Advanced Topics in Trade

Transcription

1 Advanced Topics in Trade Lectures Trade and Inequality Heiwai Tang - SAIS April 11-18, 2017

2 Today s Agenda Evidence on inequality in developed countries (developing countries next lecture). Can Heckscher-Ohlin model explain the inequality trends globally? Brief discussions of the recent direction of research on the intersection between trade and income inequality. Autor, Dorn, and Hanson s series of papers about the effects of China shocks on US labor market and other outcomes (using the local labor market approach). A very recent paper by Acemoglu and Restrepo (2017) on the effects of robots on labor market outcomes.

3 Reading Autor, David H., David Dorn, and Gordon H. Hanson (2016) The China Shock: Learning from Labor Market Adjustment to Large Changes in Trade Annual Review of Economics. Harrison, Ann, John McLaren, and Margaret McMillan (2011) Recent Perspectives on Trade and Inequality. Annual Review of Economics. Rodrik, Dani (1997) Has Globalization Gone Too Far? Institute for International Economics. International Monetary Fund (2016) Global Trade: What s Behind the Slowdown? World Economic Outlook

4 Long-run Global Trend ture 8

5 US Wage Inequality Trend Inequality Trend in the U.S. ( ) Female Male Source: Autor, Katz and Kearney (2008)

6 US Wage Inequality by Edu Group Male Wages of different skill groups ( ) HSD = high school dropouts; HSG = high school graduates; SMC = Some college CLG = college; GTC = graduate school Acemoglu and Autor (2010) Handbook of Labor Economics Lecture 8 10

7 US Wage Inequality by Edu Group Female Wages of different skill groups ( ) HSD = high school dropouts; HSG = high school graduates; SMC = Some college CLG = college; GTC = graduate school Acemoglu and Autor (2010) Handbook of Labor Economics Lecture 8 11

8 Hollowing of the Middle Class in the US? Changes in real wages by wage percentiles Autor and Dorn (2008)

9 Hollowing of the Middle Class? Other Advanced Economies Acemoglu and Autor (2010) Handbook of Labor Economics Lecture 8 13

10 Differential Effects across Occupations ecture 8

11 Differential Effects across Occupations Lecture 8 1

12 Differential Effects across Occupations Lecture 8

13 Hollowing of the Middle Class - by Occupation Category

14 Declining Labor Shares Neiman and Karabarbounis, QJE 2014 Lecture 8 15

15 Declining Labor Shares Distribution of Labor Share Trends Labor Share Trends, Percentage Points per 10 Years CHN GER ITA FRA CAN JPN USA GBR

16 Declining Labor Shares across US Industries

17 Neiman and Karabarbounis s Explanations: Cheaper Capital Labor Shares and Relative Price of Investment (Trend in Labor Share)/(1-Labor Share) LTU EST AZE MDA ISL KOR BLR KEN CRI THA GBR PRT COL ESP TUR MAC BOL CZE SGP BEL USA CANDNK NLD AUS ITA JPN TWN CHE SVK SWE NAM PER AUT CHN TUN GER FRA LVA NZL LUXFIN BHR ZAF NOR NGA MEX POL SVN HUN ARG BRA ARM ROM Trend in Log Relative Price of Investment UKR

18 Cumulative percent change since 1948 Productivity-Wage Gap Disconnect etween productivit and a t pical worker s compensation, % : Productivit : 96.7% Hourl 250 compensation: 91.3% : Productivit : 73.4% Hourl compensation: 11.1% Productivit Hourl compensation % 112.5% Note: Data are for average hourl compensation of production/nonsupervisor workers in the private sector and total econom. "Net productivit " is the growth of output of goods and services minus depreciation per hour wor

19 The Top 1% Lecture 8 20

20 The Sources of Top 1% s Incomes Lecture 8 21

21 Top 1% in 3 Countries Lecture 8 22

22 Top 10% in 3 Countries Lecture 8 23

23 Bottom 50% Lecture 8 24

24 Income Growth by Brackets Lecture 8 25

25 Top 1% s Wealth Lecture 8 26

26 Summary Increasing college premium... slower growth recently. Hallowing of the middle class (measured by income or occupation). Declining labor shares. The emergence of the super-rich. The bottom half was falling rapidly in their shares of GDP.

