CoChat
Sign Up
Cited by 2,111Labor market dynamics and wage inequalityGlobal trade and economicsEconomic Theory and Policy

Explaining Job Polarization: Routine-Biased Technological Change and Offshoring

Open access
Goos, Maarten, Manning, Alan, Salomons, Anna·2014·American Economic Review
DOI 10.1257/aer.104.8.2509
Open this paper in CoChat — free accountOpen PDF

Abstract

This paper documents the pervasiveness of job polarization in 16 Western European countries over the period 1993–2010. It then develops and estimates a framework to explain job polarization using routine-biased technological change and offshoring. This model can explain much of both total job polarization and the split into within-industry and between-industry components. (JEL J21, J23, J24, M55, O33)

AI snapshot

Written from the paper’s full text.

TL;DR

Job polarization is documented across 16 Western European countries from 1993 to 2010, with rising employment shares in high- and low-paid occupations and falling shares in middling occupations. A model combining routine-biased technological change and offshoring explains much of this total polarization and its within- and between-industry components, with routine-biased technological change mattering far more than offshoring.

Key findings

  • Pooled across the 16 countries, high-paying occupations gained 5.62 percentage points of employment share, middling occupations lost 9.27 points, and low-paying occupations gained 3.65 points between 1993 and 2010. Both within-industry and between-industry components were positive for high- and low-paying groups and negative for the middling group.
  • Polarization appeared in every country studied: Table 2 shows high- and low-paying occupation shares rising relative to middling occupations in each of the 16 countries.
  • In the labor-demand estimates, an occupation one standard deviation higher in routine task intensity grew about 0.90 percentage points less fast per year (RTI coefficient −0.900 in column 1). Offshorability had a small and statistically insignificant coefficient when RTI was included (−0.013 in column 1; −0.006 in column 3), while offshorability alone was negative and significant (−0.383 in column 5) but lost the horse race against RTI.
  • The model predicted 79 percent of the observed high-paying occupation share increase (4.45/5.62), 74 percent of the middling occupation decrease (6.86/9.27), and 66 percent of the low-paying occupation increase (2.41/3.65). The estimated product-demand elasticity was 0.42, implying that demand shifts attenuate the between-industry effect by 42 percent but do not reverse it.

Methods

  • Employment data come from the European Union Labour Force Survey for 1993–2010, using weekly hours worked, two-digit ISCO occupations, and one-digit NACE industries; Germany is covered using the SIAB social-security sample for 1993–2008. Output, marginal costs, and relative prices come from the OECD STAN database.
  • Routine task intensity is based on the Autor–Dorn RTI index mapped into European occupations; offshorability uses Blinder and Krueger’s preferred professional-coder measure. Both are normalized to mean zero and standard deviation one.
  • The paper develops a two-stage production model in which industry output is produced from tasks and tasks use domestic labor plus another input whose cost changes with routineness and offshorability. It estimates labor-demand equations with occupation–industry–country fixed effects and clustered standard errors, then uses a shift-share decomposition to separate within- and between-industry changes.

Limitations

  • The paper states that the RTI and offshorability measures are imperfect, may contain different amounts of measurement error, and make it difficult to distinguish the separate effects of routine-biased technological change and offshoring beyond reasonable doubt.
  • The paper states that wage data by country, occupation, and time are poor and potentially endogenous, so wages are modeled with dummies rather than included directly. It also notes that the homotheticity assumption is restrictive, though it reports no evidence of non-homotheticity at its level of industry aggregation.
  • The paper states that the model leaves room for additional factors. Evident from the text: the constant-returns assumption is rejected in one specification because the output coefficient is below one, so the authors impose a coefficient of one in later columns; offshoring is statistically insignificant when RTI is included; and the model’s predictions do not match actual changes exactly.

Model-written summary. Check it against the paper before relying on it.

Integrity signals

Registry

Crossref · 10.1257/aer.104.8.2509

Cites

33 of 33 references verified0 citation errors · 0 not checkable

Statistics

Nothing to recompute.

