Still the Iron Age

by Vaclav Smil

  Moreover, 30% of those 1.6 Gt produced in 2014 came from recycled metal, and the primary output of 1.1 Gt would extend the life of estimated iron resources to more than two centuries: think of the Congress of Vienna in 1814 worrying about iron ore supply in 2014 rather than about the organization of post-Napoleonic Europe. McKinsey (2013a) modeled the future iron ore demand using three global scenarios: a high-demand version would see 2.69 Gt extracted in 2020 and 3.38 Gt in 2030, a steady-growth scenario would end up with 2.44 Gt in 2020 and 2.63 Gt in 2030, and an early market saturation would call for only 2.4 Gt by 2020 and 2.35 Gt by 2030.

  And, as the review of flows and stocks made clear, yet another key adjustment must be made: to account for the secondary resource of the metal that is being constantly created by increasing steel stocks in buildings, infrastructures, vehicles, appliances, and other steel products, and whose increasing shares will be recycled, not only because it is conveniently available but also because its reuse reduces the industry’s energy intensity and environmental impacts, above all its substantial greenhouse gas emissions. Pauliuk et al. (2013) modeled global steel demand during the entire twenty-first century (an inherently uncertain exercise) and concluded that the demand for primary steel may peak as early as the year 2025 (creating an increasing excess of BF capacity during the century’s third and fourth decades); that the EAF route (including DRI) will reach parity with the BF–BOF path by 2050; that it will deliver nearly twice as much metal by the century’s end than BOFs; and that in some regions the rising availability of scrap will be larger than the final steel demand.

  Forecasts

  As in so many other instances, the simplest forecasts of steel production or steel demand are done by extrapolating recent growth rates in major steelmaking countries. But, obviously, sudden economic downturns or unexpected output spurts can take this approach way off the mark, even in fairly short periods of time. Seventy years of post-WW II changes in global steel production show how unadvisable it would be to take even a decade (in economic affairs a moderately long time span) as a basis of long-range forecasts. Between 1945 and 1974 worldwide steel production grew steadily, its increment averaging nearly 6% a year—but that growth (much like that of virtually all important economic variables) came to an abrupt end with OPEC’s quintupling of oil prices in 1973–1974. What followed was more than a quarter century of stagnation (less steel was produced in 1993 than in 1978) and slow growth as overall declines of Western production were only modestly surpassed by expansion in a few populous modernizing countries, above all in China, India, Brazil, and South Korea.

  In 2001, the global output of 852 Mt was only 20% above the 1974 level of 708 Mt, implying annual exponential growth of just 0.7%. But then came the Chinese takeoff, and output stabilization and some marginal recovery in affluent countries. Average annual growth of the global steel demand, which was just 1.9% per year during the last decade of the twentieth century, jumped to 7% per year between 2000 and 2005; it was 4.4% between 2005 and 2010 and almost 3% between 2010 and 2015. A slowdown has clearly been under way, and even slower growth is most likely in the near future as the China-driven rise between 2000 and 2013 was an exception that is unlikely to be repeated. McKinsey (2014) expects an average annual growth rate of 2.8% until 2020 (extremes of 2.2–3.5%) and 1% during the 2020s (range of 0.6–1.4%), close to the pre-2000 rate. Obviously, most of this growth will continue to be dominated by Asia, with China’s share of global steel output declining but still claiming 42% by 2026 and 38% by 2030, and with India still lagging but rising to about 10% by 2030.

  The other common forecasting choice is to relate future steel consumption to rising GDP. But (recall those substantial differences in national steel/GDP rates) this approach has to reckon both with many idiosyncratic national consumption patterns and, moreover, it can rely on (more or less) linear extrapolation only during the early stages of economic development. Pauliuk, Wang, and Müller (2013, 22) argue correctly that in order to understand future steel demand, “a detailed understanding of the evolution of steel stocks … is indispensable.” At the same time, forecasting future stock levels cannot be done with a high degree of accuracy as any such modeling exercise is affected by the choice of analytical boundaries, by the degree of sectoral subdivision used to trace stock levels, by the rates of scrap generation and consumption, by the extent of metal’s losses in fabrication, and by the level of international trade.