27 So What Caused these inequality trends y is demand-driven (not supplyst developed countries of orkers g a Rel. supply of college skills. rease of not a College wage premium her Acemoglu (2002) 28

28 Speculative Causes: 8 broad determinants 1. Trade with labor-abundant countries (e.g., opening up of China, India, Mexico, and Vietnam, SSA). 2. Production fragmentation, GVC. 3. Multinational firms dominance. 4. Global competition implies more uncertainty and volatility. 5. Immigration of low skilled workers to developed countries; 6. Skill-biased technological change (e.g. computerization, robotization, ICT); 7. ICT allows better wage discrimination 8. Institutional Changes (e.g. minimum wages, deunionization, etc.);

29 Can the Heckscher-Ohlin model explain the trends? In the data, we see little reallocation of workers across sectors. Instead, most of the shifts in the U.S. happened within sectors (e.g. each sector hires more managers and non-production workers). Berman, Bound and Machin (1998) find a similar pattern of within-industry skill upgrading in the manufacturing sectors of the advanced economies (OECD) in the 1970s and 1980s, even when relative wages of the skilled workers were rising sharply.

30 Foreign Sourcing Feenstra and Hanson (1999) argue that we should not focus on inter-industry trade to examine the impact of trade on inequality. Instead, they look at intra-industry trade (i.e. outsourcing) It is consistent with skill-upgrading within manufacturing sectors (because unskilled-intensive tasks are usually outsourced). Intra-industry trade does not have any direct implications on the prices of skill-intensive products. Theoretically they can decrease even with a rising wage premium for the skilled.

31 Feenstra and Hanson (1999) The impact of outsourcing and high-technology capital on wages: estimates for the United States, imported intermediates i = input purchase of good j by ind i Out narrow i Out broad i imports of good j consumption of good j = imported intermediates i total nonenergy inputs i = imported intermediates i in the same 2-digit total nonenergy inputs i

32 BLS computer share. Before applying our two-stage estimation procedure, we report in Table III regressions of the share of the total industry Feenstra and Hanson (1999) TABLE III DEPENDENT VARIABLE CHANGE IN NONPRODUCTION WAGE SHARE, (1) Mean (2) Regression (3) Regression (4) Regression (5) Contribution ln (K/Y) % (0.014) (0.014) (0.012) ln (Y) % (0.008) (0.008) (0.009) Outsourcing (narrow) % (0.169) (0.175) (0.166) Outsourcing (difference) % (0.046) (0.050) (0.054) Capital services (ex post rental prices): Computer share % (0.102) High-tech share (difference) (0.129) Capital services (ex ante rental prices): Computer share % (0.171) High-tech share % (difference) (0.072) Computer investment: Computer share % (0.007) High-tech share (ex % post rental prices) (0.051) constant % (0.042) (0.039) (0.040) R N The mean of the dependent variable equals Standard errors (in parentheses) are robust to heteroskedasticity and correlation in the errors within two-digit industries. The first column shows mean values of the dependent and independent variables for All regressions and means are computed over 447 four-digit SIC industries and are weighted by the average industry share of the manufacturing wage bill. ln (K/Y) is the average annual change in the log capital-shipments ratio, and ln (Y) is the average annual change in log real shipments. The outsourcing variables and the computer and high-technology shares are in annual changes and are defined in Table II and the text.

33 Trade-induced Skill-biased technological change We have seen in earlier lectures that international trade can induce firms innovation, R&D, technological and other X-efficiency improvements. Thus, opening to trade, due to more innovation and other efficiency enhancing activities, can lead to rising wage inequality in developed countries.

34 Declining bargaining power of works Bargaining power index (BPI) and income share of the richest 0.5% Levy and Temin (2007) Lecture 8 39

35 Krugman s conundrum, Economist 2008 the manifestation of a guilty conscience (referring to his 1995 paper) It s no longer safe to assert that trade s impact on the income distribution in wealthy countries is fairly minor. More trade is with poor countries. The growing fragmentation of production means more tasks have become tradable, increasing the universe of labor-intensive jobs. How can we quantify the actual effect of rising trade on wages? The answer, given the current state of the data, is that we can t.

