Technological Unemployment

A review of the issues associated with prospective technological unemployment. This includes the outlook for universal income or guaranteed income funded by robot taxes. It also covers the U.S. fiscal capacity to undertake such a scheme.

1. Technological Unemployment Gaetan Lion March 2017 1

2. Historical Perspective 2

3. The Prophets of Technological Unemployment 3 Keynes anticipated its occurrence back in 1930. Marx anticipated the related rise in Populism [Proletariat Revolution] in 1867.

4. The Technology Conundrum: Technological Unemployment Historically, technological innovation has caused rising labor productivity and living standards. Prospectively, it may cause massive unemployment. John Maynard Keynes was well aware of the problem back in 1930. “We are being afflicted with a new disease of which some readers may not yet have heard the name, but of which they will hear a great deal in the years to come – namely, technological unemployment . This means unemployment due to our discovery of means of economizing the use of labor outrunning the pace at which we can find new uses for labor.” John Maynard Keynes “Economic Possibilities for our Grandchildren.” 1930 4

5. Karl Marx: “Das Kapital” Part I. 1867 Back in 1867, within “Das Kapital”, chapter 15: ‘Machinery and Large-Scale Industry’ Marx has already much to say about technological unemployment. Introducing machinery increases productivity and profit for capitalists. Machinery implements automation that enables capitalists to replace workers. Automation transfers excessive economic power to capitalists vs. workers (proletariat). Marx advanced capitalism was unsustainable and would be toppled by a proletariat revolution. 5

6. Marx advanced capitalism was unsustainable and would be toppled by a proletariat revolution 6 “… the emergence of Ms Le Pen matches a pattern of insurgent populism across Western liberal democracies . A fear of job losses due to automation…” The Economist March 4 th , 2017 . Wilders has tapped into deep fears among many low-skilled workers over their jobs in a world of rapid technological change.

7. Let’s review what happened since the days of Karl Marx and John Maynard Keynes to two formerly dominant sectors of the U.S. labor force, namely Agriculture and Manufacturing… 7

8. Agriculture : Output is growing, the sector job share is collapsing. 8

9. Agriculture: An International Phenomenon 9

10. Manufacturing : The output is growing. The sector job share is collapsing. The output per worker is growing rapidly. 10

11. Manufacturing: An International Phenomenon 11

12. Current Situation 12

13. Could the IT sector be next? A former Google cloud computing engineer stated that many IT jobs could be at risk. In the late 90s one IT person was managing 5 servers. Now one IT manager can manage 10,000 servers! That’s a 2000-fold effect in a period shorter than what it took to lose less than half of the manufacturing jobs. 13 I met a former IT staffer working at a bike shop. He stated a huge layer of the IT workforce had been eliminated in the past few years due to Cloud Computing that is far more efficient than pre-Cloud Computing systems.

14. The Second Machine Age The Second Machine Age: Work, Progress, and Prosperity in a Time of Brilliant Technologies is a 2014 book by Erik Brynjolfsson and Andrew McAfee. The “Second Machine Age” entails the digitation of complex cognitive tasks by software-driven machines that make humans superfluous. This is different from the "First Machine Age", or Industrial Revolution, which helped make labor and machines complementary. Examples of Second Machine Age-machines include "software that grades students' essays more objectively, consistently and quickly than humans" and "news articles on Forbes.com about corporate earnings previews" — "all generated by algorithms without human involvement .“ Other examples include computers beating world champions in chess, Go, and Jeopardy. 14 Computer technology (including Big Data, AI, etc.) has the potential of displacing many “cognitive” workers that were deemed non-displaceable.

15. Robots are taking over 15 In Japan and Russia where the labor forces are shrinking rapidly, robots could be an economic life saver. But, in the U.S. with more favorable demographics, robots could cause technological unemployment.

