Corresponding author: Rostislav Kapeliushnikov ( rostis@hse.ru ) © 2019 Non-profit partnership “Voprosy Ekonomiki”.
This is an open access article distributed under the terms of the Creative Commons Attribution License (CC BY-NC-ND 4.0), which permits to copy and distribute the article for non-commercial purposes, provided that the article is not altered or modified and the original author and source are credited.
Citation:
Kapeliushnikov R (2019) The phantom of technological unemployment. Russian Journal of Economics 5(1): 88-116. https://doi.org/10.32609/j.ruje.5.35507
|
Nowadays there are many gloomy prophecies provided by both technologists and economists about the detrimental effects of the so-called Fourth Industrial Revolution on aggregate employment and its composition. These prophecies imply that in the near future we will face Robocalypse — a massive replacement of people by machines alongside an explosion in joblessness. This paper provides theoretical, empirical and historical evidence that the phenomenon of technological unemployment is a phantom. The most general results can be summarized as follows: in the long run, reduction in labor demand under the impact of new technologies is merely a theoretical possibility that has never before been realized in practice; at the level of individual firms, there is a strong positive relationship between innovations and employment growth; at the sectoral level, technological changes cause a multidirectional employment response, since different industries are at different stages of the life cycle; at the macro level, technological progress acts as a positive or neutral, but not a negative factor; a surge in technological unemployment, even in the short-term, seems a remote prospect since in coming decades the pace of technological change is unlikely to be fast enough by historical standards; the impact of new technologies on labor supply may be a more serious problem than their impact on labor demand; technological changes seem to have a much greater effect on the composition of employment than on its level.
technological change, labor demand, compensation theory, technological unemployment, computerization, robotics
“An unlimited number of jobs are available in a world of scarcity.”
(Alchian and Allen, 1972, p. 827)
In recent years, an avalanche of apocalyptic predictions of devastation to employment prospects has struck the public, triggered inevitably by the so-called Fourth Industrial Revolution, connected with the latest technological advances — robotization, digitalization and the creation of artificial intelligence, etc. The bearers of catastrophic forecasts on this issue are politicians, journalists, sociologists, futurologists, engineers, and many others. Although the majority of economists traditionally retain a certain immunity to such prophecies, nevertheless quite a few alarmists currently figure among them. We are told that as a result of the implementation of new technologies, a huge mass of people will remain out of work — machines will finally have triumphed in the race between machines and people (
The future state of labor market is depicted in the darkest colors. A well-known American engineer from Carnegie Mellon University Vivek Wadhwa warns: “The reality is that we are facing a jobless future: one in which most of the work done by humans will be done by machines. Robots will drive our cars, manufacture our goods, and do our chores, but there won't be much work for human beings.” Bill Gates agrees with him: new technologies “will reduce demand for jobs, particularly at the lower end of skill set”. Futurist Martin Ford, who published a book in 2015 under the revealing title “Rise of the robots: Technology and the threat of a jobless future” (
The picture of the world emerging from such statements appears as a bizarre mixture of excessive optimism and excessive pessimism: over-optimism in terms of the prospects for modern technological change, and over-pessimism in terms of the ability of the economy (in particular, the labor market) to adjust to its upcoming breakthrough achievements.
Immediately, we note one oddity. Conversations about the Fourth Industrial Revolution began when its fruits — at least, so far — were not visible in general. The situation with the First (steam engines), the Second (electricity and internal combustion engines) and the Third (computers) industrial revolutions was different: first, they sharply accelerated the growth of labor productivity and radically transformed living conditions and only some time later, retrospectively, were they recognized as a “revolution”. Now, we are not seeing a sharp acceleration in the dynamics of labor productivity (rather, the situation is exactly the opposite) or signs of a cardinal breakdown in people’s usual way of life. In essence, the subject matter under discussion is not so much the real, but rather the expected state of affairs, and no one can be sure whether it will really occur. In this sense, it is characteristic that many researchers regard the technological achievements taking place today not as manifestations of the Fourth Industrial Revolution, which has already come, but only as the “tail” end of the Third Industrial Revolution — with its distant, and weakened, consequences still manifesting themselves (
As for the idea of “technological unemployment” itself, it is far from new and has a near two-hundred year history, although this term itself was introduced into the scientific lexicon by J. M. Keynes not long ago — only in the 1930s.