  Addition to stocks will, obviously, also depend on average lifetimes, and relatively small variations in assumptions can add up to substantial differences decades later. Assumptions for lifetimes of steel in construction range from 40 to 100 years, in transportation from 10 to 30 years, and in machinery and appliances from 10 to 40 years (Müller et al., 2006). Pauliuk et al. (2013) used averages of 15 years for products, 20 years for transportation, 30 years for machinery, and 75 years for construction. In another study, Pauliuk et al. (2013) used ranges of 13–27 years for transportation, 15–40 years for machinery, 38–100 years for construction, and 8–20 years for products. But Kozawa and Tsukihashi (2011) favored generally much shorter average lifetimes, just 35 years for infrastructure, 30 years for buildings, 14 years for machinery, and 9 years for cars.

  Pauliuk et al. (2013) performed analysis of in-use steel stocks for 200 countries (noting large uncertainties surrounding many of their values) and found clear saturation levels for all use categories. Per capita rates in industrialized countries range from 10±2 t for construction to 0.6±0.2 t for appliances and other steel products, while transportation steel stocks appear to saturate at about 1.5 t/capita and machinery stocks at 1.3 t/capita, and the overall steel saturation level is 13±2 t/capita. More specifically, Pauliuk et al. (2013) forecast per capita saturation levels at 13.2 t for North America by 2020, 15.4 t for rich East Asian economies also by 2020, 12.8 t for Western Europe by 2030, 13.7 t for China by 2050, and the same high rate for India and Africa but both as far in the future as 2150.

  In contrast, Fujitsuka et al. (2013) estimated that the saturation value for steel used in Asian buildings is twice as high as in other areas, the rates explained by the high incidence of earthquakes in East Asia (Japan, South Korea, China) where most of the recent steel consumption has taken place. They also estimated that by 2050 the global in-use stock of steel in infrastructures and cars will be four times the 2010 level, while for buildings it will be five times the 2010 level. And, assuming that the reuse of steel will remain at the rate prevailing in the early 2010s, they concluded that by 2035 the obsolete scrap generation will become about twice as large as the demand for the metal. And even if the near-term Chinese steel production declines by as much as 20%, annual additions to steel stocks should still surpass 1 Gt by 2020, and global in-use stocks should increase by at least 10 Gt between 2015 and 2025. Not surprisingly, Hatayama et al. (2010) forecast that these increases will be driven mostly by an expected 10-fold rise of steel consumption in Asia and that the total for structural and vehicular stock will reach 55 Gt by 2050.

  This is helpful to indicate the national and global levels of an eventual equilibrium, but it is of limited help in forecasting steel demand during the next 10–30 years, when the difference in economic development and national peculiarities may be the most important determinants of output and trade. Substantial steel demand will persist even in affluent and aging societies with stable, or even declining, populations. More durable designs may lengthen the recycling spans for cars and appliances, but if affluent societies are to enjoy continued high standards of living, they will have to address their deficits in infrastructural maintenance, and this would be a source of persistent steel demand. Continuing urbanization (particularly more highrises in expanding megacities) will keep creating strong demand for steel in construction and infrastructural sectors (especially for rapid transport, including steel-intensive subways) in all rapidly modernizing populous countries.

  The continuing rise of car ownership—the global count of passenger vehicles
had surpassed 100 million during the late 1950s (most of them still in the United States) and 1 billion in 2005, with China being the largest market since 2009 (with 23.5 million units sold in 2014)—will mask the difference between saturated, and even slightly declining, markets in rich countries and large potential growth throughout sub-Saharan Africa and among the poorest half of all populations in Asia and Latin America. As with so many other capital-intensive acquisitions, the future increases in ownership depend primarily on rising incomes, but lightweighting will also have a major impact on steel demand in the still-expanding car industry. Production will grow, but as a result of lightweighting (and also of gradual penetration of less steel-intensive electric vehicles), the sector’s total steel demand may see only a marginal increase by 2025 (McKinsey, 2013b).