Claims the paper makes

0 of 12 cited sources checked · 0 supported · 0 flagged

  • Findingown data
    § IV. Explaining Job Polarization within and between Industries

    For the eight highest-paid occupations, predicted employment-share growth equals 79 percent of the actual rise.

    “For the group of the eight highest-paid occupations, our model predicts an increase in employment share that is 79 percent (4.45/5.62) of the increase actually observed.”
    Quote in the paperSubstantial claim
  • Findingown data
    § B. The Production of Tasks

    Excluding offshorability from the regression does not affect the RTI point estimate.

    “Column 4 of Table 3 excludes offshorability from the regression equation to examine the impact of RBTC unconditional on offshoring, and this does not affect the RTI point estimate.”
    Quote in the paperNot a substantial claim
  • Findingown data
    § IV. Explaining Job Polarization within and between Industries

    The model explains most of overall job polarization.

    “A comparison of columns 1 and 4 shows that our model explains most of overall job polarization.”
    Quote in the paperSubstantial claim
  • Findingreasoning
    § IV. Explaining Job Polarization within and between Industries

    With ε estimated at 0.42, the attenuation equals 42 percent of the between-industry effect when ε is zero, for every occupation.

    “given an estimate for ε of 0.42, we must have that for each occupation the attenuation effect is 42 percent of the between-industry effect for ε = 0.”
    Quote in the paperSubstantial claim
  • Findingown data
    § IV. Explaining Job Polarization within and between Industries

    For the nine middling occupations, predicted employment-share decline equals 74 percent of the actual fall.

    “For the group of nine middling occupations, the model predicts a decrease that is 74 percent (6.86/9.27) of the total observed employment share decrease.”
    Quote in the paperSubstantial claim
  • May combine two claims. 30
  • Classified as probably not a substantial claim. 8

  • May combine two claims.

    Findingown data
    § B. The Production of Tasks

    Within industries, relative demand shifts away from more routine and offshorable occupations, with routineness much more important.

    “In sum, this is evidence that within industries there is a shift in relative demand away from occupations that are more routine and offshorable with the former being the much more important effect.”
    Quote in the paperSubstantial claim

  • May combine two claims.

    Findingown data
    § B. Routineness and Offshorability of Occupations

    The correlation between the RTI and offshorability measures is 0.46 and statistically significant.

    “The correlation coefficient between the RTI and offshorability measure is 0.46 and statistically significant suggesting that routineness and offshorability may be linked.”
    Quote in the paperNot a substantial claim

  • May combine two claims.

    Findingown data
    § B. The Production of Tasks

    The coefficient on offshorability interacted with time is negative, small in absolute value, and not statistically significantly different from zero.

    “The coefficient on offshorability is also negative, but small in absolute value and not statistically significantly different from zero.”
    Quote in the paperNot a substantial claim

  • May combine two claims.

    Findingown data
    § V. Conclusions

    Polarization appears throughout European economies from 1993 to 2010, with both within-industry and between-industry components mattering.

    “The paper establishes that job polarization is pervasive across European economies in the period 1993-2010 and has within-industry and between-industry components that are both important.”
    Quote in the paperSubstantial claim

  • May combine two claims.

    Findingown data
    § B. The Production of Tasks

    Restricting the coefficient on log industry output to unity does not affect the RTI and offshorability coefficients and yields an estimate of η of 0.90.

    “This does not affect the RTI and offshorability coefficients and results in an estimate of 0.90 for η.”
    Quote in the paperNot a substantial claim

  • May combine two claims.

    Findingown data
    § IV. Explaining Job Polarization within and between Industries

    In column 3 of Table 3, the RTI index carries much more weight, so nearly all predicted changes are driven by the RTI variable.

    “Because the RTI index is much more important in column 3 of Table 3 almost all of these changes come from the RTI variable.”
    Quote in the paperNot a substantial claim

  • May combine two claims.

    Findingown data
    § B. The Production of Tasks

    The coefficient on log industry output is less than unity, rejecting constant returns to scale in goods production.