When analyzing the possible impact of technological progress on employment, two aspects of the problem should be clearly distinguished —longterm and shortterm. In the first case, we are talking about a permanent reduction in the demand for labor under the influence of new technologies, in the second — a temporary increase in unemployment due to the increased discrepancy between the structure of labor demand and the structure of labor supply (meaning that when the transition period associated with adaptation to new conditions is over, unemployment will return to the “normal” level). The long-term and short-term effects of technological change need not be the same. In the following discussion, we will consider both.
Over the long years of studying the problem of technological unemployment by several generations of economists, economics has accumulated many theoretical arguments, empirical facts and historical evidence to assess whether the predictions of today’s techno-alarmists are justified and whether in the near future we should expect complete, or at least partial, but nevertheless large-scale, crowding out of employees.
According to the results of the analysis, which stretched for almost two centuries, in order to denote a logical error, which many easily fall into when discussing this problem, economic theory even developed a special term — the lump of labor fallacy (misconception implying a fixed volume of labor). We are talking about conclusions by type: “if labor productivity as a result of the introduction of new technologies increased by X percent, then the demand for labor will also decrease by X percent.” This syllogism is false, since it is based on the assumption that the volume of output is fixed and does not take into account the actions of various macroeconomic feedback mechanisms. In fact, while increasing labor productivity, output does not remain unchanged: its growth entails an increase in incomes of either entrepreneurs who have introduced innovations, or workers who begin to use more advanced equipment, or consumers who receive access to cheaper goods, and most often, of all at a time. Increased incomes are translating into higher consumer and investment demand, and it cannot be satisfied without attracting additional workers. As a consequence, the relationship between the dynamics of labor productivity and the dynamics of demand for labor appears to be extremely complex and not unidirectional. Due to indirect macroeconomic effects, it may well turn out to be not negative, but, rather, positive. In other words, both in theory and in practice, a situation is quite probable when the introduction of new technologies will not reduce, but increase the number of jobs in the economy. However, from the statements of current techno-alarmists, it is clearly seen that most of them (unless they are professional economists) in principle are unaware of the existence of such indirect links and still continue to use “frontal” arguments, the failure of which has been convincingly revealed by economic theory almost two hundred years ago at the beginning of the 19th century.
Indeed, the main theoretical ideas concerning technological unemployment as a long-term phenomenon were also expressed by classical economists (
David Ricardo started the discussion of this issue among classical economists when he included a new chapter “On Machinery” (
Ricardo’s conclusion that the introduction of machines would result in a reduction in the demand for labor in the long term caused a detailed response from other classical economists — J. S. Mill, J.-B. Say, J. McCulloch, N. Senior.
The first, most fundamental, compensation mechanism is related to the fact that under perfect competition, the growth of labor productivity will lead to a proportional reduction in prices for goods, which in turn will stimulate additional demand for them (it is believed that the first to pay attention to this compensation channel was a contemporary of Smith J. Stewart).
The second — “investment” — mechanism, is connected to the fact that productivity growth means an increase in the profitability and competitiveness of innovation firms. In response, they will begin to increase investment, which will automatically entail the creation of new additional jobs. The first to describe the operation of this compensation mechanism was Ricardo himself; later, its importance was emphasized by A. Marshall, J. Hicks and many other authors.
The third mechanism was identified by
The fourth mechanism, to a certain extent opposite to the previous one, is based on the assumption of wage growth, when, as a result of technological innovations, more advanced equipment is at the disposal of workers. Indeed, replacing people with machines means increasing the capital-labor ratio, and this, according to the neoclassical growth theory, should increase marginal labor productivity. Correspondingly, the wages of employees who have retained employment should also increase, and since their increased incomes will be translated into additional demand for goods and services, additional jobs will be created to satisfy it.