  At the national level, it is safe to conclude that China’s experience will not be repeated: between 1992 and 2000 the country accounted for 35% of the world’s growth of steel production, between 2000 and 2007 that share rose to 62%, and between 2007 and 2014 it reached 98%. India is the only similarly sized country that could replicate that achievement, but its economy clearly has not taken off as rapidly as China’s. So far, the slowdown in China’s steel output has been gradual, but with the expansionary period over, it is natural to ask how far the maximum output will go and how soon a new production plateau will get established.

  Arguments for retreat are obvious. While the unmet demand in the poorest provinces (Guizhou, Yunnan, Gansu) leaves more room for growth, the industry’s overall large overcapacity, already high consumption levels in the most economically advanced provinces, unprofitability of small enterprises, and enormous environmental burdens point in the other direction. Most notably, Tianjin and Shanghai, two municipalities under the central administration, and the provinces of Jiangsu, Zhejiang, Liaoning, and Nei Monggol (with about 240 million people) already have average per capita steel use surpassing the peak US levels of the 1970s, before the American steel production began to decline.

  Not surprisingly, 2015 brought the first signs of retreat after decades of advances: in January 2015, China’s crude steel output was nearly 5% below the level of January 2014. But that does not mean that a secular decline has begun. Perhaps the best bottom-up analysis of China’s long-term steel demand offers a plausible sequence of rising and falling output. Yin and Chen (2013) concluded that the demand will peak at 753 Mt in 2025, and then decline to 510 Mt by 2050. As for sectoral use, construction demand should fall while automotive demand should rise rapidly before 2035, and it should claim 19% of the total output by 2050 (compared to just 6% in 2010).

  The other key determinant of future output will be the fate of the country’s effort to improve the quality of its incredibly polluted air. In February 2014, more than 15% of China’s territory experienced exceptionally heavy smog, with maximum daily levels of fine particulate matter (PM 2.5) in Beijing surpassing 900 μg/m3, while the WHO’s acceptable maximum is 25 μg/m3 (Dai and Gutierrez, 2014). Coal-fired electricity generation is by far the largest source of China’s air pollution, but steelmaking ranks as the second largest emitter of particulates and CO2, and the worst situation is in the regions of high steel mill concentrations in northern Hebei province (especially Tangshan municipality), which produce one-third of the country’s large steel output.

  In contrast, Credit Suisse (2012) argued that any steel consumption peak was “highly unlikely anytime soon,” mainly because China’s capital stock (including in-use stocks of steel) is still very low in relative terms (an order of magnitude lower than in the United States) and because the key factors driving higher consumption (demand by construction, transportation, and infrastructural improvements) will remain for years to come, while the expected structural decline “may take far longer than most expect.” Similarly, Walsh (2011, p. 36) concluded that “the China story has a long way left to run,” pointing out that the country’s average steel intensity surpassed 500 kg/capita only in 2011. In contrast, the United States steel intensity remained above that mark for 30 years, and by 2015 both Japanese and German intensities had been at such a high level for nearly 40 years, and South Korea for more than 30 years.

  An obvious counterargument is that these three countries have been exceptional exporters of steel-intensive cars and machinery and that there seems to be no early prospect of Chinese brands conquering global markets to such an extent as Toyota, Honda, Volkswagen, Daimler, Audi, BMW, or Hyundai have done. And it is also unlikely that in the next two decades India will come close to replicating China’s post-1990 rise. The combined population of the Indian subcontinent (of India, Pakistan, and Bangladesh) is already larger than that of China (1.6 billion in 2014, compared to 1.4 billion in China), and by 2025 (when their combined population will reach 1.8 billion) their steel demand will be only a third of China’s 2014 rate of about 550 kg/capita. Their annual steel consumption was about 325 Mt compared to about 90 Mt in 2013.

