Thesis #15: We have passed the point of diminishing returns
In the previous thesis, we saw that complexity is subject to diminishing returns, because of each of its facets–subsistence, information processing, sociopolitical control, economics, and technology–are not only intertwined as a single system, but are themselves subject to diminishing returns. As such, any society which pursues complexity as an answer to every stress–which is to say, any civilization (see thesis #13)–must, eventually, collapse. This is only underlined by the basic fact that nothing can grow forever in a finite universe (see thesis #12). This leaves only the question of *when*collapse will occur, or, “is our current level of complexity before or beyond the point of diminishing returns?” To answer this question, let’s again take a look at each of the elements we’ve previously broken out separately: subsistence, information processing, sociopolitical control, economics, and technology.
- Agriculture and resource production.
- Information processing.
- Sociopolitical control and specialization.
- Overall economic productivity.
- Technological innovation.
Agriculture and resource production.
Industrialism allows the resource production of modern civilization to be reduced to a single figure: fossil fuels. Not only do fossil fuels provide energy for every segment of our economy, they even provide our food. In “The Oil We Eat,” Richard Manning discusses the nature of our “petroculture”:
The common assumption these days is that we muster our weapons to secure oil, not food. There’s a little joke in this. Ever since we ran out of arable land, food is oil. Every single calorie we eat is backed by at least a calorie of oil, more like ten. In 1940 the average farm in the United States produced 2.3 calories of food energy for every calorie of fossil energy it used. By 1974 (the last year in which anyone looked closely at this issue), that ratio was 1:1. And this understates the problem, because at the same time that there is more oil in our food there is less oil in our oil. A couple of generations ago we spent a lot less energy drilling, pumping, and distributing than we do now. In the 1940s we got about 100 barrels of oil back for every barrel of oil we spent getting it. Today each barrel invested in the process returns only ten, a calculation that no doubt fails to include the fuel burned by the Hummers and Blackhawks we use to maintain access to the oil in Iraq.
The reason for the loss of caloric efficiency in agriculture, as Manning discusses in detail, is the loss of arable soil. Monoculture–planting whole fields with just one plant, as with agriculture–drains that soil very quickly. Different plants take different things from the soil, and put other things back, in much the same way as plants and animal harmonize with one another in the oxygen-carbon dioxide cycle. By planting only one type of plant in a field, those things that particular plant needs is drained, but not replenished. Meanwhile, its waste products saturate the soil.
Because of the increasing agricultural complexity of the Green Revolution, the marginal returns for agriculture have dropped to astonishingly negative values. Every calorie of agricultural product returned requires ten calories of input. This is sustainable even in the short term only because of our fossil fuel subsidies.
That subsidy may be running out soon, though. Hubbert’s Peak, more popularly known as, “Peak Oil,” is the midway point of global oil production. Energy Bulletin’s “Peak Oil Primer” explains:
For obvious reasons, people have extracted the easy-to-reach, cheap oil first. The oil pumped first was on land, near the surface, under pressure and light and ‘sweet’ and easy to refine into gasoline. The remaining oil, sometimes off shore, far from markets, in smaller fields, or of lesser quality, will take ever more money and energy to extract and refine. The rate of extraction will drop. Furthermore, all oil fields eventually reach a point where they become economically, and energetically no longer viable. If it takes the energy of a barrel of oil to extract a barrel of oil, then further extraction is pointless.
In other words, the problem is not, strictly speaking, “running out of oil.” Rather, it is a state where the oil that remains provides the same amount of energy as ever, but continues to entail greater costs for its extraction. In other words, the “Peak Oil” problem is a problem of the diminishing marginal returns for our fossil fuel subsidy.