    “However, the coefficient on log industry output is less than unity which rejects our assumption of constant returns to scale in goods production.”
    Quote in the paperNot a substantial claim
    • The authors state this as an assumption: Goods production has constant returns to scale.

  • May combine two claims.

    Findingown data
    § B. The Production of Tasks

    In Table 3 column 1, the coefficient on RTI interacted with time is negative and significantly different from zero at the 1 percent level.

    “The coefficient on RTI is negative and significantly different from zero at the 1 percent level.”
    Quote in the paperSubstantial claim

  • May combine two claims.

    Findingown data
    § II. Job Polarization

    Overall job polarization can be decomposed into within-industry and between-industry components, each quantitatively too large to ignore.

    “What this shows is that overall job polarization can be split into a within-industry and between-industry component, each of which is too large quantitatively to be ignored.”
    • Also stated in: § II. Job Polarization · plain
      “In sum, the phenomenon of job polarization is pervasive across advanced economies.Moreover, overall job polarization is driven by job polarization both within and between industries.”
    Quote in the paperSubstantial claim

  • May combine two claims.

    Findingown data
    § V. Conclusions

    Western Europe's employment structure has polarized: high-paid professionals and managers and low-paid personal service workers gained share, while manufacturing and routine office workers lost share.

    “The employment structure in Western Europe has been polarizing with rising employment shares for high-paid professionals and managers as well as low-paid personal service workers and falling employment shares of manufacturing and routine office workers.”
    Quote in the paperSubstantial claim

  • May combine two claims.

    Findingown data
    § V. Conclusions

    The model accounts for a substantial portion of total polarization and of its division into within- and between-industry parts.

    “We show that our economic model explains a sizable part not just of overall job polarization but also the split into within-industry and between-industry components.”
    Quote in the paperSubstantial claim

  • May combine two claims.

    Findingown data
    § II. Job Polarization

    Employment shares increased for some low-paid service workers (51) and some low-paid elementary occupations (91), but not for sales workers (52) or manufacturing laborers (93).

    “Similar to patterns found for the United States and United Kingdom, there has been an increase in the employment shares for some low-paid service workers (51, of which the main task consists of providing services related to travel, catering, and personal care; but not 52, of which the main task consists of selling goods in shops or at markets)”
    • Also stated in: § II. Job Polarization · plain
      “and some low-paid elementary occupations (91), which are service elementary workers including cleaners, domestic helpers, doorkeepers, porters, security personnel, and garbage collectors; but not 93, which mainly includes low-educated laborers in manufacturing performing simple tasks including product-sorting and packing and unpacking produce).”
    Quote in the paperSubstantial claim

  • May combine two claims.

    Findingown data
    § B. The Production of Tasks

    In Table 3 column 2, the coefficient on log industry marginal cost, an estimate of η, is positive and significant.

    “The coefficient on log industry marginal cost-which is an estimate of η-is positive and significant as is predicted by our model.”
    Quote in the paperSubstantial claim

  • May combine two claims.

    Findingown data
    § B. Routineness and Offshorability of Occupations

    The RTI and offshorability measures differ: physical, mathematical, and engineering professionals are nonroutine but highly offshorable, while office clerks are very routine but not very offshorable.

    “For example, physical, mathematical, and engineering professionals (21) is ranked as nonroutine but at the same time highly offshorable whereas office clerks (41) are very routine but not very offshorable.”
    Quote in the paperNot a substantial claim

  • May combine two claims.

    Findingown data
    § II. Job Polarization

    For the eight highest-paid and four lowest-paid occupations, both within-industry and between-industry components are positive; for the nine middling occupations, both are negative.

    “For the group of eight highest-paid and four lowest-paid occupations the within-industry and between-industry components are positive, and for the group of nine middling occupations they are both negative.”
    Quote in the paperSubstantial claim

  • May combine two claims.

    Findingown data
    § By Maarten Goos, Alan Manning, and Anna Salomons *

    In the data, the product-demand shift attenuates between-industry job polarization but does not reverse it.