36 Declining Manufacturing Employment Share in the US EC08CH08-Hanson ARI 29 September : ownloaded from auptbibliothek on 11/21/16. For personal use only. Manufacturing share of nonfarm employment Figure Manufacturing share of US nonfarm employment ( ). Source: FRED Economic Data ( research.stlouisfed.org/fred2/graph/?g = 1Gor). Year

37 Increasing Import Competition from China China s share of world manufacturing value added China s share of world manufacturing exports Percent Figure Year China s share of world manufacturing activity ( ). Source: World Development Indicators

38 8CH08-Hanson ARI 29 September :27 US-China Current Account Dynamics 10 United States current account balance China s current account balance Percent of GDP or personal use only Year Figure 5 The United States and China s current account balances [% of gross domestic product (GDP)] for

39 Industry-level Regression Results EC08CH08-Hanson ARI 29 September :27 L jt = α t + β 1 IP jt + γx j0 + e jt Table 2 Industry-level changes in Chinese import exposure and US manufacturing employment a 100 Annual in US exposure to Chinese imports 100 Annual log in employment (manufacturing industries) Mean (SD) Median Mean (SD) Mean (SD) Mean (SD) Mean (SD) 0.50 (0.94) 2.71 (3.07) (0.75) (3.85) 0.66 (1.33) 0.30 (3.49) 0.84 (1.61) 3.62 (4.15) 0.30 (1.68) 5.73 (5.02) a Statistics are based on 392 four-digit manufacturing industries. The change in US exposure to Chinese imports is computed by dividing 100 times the annualized increase in the value of US imports over the indicated period by 1991 US market volume in that industry. Employment changes are computed in the County Business Patterns. All observations are weighted by 1991 industry employment. Table adapted from table 1 in Acemoglu et al. (2016). se only. industry j over subperiod τ, β 1 is the estimated effect of exposure to import competition on industry

40 State of the literature up to end of the previous century Earlier empirical findings: Skill biased technical change is more important than trade. Global outsourcing affects demand for skill but only modestly. Trade with low income countries is too small to have major effects. Recent empirical observations: Since 1990, trade with low wage countries has grown dramatically. Low-wage country share in US imports: 3% in 1991, 12% in China accounts for 92% of this growth.

41 Autor Dorn and Hanson, AER 2013 The China Syndrome: Local Labor Market Effects of Import Competition in the United States 4/18/ _front.jpg ( ) bluray.com/movies/covers/43887_front.jpg 1/1

42 Autor Dorn and Hanson, AER 2013 Empirical proxy for CZ s import exposure: Instrument: IPW uit = j IPW oit = j L ijt L ujt M ucjt L it L jt 1 L ujt 1 M ocjt E it 1 i = commuting zone; j = industry; u=us; c=china Allocates to each CZ a share of total national import growth Divides this import value by a CZs total employment Yields measure of import growth per worker (in 1000s of USD) Two sources of variation: Variation in CZ s manufacturing industry mix Overall manufacturing employment share in CZ

43 EC08CH08-Hanson ARI 29 September :27 Exposure to China Shocks by Computing Zone a Unconditional trade exposure of CZs ( ) Annu. Rev. Econ : Downloaded from Access provided by University of Zurich - Hauptbibliothek on 11/21/16. For personal use only. Lowest quartile (least exposed) Second quartile Third quartile Highest quartile (most exposed) b Conditional trade exposure of CZs ( ) Lowest quartile (least exposed) Second quartile Third quartile Highest quartile (most exposed) Figure 6 Geographic exposure to trade shocks at the CZ (commuting zone) level. (a) Quartiles of unconditional exposure. (b) Quartiles of exposure conditional on manufacturing employment share. Trade-induced shocks to local manufacturing employment should affect the allocation of labor across sectors but should have no measurable impact on employment rates in directly impacted CZs relative to the national labor market. That this neoclassical prediction does not appear to

