be termed project research: direct and massive research and development efforts aimed at some specific outcome. During the early and middle part of the present long-wave upswing, project research produced atomic power, radar, the high-speed electronic computer and put a man on the moon. But project research does not always produce success. The war on cancer, begun a decade ago, has won several skirmishes but no major battles. One major barrier to producing basic innovations through project research is that the effectiveness of basic innovations often depends on the technologies of other basic innovations. The development of the high-speed electronic con puter would have been virtually impossible without the high-speed test equipment developed as a part of radar research. And radar sprang from much of the same electronic technology that produced television. Similarly, the advances in automotive technology during the 1920s and 1930s quickly created the need for chemical technology that produced a catalytic cracking refinery in

A LONG-WAVE

HYPOTHESIS

OF INNOVATION

305

1935 and nylon in 1938. Seemingly, chemical technology would have been difficult to develop without automotive technology. Mensch has shown that, with few exceptions, the order in which basic inventions occur is exactly the same order in which the corresponding basic innovations later occur [3, pp. 205-2121. This natural order suggests an interdependence among basic innovations that would make project research especially unfruitful during times of capital overexpansion. Although macroeconomic stimulation for innovation is unlikely to succeed, it would appear that there is more hope for individual research and development (R&D) efforts aimed at basic innovation.
STRATEGIES FOR INNOVATION RESEARCH

Studies of R&D management tend to focus on the tactics of R&D-how to fund projects and criteria for stopping projects, how to manage communications among staff, and so on. The long-wave hypothesis of innovation suggests that there are critical strategic questions as well: What will the technologies of the next wave be, and what can we do about them now? Assuming that present economic conditions represent a long-wave peak, the longwave hypothesis implies that many of the basic innovations of the next wave currently exist as working laboratory prototypes. More specifically, the status of the next-wave technologies looks like thisz5: Virtually all of the operating principles (the cause-and-effect relationships) that the next wave of technology will use have already been identified (perception stage). Most of the applications of those principles ceived of (conception stage). Most of the next ensemble of basic innovations in the laboratory (feasible invention stage). Half of the basic innovations to inventions have already been con-

have already worked to some extent

have begun development

(development

stage).

For less than a quarter of the next wave of basic innovations to market (decision stage). Very few of the innovations innovation stage).

there has been a decision

of the next long wave are available commercially

(basic

The extrapolated scenario above might seem to offer many opportunities to enter on the ground floor of the technologies of the future. Yet we know that in the 1930s only a few individuals and companies in fact took such opportunities. The reason is simply that, at the end of a long-wave upswing, innovation is dominated by the marginal, relatively small-scale innovations whose market and payback are close at hand. The basic inventions of the next wave aim at undeveloped markets and offer uncertain payback. They embody technologies that differ from the technologies of the last wave as radically

*5Following the six stages of the innovation

life cycle proposed by Mensch [3, pp. 189-2011.

306

ALAN K. GRAHAM

AND PETER M. SENGE

as airplanes differ from rail travel. Thus, it is not surprising that basic innovations for the next long wave tend to be overshadowed by the more attractive improvement inventions. How might one run an R&D operation to be more likely to develop the basic inventions of the next long wave? The current image guiding R&D management might be expressed as follows: Ideas come randomly from technical people; the managers task is to select those ideas with the best potential for payback and herd those ideas along. This image is not suitable for guiding a firm into the innovations of the next wave for at least two reasons: 1) Unfettered reasearch is most likely to result in improvement innovations and 2) the concept of payback is not appropriate for evaluating potential basic innovations. If most of current technology relates to the present wave rather than to the next, and if one gives researchers free rein to find new ideas, then most of the ideas will relate to the technology of the present wave. This is expecially true if researchers would rather explore familiar territory. An alternative style of management is to choose a specific line of technological development, directed, one hopes, toward the next-wave technology. For example, the electronics industry seems to have a specific line of development planned toward an office of the future fully integrated with communication and computation systems. Compared with projects that do not move a company toward the technology of the next wave, those that do offer both greater long-run profits and greater ability to generate follow-on projects. The other flaw in current R&D thinking is how one measures profit. Payback period is often used, even though financial theory says that the whole concept of payback can be misleading for project evaluation. The payback period indicates how soon a project pays for itself; the payback period indicates absolutely nothing about how much profit there can be ufrer the project pays for itself. A company that aggressively pursues a line of technological development has in effect already judged that long-run potential is more important than short-term payback. Basic innovations might almost be defined as development projects whose returns are slow at first and huge later (as the innovation penetrates the entire economy). In other words, the payback concept alone can systematically discourage R&D on future basic innovations. A long payback period does not mean that an R&D project is financially unsound. If the payback concept is to be used, it should be tempered with some recognition of payoffs after the break-even point; net present value computations produce a summary statistic that can provide such a recognition. Alternatively, year-by-year long-term projections of costs and returns would indicate the long-term implications of a given R&D project.
SOCIAL INNOVATION

