60
and away from other categories. At the same time skills like “agile, innovative and tech skills”, the set of skills that are probably most commonly regarded as the jobs of the future, are decreasing in demand, but increasing more than the other categories in pay.
The income index shows how income is related to each of the three categories in the 2013 data, and the category Interpersonal skills is here set to 100. As can be seen from this, even though “Agile, innovative and tech skills” are the best compensated and fastest rising category in terms of pay, it is shrinking in terms of demand. This seems to support the notion that the jobs and skills required by those who build automation are themselves being automated faster than they are emerging.
61
is not what we have seen. We have rather seen stable wages for three decades and a hollowing out of the middle class as the income curve becomes more extreme. Stable wages in a growing economy of course means that wages as a percent of GDP has been falling. And this is exactly what you would expect to see if certain downward trending skill sets pushed workers into using other less demanded skill sets.
At the same time we have seen the rich end of the scale getting richer. This coincides well with the idea that the “agile tech” skills are becoming ever more profitable, for an ever shrinking group of people. This could be because technological tools are empowering this group, making them more productive. At the same time the workers with maintenance and operational skills have been made redundant. Even though Murnane and Levy (2004) made some assumptions that were later shown to be wrong, this is what they predicted with their digital divide argument (recall the discussion about Murnane and Levy in the literature review).
This explanation implies that the demand for “Interpersonel Skills” is more elastic than the other two categories and is therefore able to absorb excess work from the other two categories. This move towards interpersonal skills also makes sense from a technological perspective, since interpersonal tasks are very hard to automate with current technology.
Coincidentally, the three selected categories are somewhat aligned with the four sector models as discussed in the literature review. Operations and maintenance skills could be seen as a proxy for the production / secondary sector, interpersonal skills as a proxy for the service / tertiary sector and agile, innovative and tech skills for the quaternary / innovative service sector.
Looked at this way, the data resembles very well the predictions made in Clark’s sector model as seen in figure 4. This is of course with the interesting exception of how the quaternary industry does not observe the predicted trend in terms of hours needed, but only in terms of pay.
62
15 Further extrapolating these trends
As mentioned in the statistical methodology chapter, the predicted skill importance changes (seen in table 3) are used together with the skills as required by each occupation in 2013 to calculate an overall importance change prediction for each of these
occupations. This trend is used to make a prediction of how important each occupation should have been 10 years earlier, in 2003. This predicted change was then correlated with actual change, both in terms of employment numbers and wage. The result of this
correlation analysis can be seen in table 1 in the statistical methodology chapter, where a more in-depth explanation can also be found.
Somewhat surprisingly, the correlation, or degree of explanation between skills and employment was very high at 25%, based on hundreds of data points (the 473
professions). The correlation between skills and wages were a more moderate 10%. Even though the skill trends were able to explain 25% of the observed changes in employment, it still leaves 75% to be accounted for by other factors.
From this data we will take the opportunity to look at what the data suggest about the future. Here it is important to keep in mind that the skill trends were compiled using data looking back 14 years, so while that gives us a backward looking correlation of 25%, our forward prediction will likely be even less than that. Nonetheless, our forward looking degree of explanations should statistically speaking produce better approximations than what you would get without them. Let us examine how well the predictions correspond with our expectations.
Let us first examine what the data suggest will happen to the ten largest US occupations.
Interestingly, for reasons mentioned in previous chapters, both “cashiers” and “food preparation and serving workers” are work fields that may well be expected to decline in the near future. It is unclear why we should expect to see such a strong decline in “janitors and cleaners”. On the other hand, “bookkeeping, accounting and auditing clerks” is
something you would definitely expect to see continue declining, as it has been doing for a few years now. Also, “retail sales persons”, “sales representatives” and “managers of
63
sales workers” are areas where the sort of interpersonal skills that the data suggest is becoming more important will definitely benefit.
Table 5 - Predicted change by 2023 in the 10 largest occupations.
Retail sales persons 4.67%
Cashiers 2.39%
Combined food preparation and serving workers, including fastfood 2.89%
Office clerks, general 2.82%
Waiters and waitresses 4.39%
Secretaries, except legal, medical, and executive 2.13%
Janitors and cleaners, except maids and housekeeping cleaners - 11.92%
Bookkeeping, accounting, and auditing clerks -5.49%
Sales representatives, wholesale and manufacturing, except technical and scientific products 6.12%
First-line supervisors/managers of retail sales workers 4.41%
Reviewing the occupations that are predicted to grow one finds that the selection bias of only looking at skills becomes very apparent. The data suggest that the funeral attendant industry will grow a whopping 12.55%. While the rising trend for interpersonal skill related skills can plausibly relate to an increased need for social workers, healthcare workers and so on, the need for funeral attendants likely has much more to do with demographics. This is an example where the unexplained 75% of the model is clearly visible. Now the fact that the US “baby boomer” generation is actually soon turning 60 is beside the point (the data would have showed the same had this not been the case). The same can be said for the clergy, while there might be an increased need for interpersonal skills, a much more important trend for this profession is that religion in the US is on a decline. Other than those sectors, the rest of the predictions made below seem a lot more probable.