44 Main Results 2135 THE AMERICAN ECONOMIC REVIEW october 2013 Table 2 Imports from China and Change of Manufacturing Employment in CZs, : 2SLS Estimates Dependent variable: 10 annual change in manufacturing emp/working-age pop (in % pts) I II (pre-exposure) (1) (2) (3) (4) (5) (6) (Δ current period imports 0.89*** 0.72*** 0.75*** from China to US)/worker (0.18) (0.06) (0.07) (Δ future period imports 0.43*** from China to US)/worker (0.15) (0.13) (0.09) Notes: N = 722, except N = 1,444 in stacked first difference models of columns 3 and 6. The variable future period imports is defined as the average of the growth of a CZ s import exposure during the periods and All regressions include a constant and the models in columns 3 and 6 include a time dummy. Robust standard errors in parentheses are clustered on state. Models are weighted by start of period CZ share of national population. *** Significant at the 1 percent level. ** Significant at the 5 percent level. * Significant at the 10 percent level. error term. 26 Additionally, the vector X it contains (in most specifications) a rich set

45 substantial changes in population. The regressions in Table 4 are analogous to our earlier models for the manufacturing employment share except that our dependent By Age Group VOL. 103 NO. 6 autor et al.: the china syndrome 2142 Table 4 Imports from China and Change of Working-Age Population in CZ, : 2SLS Estimates Dependent variables: Ten-year equivalent changes in log population counts (in log pts) I. By education level II. By age group All College Noncollege Age Age Age (1) (2) (3) (4) (5) (6) Panel A. No census division dummies or other controls (Δ imports from China 1.031** ** *** to US)/worker (0.503) (0.660) (0.488) (0.826) (0.572) (0.422) R Panel B. Controlling for census division dummies (Δ imports from China to US)/worker (0.513) (0.619) (0.519) (0.953) (0.474) (0.450) R Panel C. Full controls (Δ imports from China to US)/worker (0.746) (0.685) (0.823) (1.190) (0.560) (0.651) R Notes: N = 1,444 (722 CZs two time periods). All regressions include a constant and a dummy for the period. Models in panel B and C also include census division dummies while panel C adds the full vector of control variables from column 6 of Table 3. Robust standard errors in parentheses are clustered on state. Models are weighted by start of period commuting zone share of national population. *** Significant at the 1 percent level. ** Significant at the 5 percent level. * Significant at the 10 percent level.

46 Other Labor Market Effects 2143 THE AMERICAN ECONOMIC REVIEW october 2013 Table 5 Imports from China and Employment Status of Working-Age Population within CZs, : 2SLS Estimates Dependent variables: Ten-year equivalent changes in log population counts and population shares by employment status Mfg emp Non-mfg emp Unemp NILF SSDI receipt (1) (2) (3) (4) (5) Panel A. 100 log change in population counts (Δ imports from China to US)/worker 4.231*** *** 2.058* 1.466*** (1.047) (0.651) (1.128) (1.080) (0.557) Panel B. Change in population shares All education levels (Δ imports from China to US)/worker 0.596*** *** 0.553*** 0.076*** (0.099) (0.137) (0.058) (0.150) (0.028) College education (Δ imports from China to US)/worker 0.592*** *** 0.304*** (0.125) (0.122) (0.039) (0.113) No college education (Δ imports from China to US)/worker 0.581*** 0.531*** 0.282*** 0.831*** (0.095) (0.203) (0.085) (0.211) Notes: N = 1,444 (722 CZs two time periods). All statistics are based on working age individuals (age 16 to 64). The effect of import exposure on the overall employment/population ratio can be computed as the sum of the coefficients for manufacturing and nonmanufacturing employment; this effect is highly statistically significant (p 0.01) in the full sample and in all reported subsamples. All regressions include the full vector of control variables from column 6 of Table 3. Robust standard errors in parentheses are clustered on state. Models are weighted by start of period CZ share of national population. *** Significant at the 1 percent level. ** Significant at the 5 percent level. * Significant at the 10 percent level.

47 Employment and Wage Effects by Education Group 2147 THE AMERICAN ECONOMIC REVIEW october 2013 Table 7 Comparing Employment and Wage Changes in Manufacturing and outside Manufacturing, : 2SLS Estimates Dependent variables: Ten-year equivalent changes in log workers and average log weekly wages I. Manufacturing sector II. Nonmanufacturing All workers College Noncollege All workers College Noncollege (1) (2) (3) (4) (5) (6) Panel A. Log change in number of workers (Δ imports from China 4.231*** 3.992*** 4.493*** to US)/worker (1.047) (1.181) (1.243) (0.651) (0.590) (0.764) R Panel B. Change in average log wage (Δ imports from China *** 0.743** 0.822*** to US)/worker (0.482) (0.340) (0.369) (0.260) (0.297) (0.246) R Notes: N = 1,444 (722 CZs two time periods). All regressions include the full vector of control variables from column 6 of Table 3. Robust standard errors in parentheses are clustered on state. Models are weighted by start of period CZ share of national population. *** Significant at the 1 percent level. ** Significant at the 5 percent level. * Significant at the 10 percent level. The effect of import exposure on mean wages found in panel B of Table 7 is