    “We show that, in our data, this effect attenuates between-industry job polarization but does not overturn it.”
    Quote in the paperSubstantial claim

  • Probably not a substantial claim: may not pass S3 (load-bearing).

    Findingreasoning
    § IV. Explaining Job Polarization within and between Industries

    In equation (13), the factor preceding the curly brackets equals occupation j's total employment share times the negative RTI coefficient from Table 3.

    “The term before the curly brackets, s jt [ 1 -η ] [1 -κ] γ R , is occupation j's share of total employment, s jt , multiplied by the RTI coefficient in Table 3, [1 -η][1 -κ] γ R , which is negative.”
    Quote in the paperNot a substantial claim

  • May combine two claims.

    Findingown data
    § V. Conclusions

    Without conditioning on RBTC, offshoring has some explanatory power but fits worse; it cannot, for instance, account for office clerks' within-industry employment-share decline.

    “offshoring while not conditioning on RBTC does appear to have some explanatory power but it is noticeable that the fit is poorer e.g., offshoring cannot explain the decline in the employment share of office clerks within industries.”
    Quote in the paperSubstantial claim

  • May combine two claims.

    Interpretationreasoning
    § IV. Explaining Job Polarization within and between Industries

    The estimates indicate that routine-biased technological change and offshoring generate polarization both inside and across industries.

    “our estimates are such that RBTC and offshoring lead to job polarization within and between industries”
    Quote in the paperSubstantial claim
    • The authors state this as an assumption: The model assumes homothetic demand, meaning aggregate income's level and distribution leave the cross-industry demand distribution unchanged.
    • The authors state this as an assumption: For each industry, task j's cost share is assumed to equal occupation j's employment share.

  • May combine two claims.

    Interpretationreasoning
    § IV. Explaining Job Polarization within and between Industries

    The paper's model can produce the shift-share decomposition, and occupational employment share changes can be split into within-industry and between-industry parts.

    “equation (13) shows how a shift-share analysis-that is commonly thought of as merely an accounting exercise-can be rooted in our framework.”
    • Also stated in: § IV. Explaining Job Polarization within and between Industries · plain
      “equation (13) shows how changes in occupational employment shares can be decomposed into a within-industry and between-industry component”
    Quote in the paperSubstantial claim
    • The authors state this as an assumption: The model assumes homothetic demand, meaning aggregate income's level and distribution leave the cross-industry demand distribution unchanged.
    • The authors state this as an assumption: For each industry, task j's cost share is assumed to equal occupation j's employment share.

  • Probably not a substantial claim: may not pass S3 (load-bearing), S4 (based).

    Study design and conductnone given
    § V. Conclusions

    Using unweighted country averages rather than pooled data would leave the figures in columns 1 and 2 of Table 1 qualitatively unchanged.

    “Instead of pooling countries, reporting unweighted average employment shares and their changes across countries would not qualitatively change the numbers in columns 1 and 2 of Table1.”
    Quote in the paperNot a substantial claim

  • May combine two claims.

    Interpretationreasoning
    § V. Conclusions

    Equation (13)'s predicted changes reflect only RBTC and offshoring effects, not industry-specific or industry-time-specific employment-share shifts.

    “Note that the predicted changes from equation (13) are precisely the effects from RBTC and offshoring and cannot be explained by any industry or industry-time specific changes in employment shares.”
    Quote in the paperSubstantial claim
    • The authors state this as an assumption: The model assumes homothetic demand, meaning aggregate income's level and distribution leave the cross-industry demand distribution unchanged.
    • The authors state this as an assumption: For each industry, task j's cost share is assumed to equal occupation j's employment share.

  • May combine two claims.

    Interpretationreasoning
    § IV. Explaining Job Polarization within and between Industries

    RBTC usually reduces routine labor needed per unit of output, so cross-industry routineness differences shift industry employment shares even when output shares stay constant.

    “RBTC will typically cause less routine employment to be used in the production of a given level of output, so differences in routineness across industries will cause industry employment shares to change even if industry output shares do not.”
    Quote in the paperSubstantial claim
    • The authors state this as an assumption: The model assumes homothetic demand, meaning aggregate income's level and distribution leave the cross-industry demand distribution unchanged.
    • The authors state this as an assumption: For each industry, task j's cost share is assumed to equal occupation j's employment share.