Thus, the question of whether we have passed the point of diminishing returns for resource extraction is the same as the question, “Have we yet passed Hubbert’s Peak?” There is increasing evidence that we may have done just that. Jeff Vail, an intelligence officer with the United States Air Force, wrote:
I gave an intelligence briefing to the Assistant Secretary of the Interior, Tom Weimer, today. He’s in charge of “water and science”, which includes the US Geological Survey, the agency in charge of the official government calculations on oil reserves and depletion. Most Peak Oil nay-sayers rely on the USGS’s 2000 report that shows an excessively optimistic projection for recoverable oil reserves, but what does USGS really think? All I can say for sure is that Weimer didn’t have any objections to my assertion that Peak Oil may well be a Fall 2005 event, nor that the world is facing a serious energy *supply* crisis in the near future. Does the government have some master Peak Oil plan? I have no idea, but claims that they are ignorant about the problem are simply incorrect.
OPEC, which provides most of the world’s oil, may be peaking. Saudi Arabia, though very secretive about its reserves, is having difficulty selling its crude oil–it is heavy, sour crude, not light, sweet. Suggesting that the Ghawar super-field has peaked. While previous estimates for the global Hubbert’s Peak hovered around 2015 - 2025, revelations that Shell and Saudi Arabia may be lying about their reserves have revised those estimates closer to the present or recent past. The EIA released a report stating that demand would outstrip supply in 2005 Q4 (PDF). And in fact, recent oil productionhas been consistent with the “plateau” one expects at the top of Hubbert’s Peak.
While the case is as yet ambiguous, there is mounting evidence that Hubbert’s Peak is now upon us, and thus, that we are currently passing the point of diminishing returns for resource production in an industrial context.
Information processing.
In the previous thesis, we cited Jeff Vail’s analysis of what is perhaps the world’s most efficient information processing hierarchy: the United States military. Vail highlighted that the *operational* span of control for each commander is 3, since the other 2 must be dedicated to information processing due to signal degradation problems through too many levels of hierarchy.
We have recently seen a drastic increase in information processing in global telecommunications, but this has been achieved by *sacrificing* hierarchy, and developing the technological infrastructure to allow for rhizome information processing. Open source methods have proven themselves far more efficient at information processing. “The Blogosphere” circulated news about the 2004 United States presidential election well ahead of the hierarchical mainstream media, while the Iraqi insurgency has successfully used the internet and “open source warfare” to counter the most powerful hierarchical military the world has ever seen. This investment in simplicity has yielded significant marginal returns, but it was made possible only by investments in greater technological and social complexity. And already, there are efforts to reassert hierarchical information processing methods over the internet, showing that such methods cannot long be tolerated by a civilized society.
Education also shows a point of diminishing returns has been reached. In The Collapse of Complex Societies, Tainter writes:
With increasing time spent in education and greater specialization, the learning that occurs yields decreased *general* benefits for greater costs. The greatest *quantities* of learning are accomplished in infancy; learning that occurs earlier in life tends to be more generalized. Later, specialized learning is dependent upon this earlier, generalized knowledge, so that the benefits of generalized learning include all derivative specialized knowledge. Axiomatically, therefore, generalized learning is of overall greater value than specialized.
Moreover, this early, generalized learning is accomplished at substantially lower cost. Malchup has compiled figures showing that, in 1957-8, education of pre-school children in the home cost the United States $4,432,000,000 (in income foregone by mothers), which yields $886,400,000 per year for ages 0 through 5. Elementary and secondary education cost $33,339,000,000, or $2,564,538,462 per year for ages 6 through 18. Higher education cost $12,757,000,000, or $2,514,000,000 per year for far fewer students, assuming an average of five years spent in higher education. In other words, the monetary cost to the nation of a year of education between pre-school, when the most generalized, highly useful education takes place, and college, when the most specialized learning is accomplished, increases by about 284 percent. And this increase would be even more dramatic if these figures took into account the fact that college enrollment is but a fraction of the available population.
Similarly … the overall production of investment in higher education for the development of specialized expertise has declined substantially since 1900. D. Price has demonstrated, in regard to the education of scientists, that educating more scientists causes those of average ability to increase in number faster than those who are most productive. Thus, increasing investments in specialized education yield declines in both marginal and average returns.
