1. Capital Investment And Best-Practice Labour Productivity
2. Technological Innovation As The Engine Of Growth
3. R&D Budgets And The Growth Of Best-Practice Labour Productivity
4. The Link Between Grer&D Budgets And The Growth Of Blp
5. The Growth Hypothesis
6. Lpi And Long-Term Economic Periods
7. The Determination Of The Efficiency Factor A
8. The Meaning Of The Efficiency Factor
9. Towards Ever Diminishing Economic Growth
Now that we have obtained a reasonable picture of the progress of the labour productivity growth rate in the leading nation over the last centuries, let us try to see what goes into its make-up. More particularly, let us concentrate on the American economy of the last 100 years.
The economy is made up of agriculture, industry and services, including government services, and its average labour productivity has grown year after year because the stock of installed physical capital per worker (i.e. the equipment, the buildings and infrastructure) was modernised repeatedly and the organisation of production followed; and because the human capital supplied by the education system (i.e. the education stored in the brains of those who operate the equipment and organise the work, or the average educational level) grew in parallel with the physical capital.
The dependence on the stock of physical capital, work processes and human capital is implicitly incorporated in the labour productivity index (Lpi). The latter can be considered an indicator of the level of the average state of economic advance generated by the installed technology, working methods and educational levels.
The money for education comes out of taxation (government educational services) and private pockets (the private education system) and, generally speaking, the higher the pcGDP, the more a country invests in education, and the higher the human capital. The growth of the human capital is therefore linked to the rise in the labour productivity index.
The state of organisation and work processes, another important constituent of Lpi, is dependent on the physical capital, whose modernisation led to a continuous reorganisation of work processes. If we want to discover what makes Lpi grow, we must therefore in the first place address the question: “What makes the average stock of installed physical capital grow?”
One of the most important factors leading to the accumulation of physical capital is investment. Capital investment equips a growing labour force with up-to-date means of production. It also overcomes depreciation, replacing worn-out machinery with brand-new equipment whose productivity is higher than the machinery it replaces. Could investment therefore be at the origin of the growth of Lpi?
At the level of the individual company the means of production improve in a discontinuous manner. A company will build a new plant at a given moment and then make no appreciable improvements to the plant’s productivity for several years. At the aggregate level, however, there will be many companies that renew their equipment, and certain sectors that are in the process of changing their working methods in any particular year. At that level one consequently observes a much smoother, more continuous change from lower productivity to higher productivity, especially if one takes five or ten-year averages.
As new production techniques keep developing, only companies that have just built new plants or renewed their equipment will be at the cutting edge of productivity. The ones that did so 5 years ago will be producing at the top productivity level of 5 years ago, those that did so 10 years ago will be producing at the top productivity level of 10 years ago, and so on. The economy is thus at all times a mixture of different productivity levels ranging from the top productivity of just installed cutting-edge equipment, to the low productivity of the oldest technologies still in operation.
Let us now suppose that investment levels were to increase nationwide. Firms in all sectors would then install new equipment more readily, the age of the installed equipment averaged over the whole economy would fall, and productivity would be closer to that obtained in the case where the productive equipment of the whole economy were to be brand new. Next we consider the case where resources for investment become infinite (this is obviously a theoretical case). All private companies and government services would then install the most up-to-date equipment available worldwide the moment it came onto the market. Let us suppose that they would also constantly embrace the more productive working practices made possible by the most advanced technology. They would then in fact be producing at the top-level productivity in their particular sector at all times. In that theoretical case the whole economy would find itself producing at the cutting edge of progress and Lpi would be the highest it could be, given the existing state of technology. Let me call the aggregate labour productivity resulting from this theoretical case the best-practice labour productivity (Blp).
Best-practice labour productivity consequently expresses at any given time the theoretical labour productivity if all production in the USA took place at the top level of technology existing at that time anywhere in the world. It assumes that the substantial investment of resources needed to implement the commercial applications of innovations throughout the economy and to change organizational and working methods accordingly (which in real life takes years to complete) happens instantly. Let us furthermore assume that, in best-practice labour productivity, the level of education of the labour force is adequate for the existing technology, or that the human capital, which must be included in the total capital stock, is adapted to the top technology.
It is possible to estimate best-practice labour productivity Blp from Lpi with reasonable accuracy. An example will illustrate how this can be done. The average age of American production equipment at the aggregate level in 1965 was about 7 years (see Appendix E). Let us now simplify and say that all equipment in 1965 was exactly 7 years old. The equipment that produced Lpi in 1965 was therefore on average the brand new top-level equipment of 1958 that would have produced Blp in that year. In other words, Blp1958 was equal to Lpi1965.
The first step in the calculation of Blp is thus the determination of the average age of existing American plant and equipment during the twentieth century, and this is done in Appendix E. It turns out that the average age of production equipment (for all sectors of the American economy) fell from about 13 years in 1913 to 7 years in 1965 and to maybe as little as 4 years in 2000.
We can now start calculating Blp. Let me give a few examples. The labour productivity index Lpi of 1929 was 75.3. In 1929 the age of equipment and working practices averaged over the whole U.S. economy was about 11 years. We can therefore take Blp of 1918 to be 75.3. In 1965 the labour productivity index Lpi was 165.8 while the average age of equipment and working practices was about 7 years. Blp of 1958 was consequently 165.8. Lpi of 2000 was 287.8 at a time when the average age was about 4 years. Blp of 1996 was thus 287.8.