64
The occupations with the highest predicted decline are all production related and fit well with the overall trend of declining manufacturing. The occupations here are (like those above) all selected because they are outliers, so the actual predicted decline might seem a bit excessive, yet still plausible.
Table 6 - Predicted change by 2023 in the 10 fastest growing occupations.
Funeral attendants 12.55%
Bailiffs 11.75%
Models 11.55%
Clergy 10.43%
Child, family, and school social workers 10.31%
Probation officers and correctional treatment specialists 10.11%
Medical and public health social workers 9.62%
Residential advisors 9.07%
Mental health and substance abuse social workers 9.05%
Lawyers 9.03%
Table 7 - Predicted change by 2023 for the 10 fastest declining occupations.
Pattern makers, wood -15.63%
Grinding and polishing workers, hand -15.65%
Tool and die makers -15.81%
Brick masons and block masons -15.93%
Tile and marble setters -15.98%
Stone masons -16.73%
Fence erectors -17.58%
Cabinet makers and bench carpenters -18.44%
Milling and planing machine setters, operators, and tenders, metal and plastic -18.47%
Lathe and turning machine tool setters, operators, and tenders, metal and plastic -18.60%
65
So somewhat unsurprisingly, the data predict production type jobs to keep decreasing while service type jobs are getting ever more in demand. But of course a prediction made from the recent past will always tend to make predictions in line with the trend we
currently observe.
16 Conclusion
As it sometimes is with exploratory research into poorly understood phenomena, nothing can be proved beyond any doubt. Instead certain evidence has been presented and
analyzed and from this evidence a probable interpretation has been built.
The case that jobless growth can cause problems for a society has been made. These problems can potentially manifest themselves in either or both social or economic terms.
Economic problems can arise in a society even when technology is making manufacturing processes more efficient, since jobless growth can create inequality and lower demand.
Furthermore, the case has been made that a certain type of technology is starting to replace workers and stand to further replace workers at a potentially large scale. The specific technology in question here is information technology, which has been discussed in terms of its electronics and information processing parts. Also, it has been examined how output has become decoupled from labor, how labor is receiving less and less of total income and how labor contributes less and less to output as signs of how things might be different now.
Additionally, through extensive statistical research some interesting trends in how the demand for different skills on the job market is changing has been brought to light. The
“operation and maintenance skills” are less and less in demand, which corresponds with the observed decline of production work. At the same time, with increasingly
sophisticated tools and an expanding global reach, an ever decreasing number of the highly skilled tech elite can grab an ever increasing relative portion of total output. And finally it seems likely that the decreasing demand for labor in those other skills is pushing labor to increasingly involve more "interpersonal skills”, which by its nature is unlikely to
66
become automated in the near future. This overall trend could also be a contributing factor to wage stagnation and the increasingly extreme income disparity.
17 Perspective
When correlating the best fitting skill change slopes with jobs, a 25% correlation with actual employment data was found. But if a job market survey with the primary objective to track these skills were to be conducted, you would likely be able to get a higher
correlation between work and skills, as the skills chosen could be reevaluated to suit this purpose. Additionally, directly tracking the time spent on each task done at work, (for example “information searching”, “physical work” etc.) would perhaps be a better
approach, as it arguably measures the effect of changing work patterns more directly than asking workers about the importance of each skill.
What has been examined in this paper are the skill changes in the US during the last 14 years. It could be useful to conduct a study that probes how the nature of work has changed over a longer period of time (using another data source of course, since O*NET does not go back further). Additionally, the US was used as a proxy for any
technologically developed country since this thesis set out to assess the future of work in developed countries. But my study is of course very much “US biased”, and it seems to me that at least the nature of work in for example Japan could be very different to that of the US. Therefore, a broader study, including more countries, might bring additional interesting insights.
Additional potentially informative research, arguably based in another field of study, may examine from a technical perspective how much work can actually be automated in the long run. It seems reasonable for example that the price of any robot or product should over time decrease towards the price of its metal or other material constituents. Research into this should probably be done by a computer scientist or similar, but perhaps one could break down the different tasks that make up work into specific categories, and then
examine if these can theoretically be cost competitively automated by technology in the future.