The long-wave hypothesis says that we are close to a point of maximum accumulation of physical capital and hence maximum commitment to past technologies. Although opportunities for new technologies exist, a major surge in technological innovation may be a decade in coming. Policies aimed at stimulating technological innovations will probably have low leverage in coping with the long-wave downturn. Where are the areas of higher leverage ? One possibility is social (rather than technological) innovation. Historically, the peaks through the troughs of long waves have been times of considerable social change. In the United States, movements for womens rights have appeared in later stages of successive long-wave peaks, with womens liberation in the 197Os, suffrage (the right to vote) in the 192Os, and so on [5]. The hardships of the Great Depression gave rise to many social experiments, some of which survived and flourished (the New Deal role of government in economic matters), while others did not (national socialism and the Third Reich in Germany).

A LONG-WAVE

HYPOTHESIS

OF INNOVATION

307

The long-wave peak is the time of maximum physical wealth in society. Only a small fraction of the population produces the physical goods on which society is founded. The remaining people (in management, government, services, and so on) are in a sense overhead that could be redirected through social innovation. Organizational design is one area of potentially effective social innovation. The hierarchical structure common to most organizations represses creativity and personal responsibility and may be particularly ill suited to a long-wave downturn. New concepts in organizational design that emphasize local profit centers encourage entrepreneurial development and vitality while increasing overall corporate profits [28]. Moreover, such decentralized organizations may be much better suited to cope with a long-wave downturn than topheavy hierarchical structures. A decentralized organization may more readily cut back unprofitable ventures without financial losses damaging all lines of business. Employees who share responsibility for small profit centers have far greater incentives to stay attuned and adapt to shifting economic conditions, rather than waiting for word from higher up as to how to respond to major economic change. The long-wave hypothesis implies a pressing need for social innovations to cope with the downturn. Companies and institutions expire and people lose jobs that never return. If a long-wave downturn does occur in coming years, current philosophies of government intervention may not only fail to soften crises, but may actually intensify them. For example, consider the current handling of unemployment in the United States. If people lose their jobs, unemployment compensation payments from. the government provide income to sustain them until they regain their jobs. But during a long-wave downturn, jobs in the capital-goods production sector can disappear for years. When jobs reappear, they will use new technologies and require different skills. Unemployment compensation encourages people to wait, as dependents of the government, for capital-sector jobs that will not return. Social innovations are needed to help move workers out of collapsing industries. People will not remain trapped on unemployment compensation if it is made easy enough to move into new sectors and to locate positions that provide on-the-job training. For example, consider modifications to income tax laws that could facilitate movement between sectors. Current U.S. income tax laws provide a small amount of relief from drops in income, by taxing the average of several years of income. Suppose this income averaging feature was considerably strengthened, so that income tax would be minimal or negative for people with a clear record of earnings who experience a sharp drop in income. If this income tax cushion convinces people to get off unemployment compensation and to start new careers outside the capital sector, the subsidy to unemployment has turned into a subsidy to rejoin the productively employed. As another policy that would facilitate movement between sectors, businesses could be relieved of some of the taxes they pay on employees (social security in the United States, for example) if the employees have been on unemployment compensation or had a drop in income. In effect, this would provide a short-term subsidy for firms to train people whose experience is no longer directly relevant. An analysis of these tax programs with regard to costs, benefits, and loopholes is beyond the scope of this paper. Nonetheless, there are clearly opportunities for innovation in the area of labor mobility. Private sector innovations to increase worker mobility are also possible. For example, new firms have begun to appear specializing in the personnel problems of closing down plants in declining industries. One New York firm intensively trains workers about to be laid off to reestablish a sense of personal worth and trainability [29]. One result is greater willingness and ability of workers to seek new jobs requiring new types of skills.