Higher capital investment would have brought Lpi closer to Blp. In the theoretical case where capital investment could have become infinite, average ages of equipment would have fallen to zero. Lpi would then have equalled Blp all through the 20th century, but would not have surpassed it.
The conclusion one can draw from this is that capital investment, however important its role may be in equipping a growing labour force with increasingly up-to-date means of production, is not the driving force behind Blp. So, if capital investment does not explain why Blp grew from 75.3 in 1918, to 165.8 in 1958 and 287.8 in 1996, we need to look elsewhere for the answer to the question: “What made and continues to make Blp grow?”
Let us now look at the role played by technological innovation. Evidence of the enormous contribution made by technological innovation towards the growth process since the Industrial Revolution is overwhelming. Without the extremely long list of inventions, starting with the steam engine, we would still be living and producing today as our forefathers did in the early eighteenth century.
Imagine for a moment that all inventions and improvements had ended in 1900 (when the American pcGDP was $5,650). We would then be living in a world without airplanes, modern electrical appliances, assembly lines, mass-produced cars, comfortable houses, power tools, radios, TVs, electronics, high precision tools, plastics, telecommunication, satellites, microchips, computers, composite materials, specialised chemicals, antibiotics and all the most common drugs; in fact, without most of what we take for granted. And production methods, working practices and organisation would still be those permitted by the leading technology of 1900.
No capital investment, however high, would have increased the existing best-practice labour productivity. Investment would only have been able to install more capital goods offering the best level of productivity of 1900. By around 1930 even the oldest equipment remaining from before 1900 would have been replaced and the labour productivity index Lpi would finally have equalled the best-practice labour productivity Blp of 1900. Furthermore, as women would have had to continue doing most of the domestic chores due to the absence of washing machines, modern kitchen equipment and the like, the activity rate (as a proportion of total population) would have remained below 0.40. American per capita GDP would therefore not have grown much above $7,000. That it was just over $34,000 in 2000 must be attributed to the technological innovation that took place during the 20th century. Technological innovation should therefore be put at the centre of the growth process.
However, economists point out that factors such as economies of scale, better management or the new organisation of production processes, which are not directly linked to technological innovation, also contribute to raise productivity. An outstanding example of productivity increase in industrial production not related to technological advance was the work of Frederick Winslow Taylor at the beginning of the 20th century.1 By timing each movement of workers performing routine production and analysing its utility, Taylor’s task analysis was the origin of industrial engineering and led to the rationalisation of repetitive work and to significant gains in productivity.
Robert Solow also draws attention to the fact that many instances of product improvement and cost reduction originate from the accumulation of small suggestions coming from production workers, process engineers and even customers.2 Others cite such factors as workers’ efficiency, and point to the great dedication of the Japanese workforce as making an important contribution to Japanese economic growth.
Let us first discuss the relative importance of some of the factors suggested above. ‘Shop floor’ improvements do indeed play a part in productivity gains. But we should see things in their proper perspective. It is certainly possible to claim that some of the main contributions to productivity gains in the 18th and 19th centuries were the result of shop floor improvements. But this was no longer the case after the late 19th century. The major technological innovations of the 20th century, which have been the true starting point for the creation of more productive processes, came out of Research and Development (R&D). True enough, once productive equipment is installed, adaptation to local circumstances and the personal initiative of workers and shop floor teams who know their equipment all add to productivity. But it should be clear that such initiatives lead, on average, to no more than small additional productivity increases in comparison with the quantum leap in productivity produced by the installation of new, much more efficient equipment.
The same holds for human dedication, which obviously improves the chances of the most efficient use being made of existing technology. But dedication on its own did not invent the computer-driven robots and quality control equipment that have made Japanese products and production methods so outstanding.
It is also important to realise that factors such as economies of scale, better management and new organisation of the production process were made possible by prior technological advance. The assembly line, for example, could not have been created had there been no electricity generation and the technology to make parts with a high degree of standardised precision. Had airplanes, container ships and telecommunication not transformed the world into a village, we would never have experienced the phenomenal growth of multinational firms in the second half of the 20th century. In the same way important improvements in the organisation of production today (such as just-in-time delivery) have been made possible by advances in computer technology and telecommunication.
Furthermore, better organisation of production processes, which is capable of raising productivity within the existing technological framework, sooner or later is made obsolete by the introduction of new technology. The efficient manual worker produced by Taylor’s measurements and imitated by Chaplin in ‘Modern Times’ has long been superseded by robots. They perform all repetitive work on assembly lines much more efficiently and with fewer faults than quasi-robot manual workers were ever able to.
One must therefore consider technological innovation created by Research and Development (R&D) as the primary cause of the growth of best-practice labour productivity (Blp). Other factors such as economies of scale, better management and organisation, or shop floor improvements and worker dedication, are secondary causes that allow the optimal use of new technology but depend on previous technological innovation. In what follows I shall concentrate on the primary cause. The way secondary causes are taken into account will be referred to later.
The search for new technologies and their development for application in the economy leads to technological innovation. It is not always easy to follow the path of technological innovation created by a single invention. The diffusion of the effects of an invention throughout the economy is not only a slow, but also an uncertain and often erratic process. Furthermore, the link between Research and Development spending and the development of new technologies is not always obvious. High spending in one firm may lead nowhere while another firm, spending much less, may strike lucky and stumble upon an innovation. Some inventions produce few applications in the sector from which they originated, but cause important improvements in other sectors. At the level of the individual company, the link between R&D spending and technological innovation is therefore not regular or even easy to establish. But at the aggregate level these micro-effects disappear. For a large economy such as that of the United States, and averaged over several years in order to smooth irregularities even further, one may expect to find a much clearer relationship between R&D spending and technological innovation.