All the market compensation mechanisms discussed above describe the possible consequences of process innovations, that is, changes in production methods. But classical economists have not ignored the possible consequences of product innovation, connected with the introduction of new types of goods and services to the market (although, of course, no such terminology existed at that time). They believed that such innovations, by definition, are labor-intensive and, accordingly, should be accompanied by an increase rather than a reduction in the total number of jobs, which even
As we see, in the way as economic theory portrays it, the relationship between technological change and employment appears to be extremely complex and ambiguous. The combined effect of various market mechanisms can compensate for the initial labor-saving effect of new technologies, either partially or completely, or even overcompensate for it, so that compared with the initial situation, the total number of jobs in the economy will not only decrease, but increase. Based on purely theoretical considerations, one cannot a priori say which of these scenarios will be implemented in a particular case: in fact, this is an empirical question. The net (“pure”) effect of technological changes in terms of their impact on employment dynamics will depend on the ratio between product and process innovations, as well as on the balance between various compensation mechanisms. But even in the conditions of only partial compensation, it is safe to say that, taking into account the various indirect effects, the reduction in employment will in any case turn out to be much less than is assumed by naive arguments like “if labor productivity has increased by X percent, then, the demand for labor must fall by the same X percent”.
If we turn from theory to the facts of economic history for various countries, it becomes obvious from them that periods of accelerated growth in labor productivity almost always turned out to be simultaneous with sharply increased demand for labor. (As an illustration, we refer to the recent experience of the Russian economy: in 2000s, high growth rates of labor productivity were accompanied by a significant increase in employment.) As noted by
A huge amount of empirical research is devoted to the impact of innovations on employment. Such an analysis can be conducted at several levels — micro (individual firms), meso (sectors or regions) and macro (the whole economies). For many reasons, analysis at different levels can produce divergent results.
One of the key problems arising in this context is measurement. How can we quantify the pace of technological change? In the literature, a variety of indicators are used as a proxy for measuring the pace of technological change — expenditures on research and development, investments in new equipment, number of patents, activity in the use of ICT and many others; recent work is increasingly using such a new important indicator as the level of robotization. Obviously, when using different metrics, the results of the analysis will not necessarily coincide.
What does the available empirical evidence say?
Almost all studies at micro-level point to the strong positive effect of technological change on employment.
A number of micro-level studies have been based on cross-country data. The study on four European countries (Great Britain, Germany, Spain and France, 1998–2000) showed that in the manufacturing industry of these countries product innovations were associated with a significant increase in employment at the firm level, while process innovations were associated with weak decline in it. By contrast, in services the former provided only a small increase in employment, but the latter did not affect it either positively or negatively (
In general, research at the micro level allows us to conclude that there is a strong positive effect of technological change on employment, especially in the service industries and high-tech manufacturing industries; at the same time, robotization may not be as positive for the dynamics of employment as many other types of modern technological innovations.
However, analysis at the level of individual firms has serious limitations. First, it rarely turns out to be able to effectively solve the problem of causality: it is possible that innovative firms do create jobs more actively, but it is also possible that more successful and therefore more rapidly growing firms will be more likely to introduce innovations. Secondly, such an analysis can significantly overestimate the positive effect of technological changes in terms of employment dynamics if innovative firms manage to increase their market share at the expense of non-innovative firms, which, therefore, have to curtail employment. Thirdly, it is not able to take into account the indirect effects of compensatory market mechanisms operating at the level of entire sectors or the entire economy.
Partially, these limitations are overcome in the analysis at the sectoral level. With its use, the variance in estimates is much larger due to the idiosyncratic situation in different industries. Some studies using meso-data conclude that modern technologies exhibit “hostility” toward employment.
Thus, using the example of 21 manufacturing industries in five European countries (1989–1993), it was demonstrated that job losses were higher in sectors characterized by higher innovation activity (
At the same time, in the already mentioned study on France, which covered 18 branches of the manufacturing industry of this country, it was shown that in more innovative industries jobs were created at a higher rate than in less innovative ones. Moreover, unlike the results obtained in the analysis at the firm level, the effect of product innovations was stronger than that of process innovations (
Important results were obtained in a recent paper by
The most general conclusion that can be drawn from research at the meso level is that in various sectors the demand for labor responds unequally to the introduction of new technologies: as a rule, its response is positive in services, as well as in “young”, high-tech manufacturing industries, but negative in its “mature” low-tech industries. However, analysis at the sectoral level has many limitations similar to that at the firm level. Nor can it take into account the effects of various compensating mechanisms at the macro level, and, therefore, is unable to answer the question regarding the overall effect of technological changes in terms of employment dynamics.