48 Government Transfers EC08CH08-Hanson ARI 29 September : $ : Downloaded from of Zurich - Hauptbibliothek on 11/21/16. For personal use only. Change (dollars) Figure 7 $3.65 Unemployment and TAA benefits $8.40 SSA disability benefits $10.00 SSA retirement benefits $15.04 Other government income assistance $18.27 Government medical benefits Total benefits Effect of a $1,000-per-worker increase in imports from China on dollar change of annual government transfer receipts per capita in commuting zones ( ). Abbreviations: SSA, US Social Security Administration; TAA, Trade Adjustment Assistance. Although trade theory has typically emphasized the impact of trade shocks on wages, these results suggest that adjustments at the employment margin might have an even larger quantitative impact on workers earnings. 36 A direct consequence of reduced employment and wages in trade-exposed local labor markets

49 Feler and Senses, 2017 Trade Shocks and the Provision of Local Public Goods 34 AMERICAN ECONOMIC JOURNAL MONTH YEAR Table 5 Effect of Chinese Import Exposure on Home Values (1) (2) (3) (4) (5) (6) (7) ln median val. own. occ. housing Median val. own. occ. housing Share homes < 150K Share homes K Share homes > 300K ln median rent Median annualized rent Chinese imports per worker (1.489) ( ) (1.264) (1.455) (1.773) (0.648) (62.20) R N 1,444 1,444 1,444 1,444 1,444 1,444 1,444 N Clusters Notes: All regressions include a constant, a dummy for the period, geographic dummies for the nine Census divisions, and the following startof-period locality control variables: the share of employment in manufacturing, the college-educated share of the population, the foreign-born share of the population, the share of women in the population, the share of routine occupations in employment, and the average offshorability index of occupations. Observations are weighted by the start-of-period commuting zone share of the national population. Robust standard errors, clustered at the commuting zone level, are reported in parentheses. denotes significance at the 1% level; denotes significance at the 5% level; denotes significance at the 10% level.

50 Feler and Senses, 2017 Lower Local Governments Tax Revenue Table 7 Effect of Chinese Import Exposure on Local Revenues Per Capita (1) (2) (3) (4) (5) (6) (7) (8) (9) (10) Total Panel A: Percent Changes (Per Capita) Intergov. Transfers Total from Own Sources General from Own Sources Total Taxes Property Taxes Sales, Inc., and License Taxes Other Taxes Charges Other Rev: Liq. Stores, Utility and Ins. Trust Chinese imports per worker (0.761) (0.650) (0.949) (0.776) (0.768) (0.849) (3.357) (25.529) (1.149) (3.421) R Panel B: Value Changes (Per Capita) Chinese imports per worker (52.921) (9.928) (49.393) (27.632) (23.252) (11.319) (13.084) (4.511) (14.460) (29.898) R N 1,444 1,444 1,444 1,444 1,444 1,444 1,444 1,444 1,444 1,444 N Clusters Notes: Total Revenue (1) is equal to Intergovernmental Transfers (2) plus Revenue from Own Sources (3). Revenue from Own Sources (3) is the sum of General Revenue from Own Sources (4) and Revenue from Liquor Stores, Utilities and the Insurance Trust Sector (10). General Revenue from Own Sources is the sum of Total Taxes (Property (6), Sales, Income and License (7) and Other Taxes (8)) and Charges (9). All regressions include a constant, a dummy for the period, geographic dummies for the nine Census divisions, and the following start-of-period locality control variables: the share of employment in manufacturing, the college-educated share of the population, the foreign-born share of the population, the share of women in the population, the share of routine occupations in employment, and the average offshorability index of occupations. Observations are weighted by the start-of-period commuting zone share of the national population. The dependent variables are calculated as log changes per capita in Panel A and changes in per capita dollar values in Panel B. Robust standard errors, clustered at the commuting zone level, are reported in parentheses. denotes significance at the 1% level; denotes significance at the 5% level; denotes significance at the 10% level. 36 AMERICAN ECONOMIC JOURNAL