308

ALAN K. GRAHAM

AND PETER M. SENGb

Expenses for this training are assumed by the former employer. In most cases, the training cost is more than offset by 1) reduced unemployment compensation when workers find new jobs more quickly and 2) reduced vandalism and sabotage often associated with plant closings. In some cases, productivity in soon-to-be-closed plants has actually increased when workers feel that their employers are committed to assisting them in finding new employment. Another area needing social innovation is in preparing institutions for failure. The bank failures during the 1930s were sudden and massive. Yet the lessons and the discipline deriving from those failures have been forgotten over the years, as Minsky shows: Individuals, corporations, and governments have built up large debt obligations and have become increasingly vulnerable to adversity [30]. Governments are often called on to make loans to troubled banks (such as Franklin National) or corporations (Lockheed, Chrysler), and even to run railroads. The justification usually boils down to the need for short-term help to achieve long-term health. But in a long-wave downturn, there may not be any long-term health for these organizations. Government bailouts may not serve society when limiting the downside risk encourages more risk taking by those not yet in trouble. Thus, it may be more constructive in the long run not to come to the aid of corporations in trouble. Indeed, after watching corporations approach bankruptcy and systematically neglect needed long-term investments (track maintenance in the Pennsylvania Railroad, engineering in Chrysler Corporation), one might conclude that the bankruptcy laws should make bankruptcy much more easy and rapid to prevent inefficient use of societys resources. Summary This paper has set forth an hypothesis concerning long-wave behavior in innovation. The hypothesis begins with several tenets on the origin of the long wave:
l

Economies move through long waves of approximately from over- and underexpansion of the capital-producing

50 years duration, sector.

arising

The upturn of a long wave, which lasts about 30 years, is characterized by self-reinforcing pressures to acquire more physical capital to meet rising demand for capital, increase capital intensity of production, and take advantage of high returns on investment. Productivity per person increases during the upswing of the long wave, due at least in part to increasing physical capital per person. When sufficient physical capital has been accumulated,, adding capital is no longer more attractive than adding labor. Capital investment peaks and shows signs of declining, and the economy enters the peak period of the long wave, which can last for a decade or so. Capital investment eventually falls off dramatically; the economy needs much less new investment to replace depreciation than it did to expand its capital plant. The capital-producing industries collapse and many of the people in them become unemployed.

A LONG-WAVE

HYPOTHESIS

OF INNOVATION

309

Physical capital begins to deteriorate and obsolesce during the depression. Eventually, there is a need to replace it, and demand for capital rises. Again, the process of capital accumulation is begun in the upswing of the next long wave.

The long-wave hypothesis of innovation suggests that the long wave in accumulation of physical capital profoundly affects the progress of innovation, as follows:
l

During the later upswing and peak of a long wave, the economys physical, technological, and managerial infrastructure is committed to older technologies. There are numerous opportunities for improvement innovations and few economic incentives to turn away from the established technologies. During the downturn of the long wave, very little new investment there is little market for technological innovations. occurs, and

During the late upswing, peak, downturn, and trough, scientific and technical progress continues, even though most of the basic inventions are not commercialized. As a new upswing begins, investors gone untapped for decades. begin to exploit new technologies that have

The cluster of basic innovations near the trough and in the early upswing of a long wave molds the technological character of later investments, and the cycle of basic innovations repeats itself.

The long-wave hypothesis on innovation is supported by a variety of facts. If it continues to be upheld, there are a variety of important implications for corporate and national policy:
l

The current excess of physical capital renders conventional macroeconomic investment stimuli ineffective. Even if investment occurs, economic and technological constraints favor improvement innovations rather than basic innovations. Productivity would not be substantially improved. Most of the technologies that will underlie the next long wave now function in the laboratory, but have not yet been marketed. Highly directed research alone can move organizations the next long wave. Short-term, tactically motivated minor improvements in existing technology. toward the technologies of research tends to produce

Use of payback period as the sole measure of project profit favors short-term commitments and systematically discriminates against basic innovations. Profits after the break-even point are potentially huge, though highly uncertain. Opportunities for social innovation exceed opportunities tion through the peak and downturn of a long wave. for technological innova-

310

ALAN K. GRAHAM

AND PETER M. SENGE

Some

of the possible

social

innovations

in the coming

years

are innovations

in

decentralized organizational designs, changes workers to move out of declining industries, support for failing companies.

in welfare and taxation to enable and changes in attitudes toward

References 1. Bowen, W., Better Prospects for Our Ailing Productivity.