Best-practice labour productivity (Blp) is the theoretical labour productivity we would obtain if the whole American economy used cutting edge technology. Blp grows because this cutting edge keeps advancing year after year as a consequence of R&D and the resulting technological innovation. The growth of Blp must therefore be related to R&D. But what exactly is the relationship?
Could total labour employed in research laboratories be the factor causing Blp to grow? Inventing something remains a ‘eureka’ event. Examples are numerous of scientists or engineers who had all the relevant data in front of their eyes, yet failed to make the connection. Discoveries often contain an element of inspiration, of sudden intuition and where nine scientists will sift through the data without noticing anything particular or perform tests that are not relevant, the tenth will suddenly see the light or make the right test. Inventions therefore become more probable, the more numerous the scientists and engineers actively involved in research.
However, scientists and engineers need more than just a pencil and paper if they are to be successful in research. They will need sophisticated equipment to perform experiments: machines that can produce reactions and computers to control these machines; other apparatus to measure the reactions; and yet more computers to store and analyse the data. Nowadays, it is difficult enough to produce anything new, even with the support of large research budgets. Without a decent budget, the research scientist or engineer might just as well look for another job. And developing and perfecting practical applications (the D part of R&D) is just as costly. It thus seems logical to conclude that the relevant quantity is not so much the number of scientists and engineers working in R&D, as their number multiplied by the average budget they have at their disposal or, in other words, the R&D budget.
But what kind of R&D budget should we consider? Should we count only R&D privately financed by business and industry? Or should we also include government-funded R&D? And, finally, do we also need to count foreign R&D spending as contributing to the growth of American Blp?
Many economists tend to consider only private R&D budgets as contributing to the growth of productivity. This view appears to me too limited. Government-funded R&D does contribute to technological advance. Government-funded research units in universities or national or international laboratories generally carry out basic research, and usually do not produce direct practical applications. It is nevertheless true that the discovery of new practical applications in solid state physics, chemistry and so on, have become impossible today without previous fundamental advances in physics, chemistry or biochemistry. Basic research should therefore be regarded as the cornerstone upon which applied research rests, or as the underlying basis that makes productivity improvement possible. It is a gift from which industry is able to benefit without having to pay for it. So it seems correct to count not only private, but also government-funded R&D budgets.
However, it would be difficult to argue that government-funded defence and space-related R&D should be treated like the rest. There is no direct gain in economic productivity from perfecting nuclear weapons or from putting a man on the moon. A large part of military and space-related R&D is consequently a waste from the point of view of general economic progress. It is, on the other hand, undeniable that certain aspects of military innovations have, with some delay, found their way into commercial applications. Examples are many: miniaturisation, new materials, guidance systems, and the like. Space exploration too has permitted a host of practical applications: communication satellites, GPS systems, and many others.
It seems therefore appropriate to reduce defence and space-related government-funded R&D budgets to the part that leads, with some delay, to commercial applications contributing to the increase in U.S. best-practice labour productivity. As there is no scientific method that would allow us to measure this part, I suggest to fix it arbitrarily at fifty per cent and to estimate later the degree of imprecision this might introduce (see Appendix D).
In counting R&D budgets that contribute to the growth of Blp I consequently propose to count not only private R&D budgets but also government-funded R&D in which the defence and space-related part is reduced to half, and call the sum the ‘Relevant R&D’ budget (ReR&D).
It also seems obvious that we should not limit ourselves to U.S. ReR&D. Although the latter constituted over half of the world’s total ReR&D budgets during most of the 20th century (see Appendix F), foreign inventions and industrial processes have always, sooner or later, found their way into the American economy. American firms have been particularly skilful at the commercial exploitation, not only of American inventions, but also of foreign ones, especially after World War Two when the increased trade contacts between North America, Western Europe and Japan led to the rapid diffusion of new technologies wherever they were developed. This implies that foreign Relevant R&D budgets also contributed to the growth of American labour productivity.
So far the OECD countries (basically the most advanced nations) between them accounted for nearly all the technological advance. In practice we may therefore limit ourselves to the sum of the ReR&D budgets of the OECD countries, and I will call this sum the Global Relevant R&D budget (GReR&D).
Global Relevant R&D Budgets (in billion U.S. dollars), 1900-2000, logarithmic scale
The first step is thus the determination of GReR&D budgets. The problem is that statistics on R&D are of comparatively recent origin. The first effort to survey R&D funds (in the USA) dates from 1954 and the sampling of data in other industrial countries began even later. From 1965 onwards, complete data on R&D budgets can be found in published statistics, but estimating GReR&D before that time is extremely difficult. Problems begin in earnest before 1950. As we move back to the first half of the 20th century, we find that the importance of large-scale industrial research diminishes. Around 1900, private individual inventors were responsible for almost all the technological progress that was made. In order to obtain a complete long-term picture of R&D budgets we therefore have to find a way of counting the contribution of private individual inventors before 1950 as an R&D budget. An attempt to do this and estimate GReR&D budgets from 2000 back to 1913 (and even to 1870) has been made in Appendix D, but the reader should keep in mind that pre-1950 GReR&D budgets are not real budgets but have been obtained by estimating the money equivalent of the R&D efforts of private individual inventors.