In 1924, S.G. Strumilin collected in the Soviet Union a set of educational data that reveal a corroborative pattern. He showed that the marginal return on investment in education declines with increasing education. The first two years of education, according to Strumilin, raise a Soviet worker’s production skills an average of 14.5 percent per year. Yet the third year of education yields an increase of only an additional 8 percent, while the fourth through sixth years raise skills only a further 4.5 percent per year.
So, there is a definite diminishing marginal returns curve for each individual’s education. This compounds to create a society’s point of diminishing returns because, as Tainter points out, a society that can satisfy its needs based on general education will return far more on its investment than those that require more specialized education. In the modern United States, intensifying complexity has led to the rise of the four year college Bachelor degree as the expected minimum of education, rather than simply the high school diploma. This is driven by the need for workers with more specialized knowledge to handle the various components of a more complex society. As such, society’s complexity is requiring heavier costs in education–passing a point of diminishing returns.
Sociopolitical control and specialization.
In the previous thesis, we mentioned the Bush administration’s creation of the Department of Homeland Security in response to the terrorist attacks of 9⁄11. At one point in time, this move may have yielded significant returns. However, in 2002, all of the departments it was unifying already had significant hierarchies and complexities of their own. Creating another level of complexity to subsume them merely exacerbated this situation.
Tainter provides still more evidence:
Between 1914 and 1967, the number of capital ships in the British Navy declined by 78.9 percent, the number of officers and enlisted men by 32.9 percent, and the number of dockyard workers by 33.7 percent. Yet during this period the number of dockyard officials and clerks increased by 247 percent, and the number of Admiralty officials by 769 percent. … Between 1935 and 1954 the number of officials in the British Colonial Office increased by 447 percent. During this same period, of course, the empire administered by these officials shrank considerably.
…
Bendix has compiled for private industry, in several nations, data similar to those Parkinson has uncovered in government. He was able to show that a pattern of increasing hierarchical specialization characterizes the private sector as strongly as Parkinson has demonstrated for the public. Clearly in the private sector, where economic succeess depends on efficiency, this pattern cannot be attributed to self-serving inefficiency. The reason why complex organizations must allocate ever larger portions of their personnel and other resources to administration is because increased complexity requires greater quantities of information processing and greater integration of disparate parts.
Even in 1977, Elgin and Bushnell concluded in “The Limits to Complexity: Are Bureaucracies Becoming Unmanageable?” (PDF) that the United States government was a Stage III organization, marked by severe diseconomies of scale, and due to the “ratchet effect” (a specific case of the type of positive feedback loop discussed in thesis #12) must soon become stage IV, critical and prone to collapse.
Nor is this only a burden for the public sector. In recent years, enterprise search has become a necessary commodity for any large-scale enterprise. The information processing burden is simply too great. Even enterprise search products are now becoming insufficient for the complexity such organizations face, creating a niche that my employer, Vivísimo, has very successfully exploited, with the development of a sophisticated “clustering engine” to organize such an overwhelming amount of data. In a whitepaper distributed by Vivísimo (PDF), the annual savings for an organization with 100 employees over conventional search products is calculated to be $1,012,000. This suggests the amount of investment being made into information processing even in the “efficient” private sector for such complexity.
Overall economic productivity.
As the information processing burden increases, and as the marginal returns of sociopolitical complexity diminish, the overall economy cannot help but suffer the same curve. Tainter writes:
Complex societies with large, well-developed economies have historically been able to sustain only rather inferior rates of economic growth. Latecomers to economic growth tend to have higher growth rates than early starters. … [R]ates of economic growth are highest in middle income countries, followed by high income and low income nations. Kristensen infers from these data that, through time, rates of economic growth tend to slow down … Such a trend suggests that societies with more developed economies face a situation in which the productivity of GNP for stimulating further growth tends to decline.
…
Zolotas has argued that the productivity of industrialism for producing social welfare is declining. In partial support of this assertion he points out that while U.S. per capita product increased 75 percent from 1950 to 1977, weekly work hours declined by only 9.5 percent.
Technological innovation.