51 Feler and Senses, 2017 Lower Public Good Provision Panel A: Percent Changes (Per Capita) Table 8 Effect of Chinese Import Exposure on Local Expenditures Per Capita (1) (2) (3) (4) (5) (6) (7) (8) (9) (10) (11) Sewerage Liq., Parks and Other Public Public Public and Solid Stores, Total General Educ. Transp. Natural General Safety Welfare Housing Waste Utility, Resources Exp. Manag. Ins.Trust Chinese imports per worker (0.531) (0.506) (0.457) (0.796) (3.772) (2.500) (1.282) (1.532) (1.380) (1.386) (1.482) R Panel B: Value Changes (Per Capita) Chinese imports per worker (25.551) (21.258) (7.986) (4.244) (2.653) (2.311) (2.489) (1.977) (2.509) (19.519) (20.421) R N 1,444 1,444 1,444 1,444 1,444 1,444 1,444 1,444 1,444 1,444 1,444 N clust Notes: Total Expenditure (1) is equal to General Expenditures (2) plus Other Expenditures on Liquor Stores, Utilities and the Insurance Trust Sector (11). General Expenditure consists of expenditure on Education, Public Safety, Public Welfare, Public Housing, Transportation, Parks and Natural Resources, Sewerage and Solid Waste Management and Other General Expenditures. All regressions include a constant, a dummy for the period, geographic dummies for the nine Census divisions, and the following start-of-period locality control variables: the share of employment in manufacturing, the college-educated share of the population, the foreign-born share of the population, the share of women in the population, the share of routine occupations in employment, and the average offshorability index of occupations. Observations are weighted by the start-of-period commuting zone share of the national population. The dependent variables are calculated as log changes per capita in Panel A and changes in per capita dollar values in Panel B. Robust standard errors, clustered at the commuting zone level, are reported in parentheses. denotes significance at the 1% level; denotes significance at the 5% level; denotes significance at the 10% level. VOL. VOL NO. ISSUE TRADE SHOCKS & LOCAL PUBLIC GOODS

52 Can the use of robots shape these patterns instead? Acemoglu and Restrepo (2017) Frey and Osborne (2013): in the next two decades, 47% of US workers are at risk of automation (they classify 702 occupations based on their susceptibility to automation). Similarly, McKinsey estimates the potential job replacement to be 45%. World Bank s World Development Report (2016) puts the number at 57% for the OECD countries. Acemoglu and Restrepo (2017) build a general-equilibrium model of trade, with the degree of substitutability between labor and robots varying across sectors. Robots replace workers; but they also make firms more productive, possibly increasing demand for labor and intermediate inputs.

53 Acemoglu and Restrepo (2017) Main Ideas Notice that the actual adoption of robots at the firm or CZ level is not observed (similar: Autor, Dorn, and Hanson (2013) never observed China s imports at the micro level). Similar to ADH, adopt the local labor market (US commuting zones (CZ)) approach to study how CZs exposure to robots, based on the initial industrial structure, is affected by the increasing adoption of robots across industries. Instrument: use industry-level spread of robots in other advanced economies to proxy for the improvement in the global technology frontier of robots. Main variable of interest: Count (per capita) of industrial robots, based on data from the International Federation of Robotics (IFR). An industrial robot as an automatically controlled, reprogrammable, and multipurpose [machine]. The IFR estimates: between 1.5 and 1.75 million industrial robots in operation; could increase to 4 to 6 million by (see Boston Consulting Group, 2015).

54 Main Findings Acemoglu and Restrepo (2017) One more robot reduces employment in a commuting zone with the average exposure to robots by 6.2 workers. One more robot per thousand employees reduce wages in such a commuting zone by about 0.73 percent. Computing the aggregate employment effect requires several parametric assumptions based on the trade model. Incorporating trade but not local demand spillover, 3 and 5.6 workers lose their jobs as a result of introducing one additional robot in the US economy Wages decline between 0.25% to 0.5% as a result of one more robot per 1000 workers.

55 The Adoption of Industrial Robots Figures and Tables Figure 1: Industrial robots in the United States and Europe. Note: Industrial robots per thousand workers in the United States and Europe. Data from the International Federation of Robotics (IFR).