16. “The Great Decoupling” by the authors of “The Second Machine Age” Note accelerating of decoupling after 2000. 16

17. Focusing on “Decoupling” of Real GDP per Capita vs. Real Wages & Earnings 17

18. Focusing on “Decoupling” of US Labor Productivity vs. US Private Employment Notice strong decoupling since 2000. 18

19. Decoupling of Productivity vs. Hourly Compensation. See 1973 “Decoupling” Point. 19

20. See accelerating trend after 2000. That’s behind the Populist-Proletariat Revolution. Decoupling of Corporate Profits vs. Wages 20

21. Decoupling by Cognitive vs. Manual Jobs See the decoupling starting in mid 1980s. That’s when implementation of desktop software accelerated. It rendered routine cognitive jobs superfluous. 21

22. About 50% of jobs could be replaced by automation 22

23. When this paper came out in 2013, it was received with skepticism. Just over three years later with rapid progress in robotics and artificial intelligence, the paper is now viewed as being realistic. A McKinsey report of January 2017 confirms their findings. 23

24. The probability of job computerization by job categories Those are the 47% of jobs that are at high risk of being computerized (prob. > 0.70) 24 Source: “The Future of Employment” paper. 2013.

25. Job computerization probability: Engineers & Quants Engineers Quants 25 Large difference in probability of job computerization between different engineering specialties. Who needs economists? Nate Silver would agree. Source: “The Future of Employment” paper. 2013.

26. Job computerization probability: Banking/Finance & HR Banking/Finance HR 26 Many jobs in Banking/Finance associated with very high probability of computerization. Source: “The Future of Employment” paper. 2013.

27. Job computerization probability: IT & Legal & Other IT Legal Other 27 Professional sports without umpires and maybe even without athletes?! How come travel agents still exist? Remember comments by Google cloud computing engineer and former IT worker in a bike shop. Las Vegas without dealers?! Hollywood without actors?! Still need lawyers but many jobs in legal field to be computerized Source: “The Future of Employment” paper. 2013.

28. 28 Job computerization probability: Education & Healthcare Education Healthcare Except for teacher assistant, we still will need educators. Except for dental hygienists, we still will need “personal” healthcare. Source: “The Future of Employment” paper. 2013.

29. 29 50% of jobs done by humans today are vulnerable to replacement by robots. That could amount to a loss of $15 trillion in wages worldwide and $2.7 trillion in the U.S. This may fully occur by 2055 + or – 20 years. Thus, this level of tech unemployment could hit as soon as 2035. McKinsey’s report January 2017.

30. 30

31. Surviving Automation: People Skills 31 Many functions associated with people skills are less likely to be replaced by automation. Imagine a robot-manager, a robot-lobbyist, or a robot-PR executive. Not good!. A human has a marked edge in those “hi-touch” fields.

32. Surviving Automation: Quantitative skills 32 Whatever can be replaced by an algorithm will be. Today’s “expert” human cognitive functions can become tomorrow’s automated ones . Remember the former IT guy who felt his job was safe; who now works in a bike shop because of the impact of Cloud Computing.

33. Expertise is a lot more vulnerable than we think This is not just about sophisticated robots and complex Machine Learning models. It is also about plain software, and simple methods of aggregating and replacing human expertise. 33

34. 34 “There are more monthly visits to the WebMD network, a collection of health websites, than to all the doctors in the United States. Annually , in the world of disputes, 60 million disagreements among eBay traders are resolved using “online dispute resolution” rather than lawyers and judges — this is three times the number of lawsuits filed each year in the entire U.S. court system. The U.S. tax authorities in 2014 received electronic tax returns from almost 50 million people who had relied on online tax-preparation software rather than human tax professionals .”

35. E-Filing is taking over Tax filing 35 Many tax e-filings are prepared directly by the taxpayer. This must have a material impact on that accounting tax preparing profession.

36. Individual Income Tax filing. IRS disclosure. 2014 36 Focusing on individual returns, e-filing accounted for 89.7% of total returns filed in 2013 and 90.1% in 2014. Of those e-filings, tax professionals filed 60.8% of those in 2013, and 59.2% in 2014. Based on the depicted yearly trend, tax professionals will file less than half of such e-filings within less than 6 years.

37. 37 Not only AI is faster. It is also more accurate. 89% accuracy vs. 73% for doctors in study. How many expert-type-jobs in various industries will be displaced by AI?

38. 38 Indexing has taken a huge market share out of overall Wall Street equity investment management industry Over the past decade equities invested in index funds have risen from $868 billion to $4.0 trillions. That should have eliminated a lot of high paying jobs such as investment advisors, portfolio managers, and security analysts

39. Among Retail Financial Investment Providers market shares are shifting to the Indexers 39 Source: Morningstar As shown over a recent 10 year period, Vanguard and iShares focused on index funds and ETFs have gained much market shares from Fidelity and American Funds that are focused on active investment management. This trend has dire implications for the mutual fund managers and security analysts at Fidelity and American Funds.