The very notion that we have passed the point of diminishing returns for technology would seem to be the only one even more absurd than the very idea that technology is subject to diminishing returns at all–at least, to the techno-salvationist. In fact, the evidence is quite clear. In “Getting better value from information management,” published by *Information Economics Journal* in October 2003, Paul A. Strassmann notes:
The prevailing view nowadays is that IT will remain stagnant for a while. … An article by Nicholas Carr in the May 2003 issue of the Harvard Business Review resulted in a lively debate about its claim that IT spending will level off permanently because IT has become strategically irrelevant.
…
Why were organisations unable to take advantage of IT capabilities? The explanation is simple. Each firm had to organise its IT department, train its managers, educate its executives, develop most of its software and integrate vendor offerings with disorderly legacy code. It was easier to junk and re-build instead of to accumulate and grow. Vendors and consultants thrived with revenues growing faster than IT budgets. Out of total 2002 worldwide IT spending of $2 trillion the vendors and consultants reaped about 30%.
Financial executives are now asking where they can find the gains from IT spending. They are not looking for a small amount of money. For US manufacturing firms IT investments accounted for over a third of all new capital expenditures. For the US financial and services sector the IT investments consumed most of the capital used for acquiring non-financial assets.
Strassman is clearly addressing concern for marginal returns–the cost of IT, versus its benefit–and finding that it does not live up to its promises. This is in information technology, the field that has seen the most strikingly successful technological development in the past 50 years. For other areas of technology, things have been even worse. Tainter writes:
Despite Malchup’s caution, a number of factors suggest that the productivity of research and development has indeed declined. … [P]atents have been declining in respect to population and number of technical workers since about 1920, well before the R&D effort of World War II and thereafter. Even more significantly, patenting relative to numbers of scientists and engineers has declined continuously since 1900. Jacob Schmookler has compiled figures showing that, *excluding* government-financed projects, the number of industrial research personnel increased 5.6 times from 1930 to 1954, while the numbers of corporate patents rose between 1936-40 and 1956-60 by only 23 percent.
…
There are, morevoer, other data suggesting declining productivity of inventing activity in the industrial world. Hornell Hart has demonstrated consistent patterns of increasing and then declining rates of patenting (logistic curves) in many fields that are partially or wholly unrelated to military R&D. These include airplanes, automobiles, cotton machinery, electric meters, radios, sewing machines, spinning machinery, sulky plows, telegraphy, telephony, typewriters, and weaving machinery. He also noticed that the same patterns are evident in the major inventions and discoveries of the Western world, and in patents sealed in Great Britain between 1751 and 1820, and between 1821 and 1938.
Thus, it seems that military R&D cannot account for more than a small part of the decline in patents. Furthermore, the decline is so widespread in so many fields, over such a long time, that declining propensity to patent can hardly account for it either. Recent research shows that there is in fact a strong positive relationship between R&D and patenting. Thus the patent statistics appear to be a reliable indicator of inventing accomplishment.
It would appear that there has indeed been a genuine drop in the inventive productivity of research and development, and that as investments in R&D have increased (from 0.1 percent of gross national product in 1920 to 2.6 percent in 1960), the marginal product of these investments has declined. Although there are some demurrals, many economists recognize this trend.
That trend has continued. Jonathan Huebner charted the same trend from 1914 to 2005 in, “A Possible Declining Trend for Worldwide Innovation.” A Japanese report (PDF) from 2003 concluded that they, too, were suffering from having passed the point of diminishing returns in technology:
[W]e do not find strong evidence that Japanese innovative capacity has actually declined. However, that capacity has failed to grow at the rate of the 1980s. As a result, US and worldwide patent statistics suggest that Japanese firms in a number of sectors have fallen behind their US counterparts, even in areas where Japanese firms were formerly quite strong and rapidly converging on US levels of inventive output.