  A draft of India’s new national steel policy aims at “transforming Indian steel industry into a global leader in terms of production, consumption, quality, and techno-economic efficiency,” and a combination of domestic and foreign investment is planned “to reach crude steel capacity level of 300 million tonnes by 2025–26 to meet the domestic demand fully” (Government of India, 2012). And there is even greater potential demand throughout sub-Saharan Africa, with both of these regions having enormous infrastructural needs. McKinsey Global Institute (2013) put the global infrastructural spending between 2013 and 2030 at about $57 trillion, with steel-intensive projects accounting for more than 75% of the total—and while during that period India should finally realize many of its long-deferred economic aspirations, it is unlikely that either region will come close to sustaining more than a three-decade-long spell of intensive economic development akin to China’s post-1980 performance.

  Europe remains a major consumer of steel, but again, the continent’s future steel demand may follow a course of very gradual decrease (easily explained by the presence of already dense infrastructures, saturation of car and appliance ownership, and declining fertility) or of a steeper decline (resulting from chronic economic problems, rapid population aging, and stresses caused by political instability and mass immigration) that might end only with a formation of a new, and much lower, in-use equilibrium. North American prospects are no less uncertain: on the one hand there are already very high ownership levels of vehicles, appliances, and houses that bring steel used by those sectors close to, or even above, what may be the most likely long-term saturation level; on the other hand are huge, and growing, infrastructural deficits whose remedy would require substantial investment in steel.

  Not surprisingly, even the organizations seen as the most able forecasters can do no better than to resort to scenarios with widely differing outcomes. McKinsey (2013a) offered three very different scenarios of future steel demand: the high-growth version (labeled Scarcity) ends up with 1.94 Gt in 2020 and 2.62 Gt by 2030; in the moderate version (Steady Growth) the total remains below 2 Gt by 2020 (1.81 Gt) and rises to 2.21 Gt by 2030; and in the low-growth version (Saturation scenario) the total rises only modestly to 1.79 Gt in 2020 and to just above 2 Gt (2.06 Gt) by 2030. But none of these forecasts considers potential consequences of accelerated global warming.

  Continuing inadequacies of our long-range models make it impossible to offer confident predictions of the future global temperature rise (Palmer, 2014). As a result, we cannot exclude the possibility that, although not highly probable, some of the more extreme scenarios (e.g., average global temperature rise of more than 3 °C by 2050) may become a new reality. Obviously, such an unprecedented warming would call for unprecedented steps to manage the challenge—and steel production, the source of nearly 10% of all CO2 emissions from the combustion of fossil fuels, would be affected by efforts aimed at lowering the annual additions of carbon to the atmosphere. If that were the case, the third century of mass steel production, that will begin during the 2060s, wo
uld be very different from the first one that provided the foundations of modern civilization, and from the second one that has been extending its benefits worldwide.

  Appendix A

  Units and Their Multiples and Submultiples

  Basic SI Units

  Quantity Name Symbol

  Length meter m

  Mass kilogram kg

  Time second s

  Electric current ampere A

  Temperature kelvin K

  Amount of substance mole mol

  Luminous intensity candela cd

  Other Units Used in the Text

  Quantity Name Symbol

  Area hectare ha

  square meter m2

  Electric potential volt V

  Energy joule J

  Force newton N

  Mass gram g

  tonne t

  Power watt W

  Pressure pascal Pa

  Temperature degree Celsius °C

  Volume cubic meter m3

  Multiples Used in the SI

  Prefix Abbreviation Scientific Notation

  deka da 101

  hecto h 102

  kilo k 103

  mega m 106

  giga G 109

  tera T 1012

  peta P 1015

  exa E 1018

  zeta Z 1021

  yota Y 1024