Therefore, research based on aggregated data is of great importance. In particular, it allows to evaluate the comparative effectiveness of alternative compensatory market mechanisms. For this, various methodological approaches can be used (
In early work, simulations based on input-output tables were used as the main analytical tool. So, from a series of simulations performed for the US economy of the mid-1980s, it followed that employment in it would grow under any possible rates of technological change, although in scenarios with higher rates its growth should be slower (
Another research direction was associated with attempts to directly assess the elasticity of employment by economic growth rates. The research for several OECD countries that covered the years 1960–1993 showed that the correlation between the rates of change in employment and in GDP is positive, and it remained so even in the first half of the 1990s — in the period that is commonly described as “economic growth without creating jobs” (
The relationship between employment dynamics and productivity dynamics is analyzed in detail in a sample of 19 developed countries for the years 1970–2007 in a recent paper by
But perhaps the most interesting studies at the macro level present an attempt to assess the effectiveness of various compensatory mechanisms in the framework of general or partial equilibrium models. As it was shown for the USA (
In
Despite discrepancies in existing empirical estimates, it is impossible not to notice that most studies at the aggregate level still tend to conclude that technological change is a factor that favors rather than hinders employment growth. At the same time, they show that much here depends on the degree of flexibility of the labor market, on price elasticity of demand (both for goods and labor), on the elasticity of substitution between production factors, etc. (
Most recently, analyses have emerged whereby regions, rather than firms, sectors or national economies, are the objects of observation. Probably the most resonant of them is the study by
Note that even if we take these estimates at face value, quantitatively, the effect of robotization on employment in the United States turns out to be barely noticeable — a decrease by 0.2–0.3% over an almost twenty-year period. Suffice it to say that only the annual labor turnover in the American labor market (the sum of hiring and firing) exceeds 120 million. Moreover, according to the calculations of
Finally, all studies that also use regional data paint a different picture from
In
Technological change can become the source of a rise in unemployment not only when it reduces the demand for labor, but also when it complicates and slows down the process of matching workers with jobs. The fact is that under its influence not only the level of labor demand can change, but also its structure. To overcome the discrepancies that have arisen between the composition of labor demand and composition of labor supply, reallocation of labor turns out to be necessary — occupational, territorial and interfirm etc. Some occupations become obsolete, others appear; new technologies put higher requirements on workers’ education and skills; the workforce has to move from regions where the need for it falls to regions where the need is greater; non-innovative firms start dismissals (or they close altogether), while innovative ones open new vacancies, so jobs start to flow from the first to the second. In fact, this is exactly what J. Schumpeter called the process of “creative destruction”.
Naturally, with sharp technological changes, adaptation to them can stretch for a long time and take extremely painful forms. Retraining, learning new trades, raising the level of education, moving to another locality, even changing the place of work, all this takes time and involves considerable costs. In this sense, there is no doubt that technological unemployment as a short-term phenomenon is quite real and, moreover, is always more or less present in modern labor markets. The meaningful question is whether it goes beyond the limits of “normal” frictional unemployment and, if so, how strongly, and whether it resolves over time and, if so, how quickly. For example,
There are many factors acting in different directions. Nevertheless, in the general case, it can be expected that the increase in unemployment will be the more significant and longer in the following hypothetical contexts: (a) the larger the required reallocation of labor, that is, the faster, deeper and broader the technological changes that give it a boost; (b) the wider the gap between the requirements imposed by the old and new technologies to the quality of the human capital of workers; (c) the higher the inflexibility of labor markets, which impedes the process of reallocation and slows down its pace.
Moreover, if we are not talking about a one-time shock, but about a constantly reproducing situation, when the rate of technological change persistently exceeds the rate of adaptation to them by firms and workers, then technological unemployment can be observed not only in the short term, but also in the long term. In other words, under certain conditions, technological change can become a cause for increasing the “natural” (equilibrium) rate of unemployment.