  • May combine two claims.

    Interpretationown data
    § IV. Explaining Job Polarization within and between Industries

    Taken together, the model and results indicate that routine-biased technological change and offshoring account for a large share of polarization and its within- and between-industry parts, while other factors may also matter.

    “Overall, our model and results suggest that RBTC and offshoring explain much of job polarization and its within-industry and between-industry components, although the results certainly leave room for additional factors to play a role.”
    Quote in the paperSubstantial claim

  • May combine two claims.

    Interpretationreasoning
    § IV. Explaining Job Polarization within and between Industries

    Routine-intensive industries experience bigger cost and relative-price declines, raising demand for their goods and thus for every occupation they employ.

    “routine intensive industries will see a larger fall in industry costs and therefore relative output prices.This will increase the demand for routine intensive goods and therefore the demand for all occupations used in routine intensive industries.”
    Quote in the paperSubstantial claim
    • The authors state this as an assumption: The model assumes homothetic demand, meaning aggregate income's level and distribution leave the cross-industry demand distribution unchanged.
    • The authors state this as an assumption: For each industry, task j's cost share is assumed to equal occupation j's employment share.

  • May combine two claims.

    Interpretationreasoning
    § IV. Explaining Job Polarization within and between Industries

    Equation (13)'s within- and between-industry effects have economic meaning and vary with the model's elasticities.

    “the within-industry and between-industry effects in equation (13) are economically meaningful and depend on the different elasticities in the model.”
    Quote in the paperSubstantial claim
    • The authors state this as an assumption: The model assumes homothetic demand, meaning aggregate income's level and distribution leave the cross-industry demand distribution unchanged.
    • The authors state this as an assumption: For each industry, task j's cost share is assumed to equal occupation j's employment share.

  • May combine two claims.

    Interpretationreasoning
    § IV. Explaining Job Polarization within and between Industries

    This second part weakens the first part but does not reverse it provided ε<1, consistent with the estimated ε of 0.42.

    “this channel will tend to attenuate the first part of the between-industry effect though will not overturn it as long as ε < 1 as our estimate for ε of 0.42 suggests.”
    Quote in the paperSubstantial claim
    • The authors state this as an assumption: The model assumes homothetic demand, meaning aggregate income's level and distribution leave the cross-industry demand distribution unchanged.
    • The authors state this as an assumption: For each industry, task j's cost share is assumed to equal occupation j's employment share.

  • Probably not a substantial claim: may not pass S3 (load-bearing). May combine two claims.

    Interpretationreasoning
    § IV. Explaining Job Polarization within and between Industries

    The second curly-bracket term in equation (13) is the between-industry component, and it has two parts.

    “The second term in the curly brackets, [1 -ε] /[1 -η] ∑ i=1 I s i | jt ( R it I -R t agg ), is the between-industry component.This can be thought of as having two parts.”
    Quote in the paperNot a substantial claim
    • The authors state this as an assumption: The model assumes homothetic demand, meaning aggregate income's level and distribution leave the cross-industry demand distribution unchanged.
    • The authors state this as an assumption: For each industry, task j's cost share is assumed to equal occupation j's employment share.

  • May combine two claims.

    Study design and conductown data
    § V. Conclusions

    Start and end years do not drive the results; yearly employment shares follow nearly linear trends.

    “One might be concerned that our results are sensitive to the start and end years.However, online Appendix A presents the employment shares for all years showing that changes are close to linear trends.”
    Quote in the paperNot a substantial claim

  • May combine two claims.

    Premisecitation
    § By Maarten Goos, Alan Manning, and Anna Salomons *

    Job polarization has been documented for Germany and there are indications that it is pervasive in European countries.