40. 40 US online travel booking has grown from very little in the late 90s to over a 40% market share of the travel booking business by 2015. This must have had a material impact on travel agents. The online travel booking business is growing very rapidly overseas (actual market share of respective travel booking business unknown).

41. 41 When did you last buy a book at a brick-and-mortar bookseller? In any given year how many books do you buy online vs. buy at a retail store? The trends depicted on the graph must have much reduced bookselling employment.

42. Solution to Technological Unemployment: Labor Union 42

43. A very old problem 43 Henry Ford II Henry Ford II: Walter, how are you going to get those robots to pay union dues? Walter Reuther: Henry, how are you going to get them to buy your cars? 1955 Walter Reuther Robots = No car sales Labor Union = No Robots = Car sales

44. A Return to Labor Unions as a Solution? 44 The conversation between Henry Ford II and Walter Reuther took place near the peak of Labor Union membership. You can observe a logical and very strong negative relationship between Union membership and share of income going to the top 10%. Some economists have recommended a return to rising Labor Union membership including Robert Reich in “Savings Capitalism” (2016). However, it remains unclear how such a revival would be possible given international supply chains. The latter have much reduced the influence of labor unions as depicted by the graph.

45. Union Membership Private vs. Public 45 Although, those trends are very interesting they have little implication on Technological Unemployment except for the declining influence of private unions (that will not buffer the impact of Technological Unemployment). The disparity between the private and public sector trends in union membership is striking. As private union membership has steadily dwindled (Globalization, automation, etc.) since 1950, public one has risen. And, the rise of public union membership has been especially rapid between 1960 (just over 10%) and 1980 (around 40%).

46. Other Solution to Technological Unemployment: Guaranteed Income and Robot Tax 46 In this section we explore the feasibility of a Guaranteed Income for all, also called Universal Income. Such Income would be on an unprecedented level (a multiple larger than Social Security), as it would replace a material portion of Consumer Spending. Robot tax is considered as a solution to partly finance Guaranteed Income and to maintain other Government services currently financed by payroll taxes and personal income taxes.

47. Tech Unemployment Solution: Guaranteed Income for all It is not deemed too socialist. Everyone talks about it including Silicon Valley. To review how feasible it is, let’s step back and review how a basic economy works. 47

48. A Basic Economy 48 The business sector hires employees (pays wages). The employees are also the consumers who purchase the goods and services from the business sector. Both businesses and household/consumers pay taxes to the Government. And, the latter incurs numerous Government expenditures (National Defense, Social Entitlements, Discretionary spending). And, the world goes round just fine.

49. Economy with very high Technological Unemployment & Guaranteed Income collapses 49 Households do not earn much wages as they are replaced by artificial intelligent robots (AIRs). So, households can’t buy the goods produced by the AIRs. Businesses fail. And, no one pays any taxes. The Government fails too in the absence of adequate tax receipts.

50. 50 The Solution!

51. The Robot Tax 51 Bill Gates idea of a Robot tax is daring. But following Walter Reuther’s logic, robots are not consumers. Without Consumer Spending, you have nothing, not even a Robot tax.

52. How to make it work? Overall framework : Robot Tax = Guaranteed Income = Consumer Spending = Sales Price From HH/Consumer standpoint : Guaranteed Income = Consumer Spending = Sales Price From Business standpoint : Cost of Robots to be less than cost of Labor. Otherwise, there is no Robot. 52 =< Wages + Benefits + Payroll taxes Capital Investment in Robots + Operating expenses of Robots + Robot Tax

53. Robot Tax Conundrum 53 Wages + Benefits + Payroll taxes - Capital Investment in Robots - Operating expense in Robots Robot Tax Guaranteed Income Consumer Spending = Sales Price =< = = The Robot Tax has to be simultaneously equal to the Sales Price and less than the difference between using a human labor force and using Robots. Are those relationships simultaneously even possible, or mutually exclusive?

54. Robot Tax Summary 54 Difference between cost of Labor Force and Robots Sales Price = Unclear how Business could stay in business given the above equality. In view of the above, the Federal Government would probably have to do much of the heavy lifting in terms of funding a Guaranteed Income (with a Robot tax playing only a minor role). Next, we will explore what the prospective fiscal costs of Guaranteed Income would be. And, how feasible it would be.