For instance, it can increase the propensity of firms to impose layoffs and weaken their propensity to open vacancies (
However, this result cannot be considered predetermined. Technological change can take various forms: sometimes it requires the elimination of existing jobs, but sometimes everything boils down to their renovation (
How often are these episodes of ultra-high short-term technological unemployment? Economic history shows that it is extremely difficult to find examples when the acceleration of technological progress would lead to a sharp jump in general unemployment. At the beginning of the 20th century, the appearance of automobiles did not cause mass unemployment among cab drivers, blacksmiths and saddlers; at the end of the same century, the advent of computers did not cause mass unemployment among typists. This means that most often the speed of technological changes and the speed of adjustment to them are comparable. In previous periods, the introduction of new technologies, as a rule, did not have an explosive character, stretching for a more or less long period, sufficient for firms and workers to adapt to the changed conditions. Thanks to this, it was usually possible to avoid any sharp jumps in unemployment due to technological factors. (For example, in the USA, where, as we mentioned, the labor turnover exceeds 120 million a year, in order to significantly affect the overall unemployment rate, a one-time inflow into it caused by the introduction of new technologies would have to be at least 0.5–1 million people.)
Nevertheless, some authors tend to associate current problems in the labor markets of developed countries with technological factors. Thus, more recently,
But do we still expect a rapid surge in technological unemployment in the near future, as techno-alarmists predict? On closer examination, such a development of events seems highly unlikely, for several reasons.
First, as the empirical analysis shows, modern technological change is focused not so much on the elimination of jobs, but rather on their renovation. Thus, computerization has contributed to the large-scale replacement of people by machines in the process of performing routine production operations (for more on this, see the next section). But, as calculated on the data for Germany for 1979–1999 by SpitzOner (2006), this was 99% due to a reduction in routine operations performed by employees within existing occupations (that is, without dying out!) and only by 1% due to the elimination of occupations that were entirely routine. But this form of technological change, as already noted, is more likely to be associated with the expansion of employment than with its contraction.
Secondly, many alarmist forecasts are based on the simplified division of all employees into two groups — low- and high-skilled. And if we assume that new technologies reduce the demand for the former, but increase the demand for the latter, then it is easy to come to the conclusion that large-scale technological unemployment is inevitable because, for obvious reasons, low-skilled workers forced out by machines are not qualified to perform high-skilled jobs. However, the division of all employees into two polar groups is just a conventional device to simplify the analysis. In reality, there is a huge variety of gradations depending on the quality of jobs. Nothing is impossible in that, under certain conditions, workers who were at the bottom of the occupational hierarchy could go up one step; the workers who occupied this step earlier also climbed one step and so on up to the very top. The possibility of such a “chain” substitution between various groups of workers dramatically reduces the risk of a sudden surge in technological unemployment.
Thirdly, the advancement of new technologies may encounter not only engineering, but also legal, social and ethical obstacles, which reduce the speed of their spread. For example, a massive transition to driverless cars is impossible without a radical revision of the legislation on liability in case of an accident, but such a revision may take many years.
Average annual growth rates of GDP and total factor productivity (TFP) in developed countries, 1890–2015 (%).
1890–1913 | 1913–1950 | 1950–1975 | 1975–1995 | 1995–2005 | 2005–2015 | |
USA | ||||||
GDP | 3.8 | 3.3 | 3.5 | 3.2 | 3.4 | 1.4 |
TFP | 1.3 | 2.5 | 1.8 | 1.1 | 1.8 | 0.6 |
Great Britain | ||||||
GDP | 1.7 | 1,3 | 2.9 | 2.4 | 3.0 | 1.0 |
TFP | 0.5 | 1.2 | 1.8 | 1.8 | 1.6 | -0,1 |
Eurozone | ||||||
GDP | 2.4 | 1.0 | 5.1 | 2.5 | 2.0 | 0.6 |
TFP | 1.4 | 1.2 | 3.6 | 1.8 | 0.7 | 0.2 |
Japan | ||||||
GDP | 2.5 | 2.2 | 8.2 | 3.7 | 1.1 | 0.5 |
TFP | 0.5 | 0.7 | 4.4 | 1.7 | 0.9 | 0.4 |
But perhaps the most important thing is that so far (contrary to the predictions of techno-alarmists) nothing indicates a sharp acceleration of technological change that awaits us in the near future. As Table shows, the last decade can be considered one of the worst periods in the economic history of developed countries. During these years, they had the lowest growth rates since the beginning of the 20th century, not only of GDP, but also of total factor productivity, which can be considered evidence of a slowdown in technological change. Of course, one can argue that such disappointing figures are nothing more than the result of the negative impact of the Great Recession of 2008–2009. However, upon closer inspection, this reference to the Great Recession is hardly convincing.