    “Job polarization has also been documented for Germany (Spitz-Oener 2006;Dustmann, Ludsteck, and Schönberg 2009) and there are indications it is pervasive in European countries (Goos, Manning, and Salomons 2009;Michaels, Natraj, and Van Reenen 2014).”
    Quote in the paperSubstantial claim0 of 3 sources checked
    • cites Spitz-Oener 2006

      Not checked in the free run.

    • cites Dustmann, Ludsteck, and Schönberg 2009

      Not checked in the free run.

    • cites Goos, Manning, and Salomons 2009

      Not checked in the free run.

  • Probably not a substantial claim: may not pass S3 (load-bearing).

    Study design and conductown data
    § V. Conclusions

    When column 4 of Table 3 supplies the estimates-RBTC's impact without conditioning on offshoring-the results closely match Table 4 and Figure 1.

    “Firstly, using point estimates in column 4 of Table 3, thereby using the estimated impact of RBTC unconditional on offshoring.Results are very similar to those presented in Table 4 and Figure 1.”
    Quote in the paperNot a substantial claim

  • Probably not a substantial claim: may not pass S3 (load-bearing).

    Premisecitation
    § V. Conclusions

    Blinder and Krueger (2013) suggest that simple, routine-task jobs are more offshorable than jobs needing complex thought, judgment, and human interaction.

    “The authors argue that it is likely that jobs that can be broken down into simple, routine tasks are easier to offshore than jobs requiring complex thinking, judgment, and human interaction.”
    Quote in the paperNot a substantial claim0 of 1 sources checked
    • cites Blinder and Krueger (2013)

      Not checked in the free run.

  • Probably not a substantial claim: may not pass S3 (load-bearing). May combine two claims.

    Premisecitation
    § V. Conclusions

    Blinder and Krueger (2013) report that offshorability differs conceptually from routineness while still being related to it.

    “This is in line withBlinder and Krueger (2013) who find that offshorability is conceptually distinct from, though related to, an occupation's routineness.”
    Quote in the paperNot a substantial claim0 of 1 sources checked
    • cites Blinder and Krueger (2013)

      Not checked in the free run.

  • Probably not a substantial claim: may not pass S2 (checkable).

    Study design and conductreasoning
    § V. Conclusions

    Because the RTI and offshorability measures are imperfect and may have different measurement error, claims separating RBTC from offshoring effects require caution.

    “As our RTI and offshorability measures are undoubtedly imperfect, and the extent of measurement error in them may also be different, one has to be careful in one's claims about the ability to distinguish beyond reasonable doubt the separate effects of RBTC and offshoring.”
    Quote in the paperNot a substantial claim

  • Probably not a substantial claim: may not pass S3 (load-bearing). May combine two claims.

    Premisecitation
    § V. Conclusions

    Blinder and Krueger (2013) conclude that offshorability is more relevant to offshoring than routineness is, though the two criteria overlap.

    “Consequently,Blinder and Krueger (2013) conclude that their measure of offshorability is more germane to the offshoring issue than is the question of routineness but that the two criteria do overlap.”
    Quote in the paperNot a substantial claim0 of 1 sources checked
    • cites Blinder and Krueger (2013)

      Not checked in the free run.

35 of 35

  • The authors state this as an assumption.

    Output in an industry is produced by combining tasks via a CES production function with elasticity of substitution η greater than zero.

    “Assume that output in industry i is produced by combining certain common building blocks that we will call tasks.In general, assume the following CES production function for industry i using tasks T 1 , … , T ij , … , T J as inputs:”

  • The authors state this as an assumption.

    Task output is produced using domestic labor of occupation j and one other input via a Cobb-Douglas production function common across industries, with 0<κ<1.

    “If industry i uses domestic labor of occupation j, N ij , and the other input, K ij , we assume that the total production of task j is given by a Cobb-Douglas production function that is common across industries:”

  • The authors state this as an assumption.

    The technology to produce task j and factor prices w_j and r_j are common across industries, so the marginal cost to produce task j has no industry subscript.

    “Note that it is assumed that the technology to produce task j together with factor prices w j and r j are common across industries such that the marginal cost to produce task j has no subscript i.”

  • The authors state this as an assumption.