67
Also, this study has only examined the nature of technology’s effect on work. Many interesting suggestions have been proposed for solutions or adaptations that we can work towards as a society, but in my opinion, more rigorous and long term thinking and
research is needed. Apparently, citizens of ancient Rome enjoyed what is by modern standards an extremely larger number of holidays, peaking during the fifth century AD with 200 holidays pr. year and plenty of these holidays including public games or events (Kamm & Graham, 2014). It seems the Romans enjoyed their time with sports, arts and social events, and were happy enough not to work too much. Perhaps then, the end of work is not as bad a phenomenon as it has been made out to be and we should perhaps spend more time researching how to engineer new technology and how to bring about the policies, and social changes without at any point disrupting our way of life too chaotically or abruptly.
68
18 Bibliography
Acemoglu, D. (1999). Patterns of Skill Premia. National Bureau of Economic Research.
Acemoglu, D., & Robinson, J. (2012). Why nations fail: the origins of power, prosperity, and poverty.
Aghion, P., & Howitt, P. (1997). Endogenous Growth Theory. MIT Press.
Albus, J. S. (1976). Peoples' Capitalism: The Economics of the Robot Revolution.
Beckett, J. (2015). Luddites. Retrieved from Luddites:
http://www.nottsheritagegateway.org.uk/people/luddites.htm Blanchard, O. (1997). The Midium Run.
Brain, M. (. (2003). Robotic Nation.
Brain, M., & Ford, M. (n.d.). The Lights in the Tunnel: Automation, Accelerating Technology and the Economy of the Future. 2009.
Brenner, M. H. (1979). Influence of the Social Environment on Psychology: The Historical Perspective.
Brynjolfsson, E. (2013). Yes, Robots Are Coming for Our Jobs--Now What? Hentet fra Scientifica American: http://www.scientificamerican.com/article/yes-robots-are-coming-for-our-jobs-now-what/
Brynjolfsson, E., & Mcafee, A. (2014). The Second Machine Age.
Campbell, A., & Warner, M. (1992 ). New technology, skills and management, human resources in the market economy.
Carlsson, B., Taymaz, E., & Tryggestad, K. (n.d.). The economic impact of factory automation. Technological systems and economic performance.
Clark, C. (1940). Conditions of Economic Progress.
Clark, G. (2002). The Agricultural Revolution and the Industrial Revolution.
Clark, G. (2004). The Condition of the Working-Class in England, 1209-2003.
69
Cowan, R. S. (1997). A Social History of American Technology. Oxford University Press.
Cowen, T. (2011). The great stagnation.
Davenport, T. H. (1993). Process Innovation: Reengineering Work Through Information Technology.
De Witte, M., & Steijn, B. (2000). Automation, job content, and underemployment. Work, employment & society.
Detroit National Park Service. (2015). Retrieved from Detroit National Park Service:
http://www.nps.gov/nr/travel/detroit/d32.htm
Fisher, A. G. (1939). Production, primary, secondary and tertiary.
Frey, C. B., & Osborne, M. A. (2013). The Future of Employment: How Susceptible are Jobs to Computerisation?
Friedmann, G. (1992). The Anatomy of Work: Labor, Leisure, and the Implications of Automation.
Gavin, W. T. (2013). The Mechanics Behind Manufacturing Job Losses.
Google. (2014). Google Blog. Retrieved from Google Blog:
http://googleblog.blogspot.dk/2014/04/the-latest-chapter-for-self-driving-car.html Granter, E. (2009). Rethinking Classical Sociology : Critical Social Theory and the End of
Work.
Grey, C. (2014). Humans need not apply. Retrieved from Humans need not apply:
https://www.youtube.com/watch?v=7Pq-S557XQU#t=455
Grint, K., & Woolgar, S. (n.d.). The Machine at Work: Technology, Work and Organization. 1997.
Hammond, J. L., & Hammond, B. B. (1912). The village labourer 1760-1832.
Harrison, J. F. (1984). The Common People: A History from the Norman Conquest to the Present.
70
Homer, J. B., & Sloan, A. P. (1982). Theories of the industrial revolution: a feedback perspective.
IBM. (2015). IBM Watson. Retrieved from IBM Watson:
http://www.ibm.com/smarterplanet/us/en/ibmwatson/ecosystem.html Jennings, K. (2015). About Ken Jennings. Retrieved from About Ken Jennings:
http://www.ken-jennings.com/about
Johnson, C., & Chartier, G. (n.d.). Markets Not Capitalism: Individualist Anarchism Against Bosses, Inequality, Corporate Power, and Structural Poverty. 2011: Minor Compositions.
Jürgens, U., Malsch, T., & Dohse, K. (2009). Breaking from Taylorism, Changing forms of work in the automobile industry.