55. The Fiscal Cost of Guaranteed Income 55

56. Guaranteed Income vs. Consumer Spending 56 Even when Guaranteed Income replaces as little as 15% of Consumer Spending, it translates into fiscal costs in the $trillions. Source BEA for Consumer Spending annual level as of 2016Q4 The red area outlines scenarios where Guaranteed Income represents 15% or more of Consumer Spending.

57. Guaranteed Income vs. Social Security 57 Source CBO for Social Security in 2016 Even when Guaranteed Income replaces as little as 15% of Consumer Spending, it costs more than twice as much as Social Security.

58. Guaranteed Income, Consumer Spending, and Federal Budget 58 Source CBO for Federal Budget in 2016 Even when Guaranteed Income replaces as little as 15% of Consumer Spending, it represents more than 50% of the Federal Budget.

59. U.S. Fiscal Perspective Even though earlier slides are already discouraging, let’s explore how much prospective fiscal capacity the U.S. has in adding an additional major social entitlement program such as a Guaranteed Income. 59

60. Fiscal Constraints 60 Source: BEA, CBO In 2016, Consumer Spending at $13 trillion is enormous relative to the scale of the entire Federal Budget of $4 trillion. Consumer Spending is also over 13 times larger than Social Security (< $1 trillion), the largest social entitlement program. Given the scale of Consumer Spending, it is most challenging for the Federal Government to implement a Guaranteed Income that could replace a material portion of Consumer Spending.

61. CBO Baseline Long-Term Outlook depicts an unsustainable fiscal position 61 Gov. spending far outpaces tax receipts. Resulting Budget Deficits rise from 3% of GDP in 2017 to 9% in 2047. Rising Budget Deficits cause the Debt/GDP ratio to rise from 77% in 2017 to 145% in 2047.

62. Social Entitlements are pushing up Budget Deficits 62 The Deficits associated with SS and Medicare increase from about – 2% of GDP in 2017 to – 6% of GDP in average over the 2038-2047 period.

63. U.S., and many countries, are in no position to implement a Guaranteed Income 63 The majority of advanced economies are in a similar situation with current Debt-to-GDP ratios close to or higher than U.S. Other advanced economies have demographics associated with more rapid population aging that renders long term fiscal outlook weaker than U.S.

64. Fiscal Perspective Conclusion The Government has known for decades that the U.S. fiscal position (Debt/GDP) is heading towards unsustainability due to rapid rise in social entitlements. The past several Administrations, regardless of parties, have paid little attention to this issue. The current one is no different. Now, society is considering a Guaranteed Income given prospective massive Technological Unemployment. Guaranteed Income would easily cost a multiple of Social Security. Given that Social Security and Medicare are already causing an unsustainable rise in Debt/GDP ratio, Guaranteed Income considerations do not appear realistic. 64

65. Will we experience Technological Unemployment as a result of Technological Disruption? 65 We have already experienced massive historical and current Technological Disruption as reviewed*. However, Technological Unemployment has not left much of a footprint in the economic data. The U.S. economy has suffered bouts of high unemployment rate (Great Depression, Oil Shock recessions, Great Recession). Spain, Greece, and South Africa have all currently suffered Great Recession-like unemployment rate levels for nearly a decade. But, none of those very high unemployment rate periods have been associated with much Technological Unemployment. Thus, Technological Disruption does not necessarily equate to Technological Unemployment. *Remember the rapidly declining labor market share of Agriculture and Manufacturing shown earlier. As indicated the same is happening in white collar professions ranging from certain IT sectors (Cloud Computing), booksellers (Amazon), tax accountants (Turbo Tax), travel agents (Expedia, etc.), investment managers & security analysts (indexing), and many other fields.

66. Back to the McKinsey Report 66 McKinsey indicated that over the next 20 to 60 years 50% of current jobs could be automated. Their mean expected outcome being around 40 years from now. Given the above speculation, the actual rate of change is a more interesting consideration than the overall change (-50% reduction in current jobs due to automation).

67. What is the % of new jobs created yearly necessary to neutralize a – 50% decrease in jobs due to automation over several decades? 67 To neutralize the mentioned -50% jobs contraction due to automation over a 20 year period, the yearly creation of new jobs is 3.5%; over a 40 year period it is 1.7%; and over a 60 year period it is 1.2%. As shown, the time horizon has a huge influence on the necessary yearly new job creation rate.