GReR&D budgets, thus obtained, rose tremendously over the past 130 years, from a possible value of $0.45 billion in 1870 (in constant 2000 dollars) to $4.7 billion in 1913, $10 billion in 1929, $125 billion in 1965, and $560 billion in 2000. Figure 6.1 shows the growth of GReR&D (on a logarithmic scale because this growth was spectacular).
Apart from some slight fluctuations the GReR&D budget seems to have risen steadily all through the 20th century. I have drawn a smoothly increasing average curve through the data points, in order to guide the eye. It is difficult to say anything with great certainty about the period before 1950, especially since the GReR&D budgets of that period are largely symbolic. After 1913 there appears to have been a downswing due to the First World War and its after-effects and there certainly must have been a downswing during the Great Depression. This appears to be followed by an upswing starting in the 1960s (or earlier) and another slight downswing around 1980. These fluctuations seem to be related to the long-term economic periods discussed in Chapter 4. They are present in a number of economic variables and mask the true very long-term trend. In order to discover the forces that shaped this long-term trend I will therefore concentrate on the average curve.
The previous section suggests that the more the world’s nations spend on Relevant R&D, the higher the chances of inventions being made, and the higher the probability of developing technological innovations from these that add to the existing stock of physical capital in the USA, and to the American best-practice labour productivity Blp. This leads me to propose the simple hypothesis that the higher GReR&D budgets are, the larger the annual addition to Blp will be.
From what I have explained so far it seems acceptable that most of the increase in Blp is indeed a consequence of GReR&D. However, to complete the picture, productivity increases due to economies of scale, better organisation, shop floor improvements and other sources outside R&D should also be taken into account. These are secondary causes of productivity increase and, averaged over longer periods, their contribution to growth can be expected to be much less important than that of technological innovation, as mentioned before.
A way of including the contribution of better management, shop floor improvements and similar lesser causes of Blp growth would be to attribute a fictitious R&D to all these non-R&D activities (it should be noted that a small part of R&D budgets is in fact already devoted to research into better management techniques, communication, social studies and market research). This fictitious R&D can be expected to be much smaller than the real R&D. If the ratio of gains from fictitious R&D to those from real R&D does not change significantly over time, neglecting fictitious R&D sources of growth will not introduce any great errors in testing the proportionality I proposed above, and I will assume that this is the case.
Before exploring the quantitative consequences of the proposed hypothesis, we shall have to introduce a delay time between the allocation of GReR&D budgets and the growth these cause in Blp. It is obvious that the technological progress of 1960, for example, was not created by the GReR&D budget of 1960 but by that of some years earlier. Between the time research budgets are first allocated and the moment a commercial application results from these budgets, several years elapse. Appendix E gives an example of this in the USA in the 1950s and shows that we can distinguish between three kinds of delay times:
- The time elapsing between the first allocation of civil R&D budgets and first commercial application in the USA was about seven to eight years.
- Foreign inventions faced an extra 1-5 years delay.
- The time elapsing between the allocation of military R&D budgets and first civil commercial application was much longer due to military secrecy.
Such times are representative of the situation in research and development and, given the mix of American civil and military R&D, and foreign R&D in the GReR&D, the weighted average time elapsing between the first allocation of research budgets and commercial application in the USA was about ten years around the middle of the 20th century. This means that the research projects that created the first commercial application of technical innovations of, say, 1960 started on average around 1950. Part of the GReR&D continued to be spent on these projects all through the 1950s and the projects were, on average, completed and led to commercial applications in 1960. The new commercial applications of 1960 thus relied on part of the GReR&D budgets of all the years between 1950 and 1960. To simplify matters, let us say that it was, on average, the GReR&D budget of 1955 that caused the growth of Blp in 1960.
The delay time between GReR&D budgets and the increase in Blp caused by these budgets can therefore be taken to have been about 5 years in the 1950s. Delay times were somewhat longer at the beginning of the 20th century and diminished towards 2000, as less time was needed to develop information technology, which makes up an ever larger part of new commercial applications.
Summarising the discussion, I propose that GReR&D budgets produced the growth in Blp after a delay of some years. This delay time was possibly 6 years at the beginning of the 20th century, diminishing to 5 years around 1950 and to maybe as little as 3 years by 2000. These are obviously approximations, which will introduce errors in further calculations.
Let us now verify whether the proposed causal link between GReR&D and the growth of Blp is valid for the period 1950-2000. I suggested the hypothesis that the higher the GReR&D budgets, the higher we can expect the annual addition to the best-practice labour productivity Blp to be (keeping in mind that there is a delay time between both) or, in other words:
the annual addition to Blp is proportional to the GReR&D budget of a number of years earlier
Moving from Blp to the more accessible labour productivity index Lpi would greatly simplify matters. Earlier in this chapter we saw how the two were linked. The Blp of 1958, for instance, was the Lpi of 1965. Moving from one to the other therefore introduces an additional delay time, which diminished from about 8 years in 1950 to 4 years in 2000.
We can now move from GReR&D to Blp, and then on to Lpi. An example will clarify this. The Lpi of 2000, for instance, was equal to the Blp of 4 years earlier, which was in turn produced by the GReR&D of another 3 years before that. Eliminating the intermediate Blp stage, it turns out that the Lpi of 2000 was produced by the GReR&D of 7 years earlier, or that the total delay time between GReR&D and Lpi was 7 years for 2000. In the same way we can calculate total delay times for other dates. We find that the Lpi of 1950, for example, was produced by the GReR&D of 13 years earlier.