First, the long-term slowdown in productivity growth began in developed countries several years before the Great Recession (in the US since 2005) (
Today, most researchers agree that the sharp deterioration in GDP and productivity dynamics in developed countries is not a short-term episode; that since the mid-2000s they all moved to a lower long-term growth path; finally, that in the first place this was due to the slowdown in the rate of TFP, that is, rate of technological change. Of the American economists, only Brinjolfsson and McAffie (2011), who are among the convinced techno-optimists, continue to insist that in the near future the annual growth rates of TFE in the US will accelerate to 2%. However, since they first made this forecast, several years have passed, but there is still no sign of the promised acceleration. All other authoritative researchers of economic growth in the United States (D. Jorgenson, R. Gordon, J. Fernald and many others), on the contrary, expect a serious slowdown in the growth rates of TFP to 0.5–1%. According to their estimates, in the best case, the American economy will have to return to the indicators observed in the “stagnant” twentieth century years of 1972–1995.
But if this was the case, then for this reason alone it makes no sense to expect any surge in technological unemployment any time soon. If the pace of technological change, as follows from most forecasts, will be even lower than the historical “norm”, then it will be no more difficult to adapt to them, but easier than before. In light of this, even a short-term spate in technological unemployment in the coming years seems unlikely.
Of course, technological progress is hardly a predictable process, being rather fraught with many surprises, so the situation can revert very quickly. Perhaps, already somewhere on the way there are new breakthrough technologies that will once again transform the world. But, so far,available data paint a completely different picture: it seems that the world economy has a long period of — not too fast by historical standards — technological change and, as a result, rather sluggish productivity growth.
At the same time, in discussions about technological unemployment, it is often forgotten that new technologies can influence not only labor demand, but also labor supply. And if effects on the demand side are largely an area of guesswork and assumptions, effects on the supply side are already a reality.
Traditionally, in addition to monetary losses, the unemployed carried no less significant non-monetary (psychological) costs, falling into idleness and social exclusion. However, thanks to computers and the Internet, such costs have rather diminished: now the non-employed have something to do (say, spending time playing video games) and have someone to communicate with (through social networks). And to the extent that the value of leisure has increased, the incentives to seek and obtain paid employment in the market should have weakened. In other words, the unforeseen result of the diffusion of new technologies could be a decrease in the employment rate, and especially so among young people.
Indeed,
In this sense, there are grounds for asserting that today the serious challenge for economic and social policy is not so much the impact of new technologies on labor demand, as their impact on labor supply.
In the past two decades, a large number of studies have appeared in which the topic “technological change/labor demand” began to be viewed from a fundamentally new angle, namely, through the prism of changes in the occupational composition of employment. The nature of modern technological change is conceptualized on the basis of how it affects the demand for certain groups of labor, for certain occupations. “Occupation” means a limited set of tasks (work functions), the fulfillment of which in the course of the production process is imputed to the worker.
Technological change inevitably leads to shifts in the economy, which in the case of the labor market are expressed in the reconfiguration of the job structure. Jobs are heterogeneous in their characteristics: some suggest high, others — low qualification of workers; some are associated with high, others — with low pay; some have attractive, others unattractive working conditions. The nature of technological change can be defined by understanding how under its impact the job structure changes — from “good” to “bad” jobs (as Smith and Marx thought), from “bad” to “good” ones or something else. Naturally, in different historical periods the nature of the impact of technological change on the job structure might be different.