Medical technology, another field of significant investment in the past half century, has also shown signs of diminishing marginal returns. Penicillin, one of the most effective drugs ever devised, had a total production cost of approximately $20,000. According to a 2003 report by Bain & Co., the average cost of a new drug today is $1.7 billion. Writing about the study for Chemical & Engineering News, Rick Mullin writes:
According to Bain, the cost of drug development–currently 55% higher than the average cost from 1995 to 2000–is rising largely as a result of an increasing failure rate for prospective drugs in clinical trials. The rising cost of commercializing new drugs is another contributing factor–12 months of sales and marketing costs are included in Bain’s cost estimate but not in the Tufts figure.
If this is true, then the cost of developing new drugs is increasing exponentially, and largely due to the fact that most prospective drugs fail in clinical trials. Medical technology is incuring greater costs for less benefit–in the case of medical technology, that would be more “misses,” or work that never produce a viable drug.
Tainter provides another example of how we have surpassed the point of diminishing returns for complexity that does not fit easily under any of the above headings, as it applies to medical research and longevity:
Medical research and application provide a good example of a declining marginal return for increased investment in a scientific field. While it is less easy to measure the benefits of medicine than its costs, one sure indicator is life expectancy. Unfortunately, ever larger investments in health care do not yield proportionate increases in longevity. In 1930 the United States expended 3.3 percent of its gross national product (GNP) to produce an average life expectancy of 59.7 years. By 1982, 10.5 percent of GNP was producing a life expectancy of 74.5 years. … [F]rom 1930 to 1982 the productivity of the U.S. national health care system (measured thus) declined by 57 percent. (In fact, it is likely that the decline in the productivity of medicine has been even greater, for the effects of improved nutrition and sanitation on increasing life expectancy have not been included.)
From this data, Tainter concluded in 1988 that collapse was neither an option, nor an immediate threat. The reason, he said, was that the United States existed in a peer polity system, and that no single polity can collapse in such a system without being immediately reabsorbed by the whole.
For such brilliant insight, Tainter shows a disappointing inability to grasp the implications of his own theory at the end. The difference he draws between the collapse of isolated civilizations (such as Rome) and peer polity systems (such as the Maya) is arbitrary. We are dealing with “global” systems, whether we are dealing with literal islands, “islands” isolated by distance, geography, or culture, or the entire globe itself. The global system of complexity must collapse as a system. No single part can collapse in isolation, this is true. This is a direct result of the fact that civilization must always pursue complexity (thesis #13, and must always grow (thesis #12. Thus, when New Orleans collapsed, the United States government eventually moved in to restore its former level of complexity. This is one, arbitrary level we could look at; or, we could look at the arbitrary level of nation-states, and cite the collapse of the USSR and its immediate reabsorption into a similar level of complexity.
However, it would be obviously untrue to conclude from this that collapse is impossible. The only caveat is that the entire system must collapse as a system, not as individual, constituent parts. Thus, the Roman Empire collapsed as a system; the Mayan city-states collapsed not as individual city-states, but as a single system. We do not face the collapse of the United States or any one nation-state; we face the collapse of industrialized society itself. The scale of the nation-state has become arbitrary as well. As globalization proceeds, multinational corporations have risen to an unprecedented level of power, bisecting nations and undercutting their influence (*see* Jeff Vail, “The New Map: Terrorism in a Post-Cartesian World”). If nation-states are vertical powers, then multinational corporations create horizontal powers across them. This trend raises the question of whether we still truly live in a peer polity system at all, or if we are seeing the rise of some new level of complexity in such organizations as the United Nations, International Monetary Fund and the World Bank–the inevitable conclusion to intensifying complexity, with the emergence of a single, global civilization? Ultimately, the distinction is still a semantic one, however; peer polity systems often behave as though they were a single civilization, because the political alliances and economic relationships between them fuse them into a single system of social complexity.
As we have seen, the crisis of the diminishing returns on complexity is not only present, it is global. The same problems can be seen in every country; we have highlighted the United States here only for its greater wealth of available data, and, as the “capital” of the globalized civilization/peer polity system, it provides perhaps the most striking example. We have passed the point of diminishing returns for agriculture, information processing, bureaucracy, technology and the economy itself. All of these are intertwined, as we saw in the previous thesis. Having passed the point of diminishing returns, the collapse of such a system is inevitable.