56 The Adoption of Industrial Robots Automotive Plastic and chemicals Metal products Electronics Wood and Furniture Food and beveradges (1) (2) (3) (4) (5) (6) (1) (2) (3) (4) (5) (6) (1) (2) (3) (4) (5) (6) (1) (2) (3) (4) (5) (6) (1) (2) (3) (4) (5) (6) (1) (2) (3) (4) (5) (6) (1) (2) (3) (4) (5) (6) (1) (2) (3) (4) (5) (6) (1) (2) (3) (4) (5) (6) (1) (2) (3) (4) (5) (6) Basic metals Metal machinery Glass and ceramics Other transport vehicles 42 Normalized indices Textiles and leather Paper and printing Miscellaneous, toys, and others Mining Education, research, and development Agriculture Construction Utilities Services and other sectors (1)-Increase in the use of robots (2)-Increase in Chinese imports (3)-Offshoring of intermediate goods (4)-Increase in Mexican imports (5)-Capital stock growth (6)-IT capital growth Figure 2: Industry-level changes in the use of robots, Chinese imports, capital stock and IT capital. Note: This figure plots the increase in the number of robots per thousand workers between 1993 and 2007, the increase in the dollar value of Chinese imports per worker between 1990 and 2007, the growth of the capital stock between 1990 and 2007, and the growth of the stock of IT capital between 1990 and 2007 for the 19 industries for which we have the IFR data.

57 Heterogeneous Effects - Across Sectors CBP manuf. Manufacturing Highly robotized Other manuf. Transportation Agri. and mining Construction Services & retail Business services Wholesale Finance Public sector (1) (2) (3) (1) (2) (3) (1) (2) (3) (1) (2) (3) (1) (2) (3) (1) (2) (3) (1) (2) (3) (1) (2) (3) (1) (2) (3) (1) (2) (3) (1) (2) (3) (1) (2) (3) 50 Point estimate (1)-Long-differences (2)-Downweights outliers (3)-Stacked differences Figure 10: Relationship between the exposure to robots and industry employment. Note: The figure shows the estimates of the change in industry employment to population ratio against the exposure to robots between 1993 and 2007 conditional on the covariates in column 4 of Table 2. The green bars correspond to a long-differences specification similar to column 4 of Table 2; The rose bars correspond to a long-differences specification similar to column 6 of Table 2, in which we downweigh outliers; the blue bars correspond to a stacked-differences specification similar to column 2 of Table 3. For comparison, we also indicate with a dashed horizontal line the magnitude of the effect on Census private employment to population ratio.

58 Heterogeneous Effects - Across Occupations Routine manual Blue collar Assembly Mech. and transp. Farmers and miners Services Clerical Retail Professionals Management Financiers (1) (2) (3) (1) (2) (3) (1) (2) (3) (1) (2) (3) (1) (2) (3) (1) (2) (3) (1) (2) (3) (1) (2) (3) (1) (2) (3) (1) (2) (3) (1) (2) (3) 51 Point estimate (1)-Long-differences (2)-Downweights outliers (3)-Stacked differences Figure 11: Relationship between the exposure to robots and occupation employment. Note: The figure shows the estimates of the change in occupation employment to population ratio against the exposure to robots between 1993 and 2007 conditional on the covariates in column 4 of Table 2. The green bars correspond to a long-differences specification similar to column 4 of Table 2; The rose bars correspond to a long-differences specification similar to column 6 of Table 2, in which we downweigh outliers; the blue bars correspond to a stacked-differences specification similar to column 2 of Table 3. For comparison, we also indicate with a dashed horizontal line the magnitude of the effect on Census private employment to population ratio.

59 Heterogeneous Effects - Across Skill Groups All levels Less than highschool Highschool Some college College More than college Male Both Male Female Male Male Female Male Both Male Female Both Female Both Female Both Both Female Point estimate for wages Point estimate for employment All levels Less than highschool Highschool Some college College More than college Male Both Male Female Male Male Female Male Both Male Female Both Female Both Female Both Both Female Figure 12: Relationship between the exposure to robots and employment and wages by education group. Note: The figure shows the estimates of the change in Census private employment to population

60 A Threat to the China Syndrome Thesis? re 8