Fortune 100 (I I), 68-86 (1979). 2. Data Resources Inc., cited in Zucker, Seymour, ed., Economic Diary, Bu.ciness Week (20, 25 February 1980). 3. Mensch, Gerhard, Sfalemare in Technology, Ballinger, Cambridge, Mass. 1979, pp. 30-3 I. 4. U.S. National Science Foundation, National Patterns ojR&D Resources-Funds und Manpower in the UnitedStates 1953-76. U.S. Government Printing Office, Washington, D.C., 1976. 5. Kondratieff, Nikolai, The Long Waves in Economic Life, Rev. Econ. Stat. 17, 105-115 (1935); and Schuman, J., and Rosenau, D., The Kondrutpv Wave, World Publishing, New York, 1972. 6. Schumpeter, J. A., Konjunkturzyklen II, Gottingen, Germany (1961). 7. Forrester, J. W., Growth Cycles, De Economist 125, 525-543 (1977). 8. Forrester, J. W., Mass, N. J., and Ryan, C. J., The System Dynamics National Model: Understanding Socio-Economic Behavior and Policy Alternatives, Tech&. Foreca.vt. Sot. Change 9, 51-68 (1976). 9. Mass, N. J., Economic Cycles, MIT Press, Cambridge, Mass, 1975. IO. Mass, N. J., Modeling Cycles in the National Economy, Tech. Rev. 78, 2- 12 (March-April 1976). I I. Senge, P. M., Alternative Views of the Rapidity of Monetary Response, System Dynamics Group Working Paper D-3065, System Dynamics Group, M.I.T.. Cambridge, Mass., 1979. 12. Senge, P. M., The System Dynamics National Model Inwstment Function: A Cornprison to the Neoclassknl fnvestment Function, Ph.D. dissertation, Alfred P. Sloan School of Management, MIT, Canbridge, Mass., 1978. 13. Mitchell, T. W., Competitive Illusion as a Cause of Business Cycles, Quart. J. Econ. 38, 631-652 (1923). 14. Forrester, J.W., Innovation and the Long Wave, Mmwgernent Review, 16-24, June (1979). IS. Van Duijn, J.J.. The Long Wave in Economic Life, De Economist 125, 544-576 (1977). 16. Mandel, Ernest, Lute Cupitulism, NLB Atlantic Highlands, London, 1975, Chapter 4. 17. Kuznets, Simon, Economic Change, Norton, New York, 1953, p, 109. 18. Kendrick. J. W., Productivig Trends if1 the United Stores, Princeton University Press, Princeton, N.J., 1961, p. 328. 19. U.S. Department of Labor and U.S. Department of Commerce, Bureau of Economic Analysis (investment expenditures). 20. Coen, R.M., Investment Behavior, the Measurement of Depreciation. and Tax Policy, Amer. Econ. Ret, 65, 59-74 (1975). 21. Wasson, R. C. et al., Alternative Estimates of Fixed Business Capital in the U.S.. 1925- 1968, Sur\,e.v Current Bus. 50, 18-36 (1970). 22. U.S. Department of Commerce, Bureau of the Census,,Historical Statistics (1975 Ed.), Unemployment Rate, total civilian labor force, Series D-86 (1890- 1970); U.S. Department of Labor, Unemployment Rate, total civilian labor force (1971-1979). Construction Data: Historical Statistics, (1975 Ed.) Total New Construction in 1929 dollars, series N-71 (1869-1955); Bureau of the Census, Total Construction Put In Place in 1967 dollars (converted to 1929 dollars) (1956- 1963), Total Construction Put In Place in 1972 dollars (converted to 1929 dollars) (1964- 1979). Population Data: Historical Statistics, ibid., Total Residential Population, series A7 (1869-1970); Bureau of the Census, Total Residential Population (19711979). 23. Jorgenson, D., Investment Behavior and the Production Function, BellJ. Econ. 3, 220-251 (1972). 24. Energy Resource and Development Administration, A National Plan for Energy Research, Development. and Demonstradon: Creating Energy Choices for the Future, ERDA 76- 1 1976, p. 15. 25. Marchetti, N., Primary Energy Substitution Models: On the Interaction between Energy and Society, Technol. Forecast. Sot. Chanp 10, 345-356 (1977). 26. Schmookler, J.. Invention and Economic Growth, Harvard University Press, Cambridge, Mass., 1966.

A LONG-WAVE

HYPOTHESIS

OF INNOVATION

311

27. Clark, P. K., Investment in the 1970s: Theory, Performance and Prediction, Brookings Papers on Economic Activity 1, 73-124, 1979. 28. Forrester, J. W., A New Corporate Design, Sloan Manage. Rev. 7, 5-17 (1965), also in Collected Papers ofJay W. Forrester, MIT Press, Cambridge, Mass., 1975, pp. 93-109. 29. Career Development Team (CDT), described in Business Week (February 4, 1980). 30. Minsky, H., Financial Resources in a Fragile Financial Environment, Challenge 18, 6- 13 (1975).
Received 7 December 1979, revised 24 March 1980