In his book Mokyr discusses the sources of long-run economic growth across time and space. “The lever of riches” covers an enormous time period, ranging from the Antiquity to the modern era. Geographically, the author’s focus is mainly directed towards Europe and its periphery, and to a somewhat lesser extent North America and China.
In what follows I will not have time to cover all the different growth regimes that occurred in the two millennia stretching from the Antiquity to the Industrial Revolution and will leave that open for a future blog post. Instead, I will briefly outline the four different types of economic growth that can occur within an economy. Moreover, I will discuss in depth what Mokyr identifies as the source of long-run economic growth, that is, innovation and technological progress.
1. Solovian growth:
Named after Nobel prize winner Robert Solow who is famous for the so-called Solow growth model from 1956, one of the first formal models describing the process of long-term economic growth. The Solow framework shows that an economy can achieve transitory growth via capital accumulation. All else equal, an increase in the capital stock will make labor more productive and can thus lead to income gains. This process, however, is bound to run out of steam eventually as more and more capital gets accumulated. That is, of course, because additional units of capital are increasingly less productive, the concept of diminishing returns. As a sidenote one could mention that the Chinese economy adopted a growth model over the last few decades, which relied to a large extent on the accumulation of physical capital. One of the reasons for the current China slowdown is that the investment share of GDP is absurdly high and that additional increases in the capital stock simply don’t yield the same return (in terms of economic growth) as previously. The country thus has to switch towards a more consumption-based economy, but the adjustment process is painful in the short-run.
2. Smithian growth:
Based on Adam Smith’s “Wealth of Nations” in which he explains the concept of specialization of labor. The basic idea is that individuals are more productive if they specialize into one occupation and then trade with each other instead of producing a variety of goods at the household level. The argument equally holds on a bigger scale, that is between regions and even countries. Increased specialization in the production of goods combined with gains from trade are thus an alternative source of economic growth. Smithian growth is, of course, only possible if a critical mass of population is reached so that increased specialization can actually occur.
3. Size/scale effects:
The scale effect is all about how population size can lead to economic growth. There are, in fact, several mechanisms through which increased urbanization, for example, can boost productivity. First, there are very good reasons that there are various externalities associated with the emergence of cities. People are more productive in cities because productivity is to a big extent dependent on the people who are around us.
Furthermore, it makes sense to spread out the fixed cost of infrastructure investments over as many individuals as possible. It is self-evident that certain types of infrastructure investments, no matter how growth enhancing, only pass a cost/benefit test at a certain population size.
4. Schumpeterian growth:
Schumpeterian growth relies on technological progress as a source of income gains. Innovations to the production process can for example increase efficiency, i.e. firms can adopt new technologies to produce the same amount of output with fewer inputs (of labor, energy, or capital). But advances in sciences and technologies are, of course, not only limited to increasing the efficiency of current production. Some innovations create entirely new products, some of them so pathbreaking that they potentially might have huge effects on the larger macroeconomy. This type of growth based on technological change is named after Joseph Schumpeter, the Austrian economist who proposed the concept of “creative destruction”, the process in which new production units replace outdated ones. Firms with new technologies or just better production methods drive uncompetitive firms out of business over time. The demise of certain industries and jobs is thus the destructive part of the process of technological progress. But at the same time innovation creates new opportunities and it is the adoption of best-practice techniques and new technologies within industries that can assure economy-wide income gains and thus prosperity in the long-run.
Macroinventions/General Purpose Technologies:
It is crucial to note here that the first three types of growth will necessarily run out of steam in the end. That is because capital accumulation exhibits diminishing returns, gains from trade will at a certain point in time be exhausted, and the scale effect might eventually turn negative because with rising agglomeration size the negative congestion effects will start to dominate (think Chinese metropolitan areas, for example, where congestion and pollution have increasingly large economic costs).
It is thus only Schumpeterian growth, i.e. growth based on innovation and technological progress, that can assure long-run income gains for society. It is useful to distinguish here between what Mokyr has labeled microinventions and macroinventions. Microinventions are basically small, incremental changes that improve and adapt existing techniques already in use. They are by far the more common as business try to improve their production techniques all the time to remain competitive. This behavior can supposedly produce the sort of gradual technological change (technological drift) that is an inherent feature of most of the neoclassical growth models, such as the Solow model, for example. In these stylized models long-run economic growth resulting from technological progress is a continuous and steady process, which does not display any kinds of long-run fluctuations.
However, economic history suggests that technological progress is a highly uneven process. The idea that it is a steady and continuous process over long periods of time is actually quite ridiculous and also contradictory to what Schumpeter himself suggested: He claimed that new technologies produce long cycles, i.e. periods of high and low growth that alternate because innovation in some years or even decades would simply be higher than in others. This, of course, could be because the invention of new technologies might come in clusters. Innovations in one sector potentially spread to other sectors, producing a general “boom”. Conversely, an innovation slowdown in one dominant industry can just as well spread across the entire economy, thus causing a general economic slowdown.