Kaldor, N. (1957). A Model of Economic Growth. The Economic Journal.
Kamm, A., & Graham, A. (2014). The Romans: An Introduction.
Karabarbounis, L., & Neiman, B. (2012). Declining Labor Shares and the Global Rise of Corporate Saving.
Keynes, J. M. (1930). Economic Possibilities for our Grandchildren.
Lambert, R., & Andries Beziudenhout, E. W. (2009). Grounding Globalization : Labour in the Age of Insecurity. Wiley-Blackwell.
Lichtl, A., & Jones, S. (2015). GPUs to Mars. GPU Technology Conference. NVIDIA.
Retrieved from http://on-demand.gputechconf.com/gtc/2015/video/S5398.html Love, D. (2014). Business Insider. Retrieved from Business Insider:
http://www.businessinsider.com/momentum-machines-burger-robot-2014-8?IR=T Malthus, T. (1798). An Essay on the Principle of Population.
Mantoux, ,. P. (2006). The industrial revolution in the eighteenth century: An outline of the beginnings of the modern factory system in England. Taylor & Francis.
71
Markoff, J. (2010). New York Times. Retrieved from New York Times:
http://www.nytimes.com/2010/10/10/science/10google.html?_r=2
McAfee, A. (2011). Andrew McAfee's Blog. Retrieved from Andrew McAfee's Blog:
http://andrewmcafee.org/2011/01/jevons-computation-efficicency-hardware-investmen/
McAfee, A., & Brynjolfsson, E. (2011). Race Against the Machine. Digital Frontier Press.
Mokyr, J. (1990). The lever of Riches. Oxford University Press.
Montgomery, D. (1980). Workers' Control in America: Studies in the History of Work, Technology, and Labor Struggles.
Murnane, & Levy. (2004). The New Division of Labor.
Okun, A. M. (1962). Potential GDP: It's Measurement and Significance. American Statistical Association.
Piketty, T. (2014). Capital in the Twenty-First Century. The Belknap Press of Harvard University Press.
Piketty, T., & Zucman, G. (2013). Capital is back.
Rifkin, J. (1995). The End of Work. Tarcher/Putnam.
Saunders, M., Lewis, P., & Thornhill, A. (2009). Research methods for business students.
Schimmack, U., Schupp, J., & Wagner, G. G. (2007). The Influence of Environment and Personality on the Affective and Cognitive Component of Subjective Well-being.
Springer Science Business Media.
Schneider, C. J., & Schneider, D. (2003). World War II.
Shaiken, H. (1984). Work Transformed: Automation and Labor in the Computer Age.
Holt, Rinehart & Winston.
Skills, Tasks and Technologies: Implications for Employment and Earnings. (2011). In D.
Acemoglu, & D. Autor, Handbook of Labor Economics,.
72
Spence, M. (2014). Labor’s Digital Displacement. Project Syndicate.
Spinney, L. (2012). Busted! The myth of technological progress. NewScientist.
Steininger, M., & Rotte, R. (2009). Crime, unemployment, and xenophobia?
Sturgess, R. W. (1966). The Agricultural Revolution on the English Clays.
Tabarrok, A. (2003). Productivity and unemployment.
Tamura, E. (1997). China: Understanding Its Past, Volume 1.
Texas A&M Transport Institute. (2013). Texas A&M Transport Institute. Retrieved from Texas A&M Transport Institute: http://tti.tamu.edu/2013/06/13/simulated-crash-testing-saves-time-and-money/
The Economic History Association. (2015). The Economic History Association. Retrieved from The Economic History Association: http://eh.net/encyclopedia/hours-of-work-in-u-s-history/
The Occupational Information Network. (2015). The Occupational Information Network.
Retrieved from The Occupational Information Network:
http://www.onetcenter.org/dataCollection.html
Thibodeau, P. (2014). Computerworld. Retrieved from Computerworld:
http://www.computerworld.com/article/2837810/automation-arrives-at-restaurants-but-dont-blame-rising-minimum-wages.html
Tonna, C. E. (1841). Helen Fleetwood.
UK Office for National Statistics. (2011). UK Office for National Statistics. Retrieved from UK Office for National Statistics: http://www.ons.gov.uk/ons/publications/re-reference-tables.html?edition=tcm%3A77-309799
Us Bureau of Labor Statistics. (2015).
Wade, P. (2014). Fast Company. Retrieved from Fast Company:
http://www.fastcompany.com/3035490/fast-feed/elon-musk-tesla-cars-could-run-on-full-autopilot-in-5-years
73
Warde, P. (2007). Energy Consumption in England and Wales.
74