We can now rewrite the above hypothesis, which I intend to test between 1950 and 2000, as:
the annual addition to Lpi is proportional to the GReR&D budget of some years earlier
If we divide the annual addition to Lpi by Lpi we obtain the labour productivity growth rate Lpgr. The above hypothesis, to be tested between 1950 and 2000, then becomes:
Lpgr is proportional to the GReR&D budget of some years earlier divided by Lpi
This proportionality implies that there exists a factor A that links Lpgr to GReR&D divided by Lpi. Lpgr would then be equal to this factor A multiplied by GReR&D and divided by Lpi.
A quick calculation shows that this factor A was about 11 times smaller in 2000 than it was in 1950. One might therefore call A the efficiency factor because it expresses the efficiency with which GReR&D budgets divided by Lpi translate themselves into the Lpgr of some years later.
Figure 6.1 gave us one of the ingredients that goes into the build-up of the factor A: the GReR&D budget, which does not follow a smooth average but fluctuates slightly over the course of the long-term economic periods. Figure 5.1 gives us the other ingredient, the labour productivity index Lpi (from which Lpgr can be derived). However, as this figure lacks detail because of the very long timescale, I have reproduced the progress of Lpi since 1900 in Figure 6.2 (also on a logarithmic scale).
In Chapter 5 we divided the period 1870-2000 into six sub-periods, chosen in such a way as to minimize the effects of business cycles, and in particular, of two occurrences of non-economic origin: the world wars and their aftermaths. The end-years of those six sub-periods were 1913, 1929, 1950, 1973, 1987 and 2000, and the Lpi values corresponding to these years are indicated in Figure 6.2. Drawing a (broken) line through these points, it is obvious that there is a significant up and down fluctuation between 1950 and 2000.
Drawing the average trend as a smooth full line through these fluctuations makes the picture much clearer. The Lpi value for 1973 lies above the average trend, and the Lpi value for 1987 below it. Although there is no data point between 1930 and 1950, one may assume that Lpi was below average at that time due to the Great Depression. I have not attempted to extrapolate the broken line before 1930, because Lpgr and the derived Lpi values of that period are not accurate enough to allow any detailed analysis.
The undershoot and overshoot movements of the real Lpi are clearly related to the long-term economic periods, with downswings occurring during the hyperbolic phases and upswings during the exponential phases of each period. After the Great Depression and Second World War, Lpi shot above the average trend around 1950, to come down again in the 1970s and plunge below trend around 1980.
In the present long-term economic period, the labour productivity index picked up again in the 1990s, to return to the long-term average by 2000. We may suppose that the real Lpi will remain above the average trend for the next few decades, and join up with the average trend again by the late 2020s, at the end of the economic period.
American Labour Productivity as an Index, 1900-2000, logarithmic scale
In the framework of the long-term economic periods, the spectacular rise of the labour productivity growth rate after 1990 can therefore be interpreted as a temporary catching-up effect. In that case the high labour productivity growth rates of the last decade, of the order of 2 percent, are not going to last. When the real Lpi begins to curve down to return to the average curve, the corresponding growth rates can be expected to slow down significantly, and will be lower than the growth rates of the average curve from 2015 onwards.
In spite of the consistency of such an explanation, some readers might prefer to ignore the average Lpi curve in Figure 6.2, or even reject the effect of the long-term economic periods on labour productivity altogether. We would then remain with a strange picture of the labour productivity index Lpi: it rose more or les at a fixed rate (2 percent) until 1973; its growth rate then slowed down radically until the early 1990s, after which it suddenly returned to its pre-1973 value.
Proponents of such a view might advance that the slowdown of the 1970s-80s was due to the difficult transition from industrial to service society. Indeed, the American economy went through three different stages over the last century. Around 1900 it was based on heavy industry, coal and steel; the economy of the 1950s was dominated by the industrial production of consumer goods; and the post-1990 economy is a service economy, based on computers. Critics may therefore argue that the characteristics of the periods 1900-30 and 1950-73 are totally irrelevant to the economy of 1990-2010 and, consequently, that there cannot be a single formula describing the growth process for three totally different types of economies.
Yet, the passage from one long-term economic period to the next has always been characterised by difficult transitions. The end of the first economic period saw the transition from agriculture to textiles and, later, to coal and steel; the end of the second period, the transition from heavy industry to light consumer industry after 1930; and the end of the third period the transition from light industry to services after 1980. Such transitions were not rapid. The passage from one period to the next always happened through the gradual spreading of new technologies, while old production processes slowly disappeared (this will be illustrated in greater detail in Chapter 8). Each of these transitions needed an economic reorganization and new investment patterns, and such reorganisations took time.
While an economy is undergoing this long-term restructuring, labour productivity growth will be slower, and the economy will forge ahead again under full steam only when the restructuring is over. Economic periods therefore induce a slowdown when the transition takes place followed by an acceleration of growth when new production patterns become widespread, and then again a slowdown as they mature. The swings, which we observe in the long-term picture of Lpi, are therefore clearly linked to the long-term economic periods.