Here Mokyr’s concept of macroinventions comes to mind. A macroinvention is a game-changing technology that transforms the entire industry, and has potentially very large affects on the entire economy. Nowadays, economists rather use the term “General purpose technology” (GPT) instead of macroinvention, but the idea is exactly the same. The steam engine, railroads, electricity, the computer, and the internet are all prominent examples of GPTs, i.e. technologies that fundamentally altered the economy’s current pattern of production and consumption because they affected numerous industries at once.
One should note that some more recent growth models (endogenous growth models) have incorporated the idea that long-term economic growth is driven by the invention of GPTs. Interestingly enough, the adoption of a new GPT might at first lead to a slow-down in economic growth. That is because with the invention of the new technology a lot of the existing capital stock in the economy becomes obsolete. Furthermore, workers might not have the right skills to work with the new technology and must first augment their human capital (via intensive training or schooling) to adapt to the new requirements. As a consequence, with the invention of the GPT comes a period of transition in which the economy might actually experience lower productivity growth because a substantial fraction of the existing physical and human capital becomes obsolete and must therefore be replaced. Moreover, Jovanovic and Rousseau (2005) suggest that during the time of rapidly increasing electricity adoption the economy experienced a rise in “creative destruction” as measured by the entry and exit of firms and by the number of mergers and takeovers. It thus seems to be that the economy enters a phase of rapid transition after the introduction of a GPT, generally characterized by many structural changes, which lead at first to a slowdown in productivity (http://www.nyu.edu/econ/user/jovanovi/JovRousseauGPT.pdf).
Over time the capital stock of industries and the skills of workers adjust to the new technology and productivity will start to accelerate. This process, however, might be very lengthy, as shown by the example of electricity. Only decades after its first adoption did electric power become a major input for all industries. Sweden was one of the countries that started electrification quite early in the 20th century, indeed earlier than many other industrialized nations. The first large hydroelectric power stations (relying on dams) were built in between 1910 and 1920. Many rural regions, however, had to wait a significant amount of time until they gained access to the electricity grid. Only by 1945 were about 86% of all Swedish households connected to the grid. Again, in comparison to other countries this number is quite spectacular. But it also shows that it took Sweden roughly three decades to construct a national electricity network also covering the more rural areas of the country. It is thus not very surprising that the large productivity gains from electrification only materialized with a significant time lag after its first introduction. That is, of course, because the penetration of the GPT has to reach a certain threshold value first until it can have a an impact on economy-wide productivity.
What are the implications?
Economic history thus suggests that economic growth is an uneven process, even in the modern era. GPTs can lead to long economic cycles where periods of slow productivity growth alternate with periods of rapid productivity growth. The length of these cycles is unpredictable and will depend on a large variety of factors, such as:
- The speed at which the GPT penetrates the different industries within the economy.
- The speed at which obsolete physical capital is replaced and at which workers can augment their human capital to learn the newly required skills.
- The duration of the productivity boom once the GPT has penetrated all industries and starts to have a transformative effect on production methods
The sobering conclusion from the study of economic growth is probably that it is close to impossible for economists to predict the pace of future long-term productivity growth because we simply do not know what the next GPT will be, when it will be invented and to what extent it will transform the economy. The amplitude and the duration of the next cycle can thus not be predicted since we even have hard time figuring out the current cycle. The computer and all related internet technologies are often classified as a GPT and led to a sudden burst of productivity growth in the 1990s. However, since mid-2000s productivity growth has slowed down considerably, even more so since the beginning of the crisis.
Indeed, world productivity growth has been flat for several years in a row now. So much, by the way, for the theory that robots are stealing our jobs! If that were the case, productivity should accelerate instead of decelerate. All the new IT-related technologies have definitely led to a large increase in economic welfare. We all benefit from having access to Facebook, WhatsApp, Instagram and many other Internet-based services all the time. However, they have not resulted in large productivity gains. One reason could potentially be that many of these new products have been directed towards improving our leisure time instead of being directed towards improvements in production methods: Being able to use Facebook and WhatsApp constantly is definitely a plus from a consumer point of view, but it doesn’t make labor more productive. To the contrary, one could even imagine that there are minor negative productivity effects just from the fact that nowadays one can have access to many more distractions than in the not so distant past (speaking from experience here).
References:
- Jovanovic, Boyan & Rousseau, Peter L., 2005. "General Purpose Technologies," in: Philippe Aghion & Steven Durlauf (ed.), Handbook of Economic Growth
- Joel Mokyr: “The lever of riches” (1992)
- Joeseph Schumpeter: Capitalism, Socialism and Democracy (1942)