    Routine-biased technological change and offshoring affect the cost of the other input according to ∂ log r_jt/∂t = γ_R R_j + γ_F F_j.

    “We model RBTC and offshoring as affecting the costs of the inputs other than domestic labor, in particular that (5) ∂ log r jt _ ∂ t = γ R R j + γ F F j , with R j the routineness and F j the offshorability measure discussed in Section I.”

  • The authors state this as an assumption.

    The log demand for industry output is given by log Y_ict = -ε log P_ict + log Y_ct + ξ_ict.

    “To do this, assume that the log demand for industry output is given by (7) log Y ict = -ε log P ict + log Y ct + ξ ict ,”

  • The authors state this as an assumption.

    There is monopolistic competition within each industry, with log P_ict = log μ + log c^I_ict and μ>1.

    “If there is monopolistic competition within each industry we have that log P ict = log μ + log c ict I with μ > 1 the price mark-up.”

  • The authors state this as an assumption.

    The industry output demand specification could be derived from a CES utility function with homothetic preferences.

    “This specification could be derived from an underlying CES utility function in which consumers have homothetic preferences.”

  • The authors state this as an assumption.

    Goods production has constant returns to scale.

    “However, the coefficient on log industry output is less than unity which rejects our assumption of constant returns to scale in goods production.”

  • The authors state this as an assumption.

    The model assumes homothetic demand, meaning aggregate income's level and distribution leave the cross-industry demand distribution unchanged.

    “The assumption of homotheticity implies that changes in both the level and the distribution of aggregate income have no effect on the distribution of demand across industries.”

  • The authors state this as an assumption.

    For each industry, task j's cost share is assumed to equal occupation j's employment share.

    “the simplifying assumption that the share of task j in industry costs is the same as the share of employment of occupation j in that industry.”
10 of 10

Claims other papers make about this one

Each is checked against this paper’s own text: does it say what the citing paper says it does?

  • Studies conducted in high income countries document that a growing number of workers are employed in occupations which are characterized as precarious.

    “Studies conducted in high income countries document that a growing number of workers are employed in occupations which are characterized as precarious (5–8).”

    From Large language model exposure and precarious occupations: Unpacking relationships in the Canadian labor force (2026) · cites this paper as [6]

    Section II, “Job Polarization” (page not available), reports rising employment shares in some low-paid service and elementary occupations across 16 Western European countries. The paper classifies those occupations by pay, not as precarious, so it does not support that characterization. (LLM, conf=0.85)

    Experimental — accuracy not yet measured

  • The automation of work was most prominent among occupations with repetitive job tasks and low skill requirements, which have been amongst the most precarious.

    “The automation of work was most prominent among occupations with repetitive job tasks and low skill requirements, which have been amongst the most precarious (6).”

    From Large language model exposure and precarious occupations: Unpacking relationships in the Canadian labor force (2026) · cites this paper as [6]

    The paper’s introductory discussion says technological change replaces labor in routine tasks. But it identifies the resulting decline as concentrated in middling occupations relative to high- and low-skilled ones; Section II reports declines in occupations paying around the median wage and increases in some low-paid occupations. The cited passages do not characterize the affected occupations as the most precarious. (LLM, conf=0.88)

    Experimental — accuracy not yet measured

  • Research conducted during past periods of technological change has indicated that the adoption of new technologies has been associated with increases in precariousness reported by workers in occupations facing the lowest quality employment conditions.

    “Research conducted during past periods of technological change have indicated that the adoption of new technologies has been associated with increases in precariousness reported by workers in occupations facing the lowest quality employment conditions (5, 6).”

    From Large language model exposure and precarious occupations: Unpacking relationships in the Canadian labor force (2026) · cites this paper as [6]

    The paper’s Conclusions (page unavailable in our copy) link routine-biased technological change to job polarization, including a rising employment share for low-paid personal service workers. But the paper analyzes occupational employment shares, not workers’ reports of increased precariousness or changes in employment conditions. Its findings therefore support the technological-change context, but not the claim’s central outcome. (LLM, conf=0.84)

    Experimental — accuracy not yet measured

  • Many entry level and manual jobs are not routine and therefore not easily codifiable.