Taking out these decelerations and accelerations, which have to do with capital investment (see Appendix E where I show the effect of working with the real Lpi curve) we notice a smooth decrease of the importance of the primary sector from the 18th to the 20th century; and a smooth rise of the industrial sector during the 19th and 20th centuries to reach a maximum in the mid-20th century, followed by a gradual fall as the service sector began to dominate the economy. These changes, from the 18th century to 2000 were brought about by inventions whose rate of occurrence changed smoothly too (see Figure 7.2). This shows that the growth process during the whole of the 20th century (and before) can be considered a continuous development, and that it does make sense to search for the possible existence of a growth formula representing this process between 1950 and 2000.
The fluctuations caused by the long-term economic periods thus introduce unnecessary complications. If we take these fluctuations out of the picture by concentrating on the average (full) Lpi curve in Figure 6.2, we are left with a story of continuity: an average economy in which the implementation of a new technological base happened smoothly and continuously, and in which it will be easier to search for a mathematical form of the efficiency factor A. In trying to pin down this form I propose therefore to work with the average Lpi curve. Average Lpgr values can be derived from this average curve (see Appendix E).
Let us recall the proposed relationship for the U.S. economy between 1950 and 2000:
Lpgr equals GReR&D multiplied by A and divided by Lpi
We can now isolate the factor A and calculate its value for different years from its ingredients Lpi, Lpgr and GReR&D, and next start our search for a mathematical expression of this factor. As this is somewhat complicated, I have removed all calculations to Appendix E, and will only present the result here. It turns out that there exists indeed a simple mathematical expression for the factor A: it behaves as an inverse power of Lpi, and the mathematical formula we tried to discover turns out to be:
Lpgr equals GReR&D divided by (a power of Lpi) and divided by Lpi
As the labour efficiency Lpi grows through the years, the factor A divided by Lpi shrinks rapidly, and GReR&D budgets will have to rise explosively to compensate for this and produce decent further Lpi growth rates. The efficiency factor A therefore has a strongly limiting effect on the capacity of research budgets to create further labour productivity growth. In other words, the higher the labour productivity index already reached, the more difficult it becomes to convert GReR&D budgets into further growth of Lpi and to continue creating high labour productivity growth rates.
Let us first of all see if this agrees with the history of R&D and inventions. If we look at the way research was done a few centuries ago we cannot fail to notice the enormous differences between those times and today. Nor can we ignore the fact that in the past even the simplest inventions or improvements brought large returns in terms of productivity increase whereas at present returns are comparatively small.
As late as the 18th century, and even the beginning of the 19th, important discoveries were well within the reach of any person having some basic education, technical skill and practical experience. Many of the discoveries that gave rise to the Industrial Revolution were what one might call ‘practical’ inventions. They were the result of the effort of individuals, often technicians, who had sufficient scientific training and good mechanical knowledge.
Around the middle of the 19th century ‘research laboratories’ were usually no more than a backroom in which one single person experimented during his free time. If one looks at the (often home-made) equipment used by inventors at that time, one can only marvel at its crude simplicity. As we move towards the end of the 19th century the situation begins to change. The knowledge required to make important new contributions to the existing stock of technology had moved up a notch and most of those who contributed to technological progress at that time were no longer technicians but engineers and scientists. Also, some successful inventors such as Edison founded companies; these, and other companies, increasingly began to contribute to research. From the beginning of the 20th century onwards, a growing proportion of inventions were born, not in the workshop but in the research laboratory.
At the beginning of the 20th century individuals continued to account for a significant proportion of all inventions, but individual contributions diminished rapidly as a share of the total towards 1940 and especially after World War Two (see Appendix D). Today individual contributions have ceased to add to productivity to all intents and purposes. The Second World War was in fact a watershed in the history of Research and Development. Not only did industry pump ever-larger amounts of money into R&D after 1945 but governments, too, actively began to fund research programmes. By the late 20th century research had become a concentrated, worldwide effort. Huge amounts of money are now poured into R&D every year, and even the most modest research laboratories use highly sophisticated and extremely costly equipment.
Yet, in spite of the spectacular rise in the number of people in R&D, their much higher level of scientific education, and escalating research budgets, the labour productivity growth rate did not keep pace. The complete formula for the efficiency factor (see Appendix F) shows that in the 18th and beginning of the 19th centuries, the labour productivity growth rate was virtually proportional to the GReR&D budget. Consequently, as more people did research through the years and the budgets they invested in those activities grew, the labour productivity growth rate Lpgr picked up. Had returns from investment in R&D remained constant, the labour productivity growth rate would have continued to accelerate and would have attained explosive proportions in the 20th century. However, the efficiency with which GReR&D budgets could be converted into labour productivity growth began to diminish slowly from the mid-19th century onwards and more rapidly during the 20th century. The enormous increase in GReR&D budgets in the last half-century has been insufficient to compensate for the increasingly rapid fall in the conversion efficiency, expressed by the efficiency factor A.
The brief history of research and technological development given above clearly reveals the pervasive influence of the diminishing conversion efficiency. Relatively few people, possibly some 20,000 in Western Europe and the USA, often no more than gifted amateurs with limited resources at their disposal and using elementary equipment, produced a level of innovations sufficient to make labour productivity grow at an annual rate of 2 percent around 1870. In contrast, no fewer than 2 million scientists and engineers are actively engaged in R&D at present in the OECD countries, and the Global Relevant R&D budget is probably more than 1,000 times higher (in constant dollars) than it was in 1870. Yet, the average labour productivity growth rate over the last few decades has been lower than 2 percent. Without the efficiency factor A, which expresses the rapidly diminishing efficiency of converting GReR&D budgets into Lpgr, there is no way of accounting for this.
