Issue, No.7 (September 2018)

Routine Task Intensity and Offshorability for the LIS

by Matthew C Mahutga, Michaela Curran (University of California, Riverside), and Anthony Roberts (California State University, Los Angeles)


In the contemporary period, globalization and technological change are among the most important drivers of income, inequality, and labor market dynamics. Both technological change and globalization are thought to reduce the demand for “low-skill” occupations, and increase the demand for “high-skill” occupations, which has implications for the structure of income and employment, the prevalence of poverty and the shape of the income distribution. New conceptualizations of what it means to be “low-skill” and “high-skill” has greatly advanced research on occupational stratification.

We recently created a new dataset on the “offshorability” (OFFS) and “routine-task intensity” (RTI) of occupations for use with the Luxembourg Income Study Database. These data characterize the susceptibility of occupational characteristics to offshoring and technological change, following the work of David Autor, Alan Blinder and Goos, Manning and Salomons (2014). Because these data are linked to the two-digit International Standard Classification of Occupations (ISCO-88) and only a subset of countries report ISCO-88 occupations to the LIS, we recoded 23 country-specific occupational schemes (74 LIS country-years) to the two-digit ISCO-88 scheme. All together, we produce individual level RTI and OFFS scores for 38 LIS countries and 160 LIS country-years. This recording effort also greatly expanded coverage of all LIS variables based on ISCO-88 (e.g. the OCCa and OCCb variables).

In a recent paper, we assess both the validity of these recodes and the utility of these data for substantive questions of interest to LIS users (Mahutga, Curran and Roberts, 2018). First, we compare average labor-income ratios predicted by recoded ISCO-88 occupational categories to those predicted by reported ISCO-88 occupational categories within countries that transitioned from country-specific to ISCO-88 codes over time. Second, we analyze the association between OFFS and RTI with work hours and labor incomes in the global North and South. In the remainder of this short note, we briefly describe the data, suggest these data can be trusted, and scratch the surface with respect to how we can use them in tandem with the rich LIS data. We ask future users of these data to cite Mahutga et al. 2018.

The data

RTI and OFFS are occupational level variables linked to ISCO-88 at the two-digit level.

The intuition behind RTI is that routine, manual, non-interactive (i.e. do not involve face-to-face interaction) job tasks are the most susceptible to automation. Conversely, non-routine/cognitive/ interactive jobs are the least susceptible to automation. RTI thus captures the degree to which occupations are routine-task intensive. Our data originate in work by Autor and Dorn (2013), as mapped on to ISCO-88 according to Goos et al. (2014). This work quantifies jobs as non-routine/cognitive, routine/cognitive, routine/manual, non-routine/manual and non-routine/interactive. Jobs with high RTI scores are high on either cognitive or manual routine tasks, and low on either cognitive, manual or both types of non-routine tasks. The RTI index now available to LIS users is linked to two-digit ISCO-88 occupations, and 0/1 standardized.

The intuition behind OFFS is that offhsorable jobs are not “place bound,” and can be done in the global South without a loss of quality. We implement Goos et al.’s (2014) operationalize of Blinder and Krueger’s (2013) measure. To produce these data, Blinder and Krueger employed export coders to assign an industry, Standard Occupational Classification (SOC) and OFFS score (from 1 to 5) to a random sample of respondents to the 2003 National Assessment of Adult Literacy. A score of 1 corresponds to “not offshorable” and a score of 5 corresponds to “easily offshorable with only minor (or no) difficulties or loss of quality.” Goos et al. cross-walked the SOC scores to ISCO-88, and we implement this 0/1 standardized variable in the LIS.

Why we can trust these data

Recoding the 23 country-specific occupational codes was herculean, and presents a potential source of measurement error. Thus, we took steps to ensure that our recodes were valid. We identified all of the countries that transitioned from a country-specific occupational scheme to ISCO-88 within the LIS data. We then used our recoded ISCO-88 categories to predict occupation-specific labor income ratios, and compared these predicted values to those predicted by ISCO-88 categories as reported by the same country in the closest year to the recode. We also compared income ratios predicted by country-reported ISCO-88 categories across the closest two years of similar distance to those in our recode comparison. This second set of comparisons give us a baseline rate at which occupation-specific mean labor incomes change over time. As detailed in the paper, our analyses suggest that our recodes are valid.

It was sometimes the case that the mix of occupations in a particular country-specific code did not map perfectly onto one ISCO-88 code. In these cases, we developed a weighting scheme to assign a country-specific occupational code an RTI/OFFS score that was proportional to the mix of ISCO-88 categories embodied within it (see Mahutga et al. 2018 and auxiliary files at for detailed information). The validity exercise described above is conservative in light of this weighting scheme (see Mahutga et al. 2018: 89).

To ensure the substantive validity of these data, we replicated work by Goos et al. (2014) linking both to polarization in work hours. We also show that both RTI and OFFS contribute to income polarization directly. Consistent with the natural intuition that RTI and OFFS should impact labor markets differently in the global North and South, we find that both contribute to work hour and income polarization in the North, but not in the South.

Why we should and how we can use these data

We contend that RTI and OFFS can provide additional explanatory power to a vast array of social-science models of income and employment, and can illuminate any theory for which income or employment are key explananda. We encourage users to consider the degree to which national (e.g. labor-market institutions and unions) and world-level (e.g. value-chains and production networks) phenomena moderate the impacts of RTI and OFFS on incomes and employment. Our paper elaborates on these themes (Mahutga et al., 2018).

Users can access these data in one of two ways. Users who wish to make use of assembled RTI and OFFS scores used in Mahutga et al. (2018), as well as new ISCO-88, occ1a and occ1b covariates that result from our recode (plus a few country-waves added after the writing of this text), can find them on the LIS website and here. We also provide user guide and codebook for the variables included in these data, as well as a very large document detailing the recoding particulars for each country-year recoded.

Users who wish to work with (or augment) our original script may find it on or by emailing the lead author. This script can be used to recode additional datasets as they come online in the LIS Database.

Autor, D. H. and Dorn, D. (2013), “The Growth of Low Skill Service Jobs and the Polarization of the U.S. Labor Market”, American Economic Review, 103(5): 1553-1597.
Blinder, A. S. and Krueger, A. B. (2013), “Alternative Measures of Offshorability: A Survey Approach”, Journal of Labor Economics, 31(S1): S97 – S128.
Goos, M.; Manning, A.; Salomons. A. (2014), “Explaining Job Polarization: Routine-Biased Technological Change and Offshoring”, American Economic Review, 104(8): 2509-2526.
Mahutga, M. C.; Curran. M.; Roberts, A. (2018), “Job Tasks and the Comparative Structure of Income and Employment: Routine Task Intensity and Offshorability for the LIS”, International Journal of Comparative Sociology, 59(2): 81-109.