    “Third, many entry level and manual jobs are in fact not routine and, therefore, not easily codifiable (Goos et al. 2014;Autor et al. 2015, Barbieri et al. 2020).”

    From Artificial intelligence and work: a critical review of recent research from the social sciences (2024) · cites this paper as Goos et al. 2014

  • As technology becomes pervasive across industries, developed economies are increasingly requesting more highly skilled workers, while reducing the number of low skilled and routine jobs.

    “Many of these studies have found that as technology becomes pervasive across industries, developed economies are increasingly requesting more highly skilled workers, while reducing the number of low skilled and routine jobs (Autor, 2014, Goos et al., 2014).”

    From Skills or degree? The rise of skill-based hiring for AI and green jobs (2025) · cites this paper as Goos et al., 2014

Cites

Flagged first, then by citations

6 of the 39 references couldn’t be matched to a known paper.

Research works printed but not linked: 2

  1. Baumol William J, 1967, American Economic Review, 57, 415
  2. Goldin Claudia, 2009, The Evolution of U.S. Wage Differentials, 1890

Cited by

Sorted by citations

Related

Authors
Goos, M., Manning, A., & Salomons, A.
Affiliations
  • KU Leuven (BE)
  • London School of Economics and Political Science (GB)
  • Utrecht University (NL)
ORCID
Venue
American Economic Review
Publisher
American Economic Association
Published
2014-08-01
Volume
104
Issue
8
Pages
2509-2526
ISSN
0002-8282 (ISSN-L)
DOI
10.1257/aer.104.8.2509
OpenAlex
W1980544129
Semantic Scholar
c219423d9836f1d81b08710f724f0e9b787e568f
Cited by
2,111
References
39
Citations by year
2026: 214 · 2025: 217 · 2024: 228 · 2023: 199 · 2022: 234 · 2021: 240 · 2020: 224 · 2019: 166 · 2018: 161 · 2017: 100 · 2016: 68 · 2015: 40 · 2014: 14 · 2013: 3 · 2012: 1
FWCI
183.49
Type
journal article
Language
en
Open access
Locations
  • American Economic Review (publishedVersion)
  • London School of Economics and Political Science Research Online (London School of Economics and Political Science) (submittedVersion)
  • London School of Economics and Political Science Theses Online (London School of Economics and Political Science) (submittedVersion)
  • RePEc: Research Papers in Economics (submittedVersion)
  • Utrecht University Repository (Utrecht University) (submittedVersion)
  • Lirias (KU Leuven) (submittedVersion)
Funders
Economic and Social Research Council (ES/H02123X/1)
Topics
  • Social Sciences › Economics, Econometrics and Finance › Economics and Econometrics › Labor market dynamics and wage inequality
  • Social Sciences › Economics, Econometrics and Finance › General Economics, Econometrics and Finance › Global trade and economics
  • Social Sciences › Economics, Econometrics and Finance › General Economics, Econometrics and Finance › Economic Theory and Policy
Keywords
job polarization, routine-biased technological change, offshoring, Western Europe, labor market structure, skill-biased technological change

About these checks

Registry, retraction, Cites and statistics results are computed, not judged, with one exception: a printed reference that no index lists is read once by a model before it counts as a citation error. Claims are extracted by a model and kept only when their quote appears verbatim in the text; each cited source is then checked, with a model’s help, for whether it says what the claim relies on. Those verdicts can be wrong — read the claim’s quote and the cited source before relying on one.

Claims read by deepseek-v4.1-flash under definition version 3 (extract-v3) on 2026-10-07.

Classified by jev-1.13-20260917 (gold-guideline-v0.2) on 2026-10-07.

Statistics and arithmetic are recomputed, not judged.

Confidence bands are shown only for fields calibrated on our annotated test set. None is calibrated yet, so no band is shown.

What counts as a claim ›How we measure confidence ›How we check this ›

Are you an author, or do you see a problem with this page? Request a review