The efficiency factor needed to be incorporated in the growth formula in order to account for the disappointing performance of Research and Development. It is the mathematical expression of the fact that the explosive growth in research budgets did not produce anything better than a more or less constant labour productivity growth rate since 1870.
We have now gauged the effect of the efficiency factor, but we have yet to understand its origin. It is interesting to note that, while the GReR&D budget is a factor endogenous to the economy, no parameter pertaining to the economy intervenes in the make-up of the efficiency factor. This indicates that it is exogenous to the economy, in other words that it originates outside the economy, and that its explanation will have to be sought in fields unrelated to economics. So, what causes this factor to exist and why is it exogenous to the economy? To find the answers to this question we need to make a brief excursion beyond the realm of economics.
Although scientific research and economics are two quite different fields and although the history of science may seem out of place in a book about the world economy, it is nevertheless scientific advance that is the foundation of all technological change. Consequently, we will have a look at the history of scientific progress and find out whether it can explain the origin of the decreasing efficiency of converting research budgets into labour productivity growth. This will be done in the next chapter.
At the end of the previous chapter I drew attention to the fact that Figure 5.2 suggests that the labour productivity growth rate Lpgr peaks around 1920-30 and seems to have declined afterwards. In spite of this, some readers might prefer to interpret the rise in growth rates after 1995 as a real rise in the growth potential, hoping that we are back to normal and that Lpgr will remain at around 2 percent for several decades. Such an interpretation would, however, be based on ‘being-used-to’ arguments or on wishful thinking rather than on scientific evidence. It flies in the face of scientific arguments, advanced in Chapter 5, that there cannot be a ‘normal’ growth rate. More importantly, it would also fail to recognise the importance of the factor A, which represents the efficiency with which GReR&D budgets can be converted into Lpgr.
The rapid increase in GReR&D has so far obscured the fact that it needs ever-higher GReR&D budgets to keep Lpgr at a decent level. But will the GReR&D budget continue to rise rapidly during the 21st century? In order to estimate future GReR&D budgets, we need forecasts of the future Gross Global Product, and an estimation of how much of their GDPs the world’s nations may be willing to invest in ReR&D.
In the year 2000 the most advanced nations spent 2.2 percent of their combined GDP on Relevant R&D. This percentage has stagnated during the last decades, indicating that saturation has been reached, and that the most advanced nations are unwilling to commit a higher proportion of their GDP to R&D. This percentage can be expected to fall somewhat in the 21st century (see further). According to the forecasts of Appendix H the GGP of 2100 may be just over 300 trillion dollars, and if at that time all the world’s nations spent somewhat less than 2 percent of their GDPs on Relevant R&D on average, the GReR&D budget of 2100 would be in the 5 to 6 trillion dollars range.
Letting the GReR&D budget rise smoothly from just over half a trillion dollars in 2000 to 5 or 6 trillion dollars in 2100, and calculating the resulting Lpgr and Lpi values using the mathematical form of the efficiency factor A, leads Lpgr to fall smoothly from 1.5 in 2000 to just above 0.5 in 2100, with an average over the 21st century of 0.8.
The problem with such forecasts, of course, is that we have to obtain a reasonable idea of future GReR&D budgets, which in turn depend on our guesses of future GGP values. One may argue, for instance, that the GGP estimates of Appendix H are based on Lpgr rates in the advanced nations, falling to about 0.5 in 2100. But what if we assume that the Lpgr stayed at 2 percent in the advanced nations during the whole of the 21st century? This would lead to higher GGP values and higher GReR&D budgets all through the 21st century.
Higher GReR&D budgets will obviously lead to higher Lpgr values, but even if the GReR&D budget of 2100 were as high as 10 to 12 trillion dollars (a very extreme assumption), this would not in any way lead to a doubling of Lpgr. Indeed, plugging higher GReR&D values into the calculations would produce higher initial Lpgr values, higher Lpi values, and a factor A that would decrease much more rapidly during the 21st century, which would then tend to quench any further rapid rise in Lpgr.
It is on the other hand entirely possible that GGP values will never reach 300 trillion dollars in 2100, because limitations imposed by the environment and the climate might well restrict global economic production. We may furthermore expect a large fraction of future research budgets to be deviated into the search for alternative energy methods and the abatement of environmental destruction and pollution, caused by global warming, polluting mining activities, high-yield agriculture and urban sprawl. While these will be worthwhile, even imperative goals, they are not relevant to the raising of labour productivity. Lower GGP values, and lower proportions of the R&D budgets going into production-raising research, would result in GReR&D budgets much lower than the ones on which we based our future assumptions of Lpgr.
While the future of the GReR&D budgets is quite uncertain, the outcome for the Lpgr of the year 2100 is relatively independent of the different scenarios. Extreme assumptions would lead to Lpgr values in 2100 maybe as high as 0.8 percent or as low as 0.3 percent, but forecast for different reasonable assumptions of future GReR&D trends turn out to fall within the 0.5-0.6 percent range.
Some readers may find this forecast unacceptable and reject the proposition that the discovered mathematical formulation of the factor A can represent future growth. There is of course no absolute guarantee that a mathematical form, which correctly expresses the efficiency factor A between 1950 and 2000, will continue to represent this factor after 2000, but a mathematical expression that accurately reproduces past average labour productivity growth rates can be expected to give at least a reasonable picture of the future. And there is no valid scientific reason to believe that the rapid shrinking of the efficiency factor A, which is caused by exogenous forces and has been going on for well over a century, is going to stop suddenly today.