68. Job Turnover in the US is very high 68 In any single month 3% to 4% of jobs are eliminated (separations that include voluntary and involuntary leaving jobs) and recreated (job openings). Given that, the U.S economy can easily create 1.2% to 3.5% new jobs per year to replace existing jobs lost. See below monthly rates of Job openings and Separations for the overall U.S. economy (nonfarm sector) and for the Information sector*. *Per BLS, this includes publishing, software publishing, motion picture, sound recording, broadcasting, telecommunication, web search portals, data processing, and information services industries. Source: BLS Source: BLS

69. Over decades the composition of the Labor force can change drastically 69 From 1900 to 1950, % of labor force engaged in farm jobs dropped from 40% to just over 10%. From 1973 to 2012, % of labor engaged in manufacturing dropped from 25% down to under 10%. The US labor force adapted well to wrenching changes as it moved progressively over time from the agriculture sector onto manufacturing and finally onto services.

70. The concern over Technological Unemployment: What’s next? 70 Agriculture Manufacturing Information Services Few companies dominating Cloud Computing, e-Commerce, and everything else. Dystopian Future

71. Labor Force is already adapting to a “dystopian” Present 71 Cloud computing and e-commerce is already dominated by just a few companies (Amazon, Google, etc.). Many industries have been materially disrupted (book publishing, music industry, travel industry, TV broadcasting). As mentioned software and aggregation mechanisms such as online tax filing and investment indexing have already taken a huge share of related professional activities (tax accountants, portfolio managers, etc.). And, employment is still growing.

72. Can you find the footprint of Technological Unemployment within our Dystopian Present? 72 The only recent major correction in employment growth was during the Great Recession associated with the Subprime Crisis that had little to do with Technological Unemployment. Great Recession

73. How about the “Decoupling” since 1970s 73 As reviewed, there has been a “decoupling” between Real GDP growth (even on per capita basis) and wages since the early 70s. Much of that trend may be due to the rising labor supply as women labor force participation rate skyrocketed from 50% in 1972 to 77.5% by 1998.

74. How about the “Decoupling” since 2000 74 The decoupling since 2000 is not explainable due to women labor force participation rate as the latter has remained stable over that period. This more recent decoupling may be due to the impact of technological disruption. And, that is the major concern of the authors of “The Second Machine.”

75. Reframing the Technological Unemployment issue Considerations of Guaranteed Income substituting for a material share of Consumer Spending given existing fiscal constraints do not appear realistic. The concern may not be massive unemployment. If 50% of today’s jobs are eliminated over several decades, our economy may well create the new jobs necessary to sustain growing employment. It has done so in the past with shifts away from agriculture and manufacturing. Whether 50% of today’s jobs are eliminated is not as important a consideration vs. how rapidly such jobs would be eliminated. If they are eliminated over 20 years, the stress on employment trends will be twice as great as if over 40 years. Our inability to foresee future employment sectors is not to be confused with inevitable unemployment related to the progressive elimination of current employment sectors. The “decoupling” trend since 2000 is associated with rising labor productivity with companies (rising profits) capturing more of the economic gains than employees (stagnant wages). This does not preclude employment from rising. However, for economic power to partly shift back to employees you need prospective labor shortages. Only a few years back, the concern was of emerging labor shortage associated with the upcoming retiring of the Baby Boomers. Now, it is no more a concern. Things change… 75

76. 76 Summary of evolution of a concept over time while conducting this research 1) Ouch, we are all going to be replaced by intelligent robots. 2) Well, maybe not if we have special expertise. 3) Woops, special expertise is vulnerable too. We are going to lose our jobs after all. 4) Wait, it is not the overall job replacement that matters (-50%), it is over what time period (20, 40, or 60 years, etc.). The economy should have time to progressively create new job sectors we do not know off today. 5) In view of item 4), we won’t need to implement concepts such as Guaranteed Income and Robot Tax that were unfeasible anyway (fiscally and politically, respectively). Return of private labor unions is equally unlikely. 6) Prospect for long term employment looks better than as first expected under item “1 & 3” . However, prospect for wages looks rather stagnant. “Decoupling” likely to continue for a while… until things change…