Early research that emerged in the early 1990s focused on how the computer revolution and the spread of information technologies changed demand for different skills. In the simplest variant, two groups of employees were distinguished — low-skilled (without higher education) and high-skilled (with higher education). Experience shows that modern computer technologies are closely related to the process of accumulating human capital, since their introduction and use requires skilled workers with high formal education (
Indeed, as shown by the very first work, where the concept of the SBTC was put forward, in the United States during the 1970–1980s the number of high-skilled workers grew at a rate of 3% per year, while the relative quantity and relative wages of low-skilled workers steadily declined (
However, the range of occupational duties can vary greatly even among workers belonging to the same qualification group. This circumstance was not taken into account in the SBTC conception, but became the starting point for an alternative conception of “routine-biased technological change” (RBTC) (
The routine occupations imply work operations, characterized by a pre-assigned, monotonous, repetitive nature. On the one hand, such operations require following a strictly defined protocol, so that they are easily codified and programmed using modern ICT. On the other hand, routine work is most characteristic for occupations located on the middle layers of the skill ladder (bank tellers, office clerks, accountants, etc.). At the same time, many occupations that do not require specialized qualifications (waiters, nurses, etc.) are difficult to computerize, because quick reactions, the ability to cultivate personal contacts with customers, etc. are necessary. Occupations that are at the top of the skill ladder (managers, specialists, etc.) are even less amenable to codification and programming since they require the ability to solve complex problems as well as intuition, creativity and gift of persuasion, etc. As a result, modern technologies act as complementary with respect to high-skilled and as neutral with respect to low-skilled labor, but as substitutes with respect to medium-skilled labor.
Three main predictions stem from the RBTC conception. First, in the general array of tasks solved by workers, we should expect a shift from routine operations (both physical and intellectual) to non-routine ones, since the former are increasingly taking over from machines. Empirical evidence of a declining significance of routine tasks was obtained for Anglo-Saxon countries, countries of continental Europe and Japan (
Here, however, it is worth noting that different technologies may not equally affect different groups of routine occupations. For example, in contrast to computerization which for the most part contributes to the extinction of routine intellectual forms of work (such as office workers), robotization contributes to the extinction of routine physical forms of work (such as machine operators). Accordingly, the changes they generate in the job structure will not necessarily coincide. Thus, in a number of studies, it is argued that robotization, unlike computerization, leads to a decrease in demand for low-skilled, and an increase in demand, for high-skilled labor, but without a drop in demand for workers of medium qualification, as suggested by the canonical version of the RBTC conception (
As can be seen from this brief review, economists have long been interested almost exclusively in how technological progress is associated with changes in the occupational composition of employment. It is only recently that they have addressed the equally important question of how changes can affect the general level of employment.
Of this new series of works, the study of two British economists K. Frey and M. Osborne, who presented a forecast of changes in the occupation composition of the American workforce, attracted the greatest attention (
These quantitative indicators are so colossal that they could not but cause a shock to politicians and the general public. Therefore, it probably makes sense to clarify the methodology by which they were obtained.
In their calculations,
However, the quantitative estimates of Frey and Osborne, as well as all those who used their methodology, look quite surreal. As
Here are some more examples of this kind. The emerging ATM techno-alarmists made predictions about the complete disappearance of the labor market for bank tellers. But in reality, their number (in full-time equivalent) has increased in the United States from 400,000 in 1990 to 450,000 at present — and this is against the background of an increase in the number of ATMs from 100,000 to 425,000 (
At the methodological level, the main flaw in Frey and Osborne’s calculations was revealed by
As
Explaining the reasons for the observed cross-country variation,
But even their own assessments are considered by
The most general results of our analysis can be summarized as follows: in the long run, the reduction in the labor demand under the impact of new technologies is merely a pure theoretical possibility that has never before been realized in practice; at the level of individual firms, there is a strong positive relationship between technological innovations and employment growth; at the sectoral level, technological changes cause a multidirectional employment response, since different industries are at different stages of the life cycle; at the macro level, technological progress acts as a positive or neutral, but not a negative factor; the question of the impact of robotization on employment still remains open, different researchers come to different conclusions; a surge in technological unemployment, even in the short-term, seems highly unlikely, since in the coming decades the pace of technological change will not be sufficiently high by historical standards; the impact of new technologies on labor supply may be a more serious problem than their impact on labor demand; technological changes have a much stronger effect on the composition of employment than on its level; as for occupations, current technological progress is associated not so much with changes in their nomenclature, but with changes to their internal content, namely, with the crowding out of routine tasks by non-routine ones within existing occupations.
As for predictions of technological unemployment, these arguments can hardly be taken seriously because they ignore the fundamental fact that humanity lives, and will inevitably continue to live, in conditions of scarcity. But, as