To show that we are heading for a future of much lower labour productivity growth rates, let us look at the relationship between GReR&D and Lpgr from a different angle. Figure 6.1 shows that GReR&D budgets are rising ever more slowly. Although the GReR&D budget of 1870 is very speculative, Appendix D indicates that GReR&D budgets rose by 6 percent annually between 1870 and 1970 and by 4.5 percent on average annually between 1970 and 2000. According to the average scenario of Appendix E, which gives a GReR&D budget between 5 and 6 trillion dollars in 2100, this budget would grow by only 2.3 percent on average annually during the 21st century.
The Labour Productivity Growth Rate, 1000-2200
High growth rates of GReR&D budgets produce much less spectacular labour productivity growth rates. Lpgr was no more than 2.1 percent on average between 1870 and 1970, and only 1.6 percent between 1970 and 2000. The annual growth rate of Lpi turns out to be roughly 2.8 times lower than the annual growth rate of the GReR&D budgets for both periods 1870-1970 and 1970-2000. If the same holds for 2000-2100, we can expect Lpgr averaged over the whole 21st century to be 2.3 percent divided by 2.8, or about 0.8 percent. This is exactly the result forecast by the more rigorous mathematical extrapolations of Appendix E. above.
Everything therefore points at falling average Lpi growth rates. This is shown in Figure 6.3, which summarizes the situation. The block diagram of Figure 5.2 has been continued up to 2030, and the values for 2010 and 2020 are based on the overshoot phase of the real Lpi curve after 2000 (see Appendix E). I have furthermore superimposed the average Lpgr curve on the block diagram. As mentioned above, Lpgr values forecast for 2100 fall within a 0.5-0.6 percent range for different reasonable future levels of GReR&D.
Figure 6.3 clearly answers the question we asked at the end of the previous chapter: “What will happen to the labour productivity growth rate?” It suggests that low labour productivity growth rates are due to arrive, not within a century, but within decades, and that measures to cope with such a new situation will have to be implemented, not within a generation or two, but within the medium-term.
As long as mankind invests in R&D, the labour productivity index will continue to grow, even in 500 years from now. Its growth rate will never become zero unless GReR&D budgets themselves become zero. It is clear, though, that the higher Lpi becomes, the more difficult it will be to convert GReR&D budgets into the further growth of labour productivity. While the latter will keep rising after 2100, it will therefore do so at ever-lower rates. In order to show this, I have carried the extrapolation of the average Lpgr curve in Figure 6.3 as far as 2200. Although moving two centuries forward is necessarily very speculative, the 0.2 percent value at least gives us a rough idea of the possible Lpgr level at that time.
After the year 2200 even extremely large global research budgets would not be able to produce labour productivity growth rates much higher than 0.2 percent. It is on the other hand quite possible that future society will no longer consider it worthwhile pouring a significant part of total economic production (some 2 percent) into R&D activities relevant to the further growth of labour productivity when this will yield such miserable returns. Being unable to produce spectacular results, scientists will lose even more of their standing than is the case today. This might well create a vicious circle: very few young people will then want to study science, less research will be done and it is conceivable that GReR&D budgets will not keep rising but instead fall to low values. Furthermore, planetary problems will probably become so acute by that time that there will be many competing demands for scarce research money. Although even low GReR&D budgets will still contribute to make labour productivity grow, such growth would then be infinitesimal.
Change has been so rapid all through the 19th and 20th centuries that economic practices are now centred on growth and constant adaptation. Yet, as I explained, the fact that change is the norm today and that we are used to it does not constitute proof that this will always remain so. Indeed, within two centuries the total technological stock might well stop growing, and even in the best of cases Lpgr would only be about 0.2 percent. This implies that changes in technology and working methods will become so slow after 2200 as to be almost imperceptible over a lifetime. Working life will then revert to fixed working methods and societal values can in turn be expected to stabilise. Such a situation would have more in common with the stability of the late Middle Ages (when Lpgr was also 0.2 percent) than with the present world.
We should expect the transition from fast growth to almost no growth, to transform present-day society radically. Indeed, the transition from very low growth before 1500 to fast growth in the 20th century left nothing intact of the society of the Middle Ages. Furthermore, living and working conditions in 1300-1500 did not give people any clue as to what life would be like in 2000. In the same way, an extrapolation of current social and working conditions with just a few cosmetic alterations will not give us an accurate picture of what life might be like after 2200, when the era of fast change will have come to an end.
It should also be obvious that efforts spent in the defence of many of the present economic and organisational practices, or of today’s social structures, will be a waste of time and energy, just as efforts to defend the economic structure and social order of the Middle Ages, and the interests of the nobility were doomed. Present-day beliefs and practices will be brushed aside by new ways of behaviour and thinking.
It is certain that future society will be very different from the one we know, and some of its characteristics may even turn out to be totally unexpected. The analysis of the future growth of labour productivity tells us that stability will be the basis of work in 2200, and this could give us some clues as to what the economic characteristics of society might be like two centuries from now. Trying to formulate the economic, and especially the social, legal and political characteristics that will prevail at that time will nevertheless require a lot of original, unprejudiced thinking, and would certainly be one of the most interesting projects that could possibly be undertaken.