Content uploaded by Jeisson Cardenas
Author content
All content in this area was uploaded by Jeisson Cardenas on May 21, 2020
Content may be subject to copyright.
Content uploaded by Jeisson Cardenas
Author content
All content in this area was uploaded by Jeisson Cardenas on May 21, 2020
Content may be subject to copyright.
Content uploaded by Jeisson Cardenas
Author content
All content in this area was uploaded by Jeisson Cardenas on May 11, 2020
Content may be subject to copyright.
Possible effects of Coronavirus in the
Colombian labour market
Jeisson C´ardenas
Jaime Montana
Documento de Trabajo
Alianza EFI - Colombia Cient´ıfica
Mayo 2020
N´umero de serie: WP2-2020-006
Possible effects of Coronavirus in the Colombian labour
market∗
Jeisson CARDENAS†Jaime MONTANA‡
This version: May 7, 2020
Abstract
The COVID-19 pandemic and its social distancing measures have brought unprece-
dented socio-economic challenges worldwide. One of the most urgent questions is how
the labour force will be affected by the pandemic. The answer to this question will have
considerable impact on the countries’ productivity, poverty and unemployment rates,
etc. Consequently, the measurement of jobs which can be performed without increasing
the risk of contagion has become a priority. However, due to the lack of proper informa-
tion, less advanced countries such as Colombia (where unemployment and informality
rates are high - around 10.5% and 46.2%, respectively in 2019) face huge challenges
in making such estimations. Thus, we contribute to the debate by adapting different
international work-from-home and proximity measures and estimated the proportion of
workers in the corresponding groups according to the context of a developing country
such as Colombia. Our results suggest that a fifth of jobs in Colombia can potentially
be done from home. While around 10% of Colombian workers have a high degree of
physical interaction with other people.
Keywords: Coronavirus, Occupations, Demographics, Households
∗J. C´ardenas thanks the “inclusi´on productiva y social: programas y pol´ıticas para la promoci´on de una
econom´ıa formal, c´odigo 60185, que conforma la Alianza EFI, bajo el Contrato de Recuperaci´on Contingente
No. FP44842-220-2018” program.
†University of Warwick. Email: jeisson.cardenas-rubio@warwick.ac.uk
‡Paris School of Economics and Turin University. jaimem.montana@gmail.com
1 Introduction
For a developing country with substantial levels of informality, how many jobs can be per-
formed from home during the social distance measures? What are the works that will suffer
in the phasing out of the confinement policies? What workers might face a more significant
risk of unemployment? This short paper presents evidence built on the Colombian household
survey GEIH 2016 - 2019 and on the occupational information from O*NET and STEP to
answer such questions. This paper complements previous efforts that assess the impact of
the virus and social lock-down in the Colombian labour market (Eslava and Isaacs, 2020;
Jaramillo et al., 2020).
The first coronavirus case in Colombia was detected on March 6th, 2020. On March
20th, just after a couple of days form the first infection, the Colombian government declared
quarantine as a social distance mechanism to prevent further spread of the virus. The
economic impact of social distancing measures is not uniform across all economic activities
and occupations (Del Rio-Chanona et al., 2020). This has been documented in the U.S.
following two approaches. The first approach characterises the jobs that can be performed
at home (teleworkable) (Dingel and Neiman, 2020). The second approach characterises jobs
that are at higher risk of contagion because their task involves high proximity with others
(Leibovici et al., 2020; Mongey and Weinberg, 2020).
Thus, this document, even if it contains essential and unknown information at the country
and regional level, is original in the methodology used. It relies heavily on recent develop-
ments (Dingel and Neiman, 2020; Mongey and Weinberg, 2020; Saltiel, 2020). This paper
fills a void for possible effects of COVID-19 at a disaggregated level, and present detailed
information for a developing economy with significant levels of informality.
The paper also presents the characterisation of the most significant potential affected
population. Moreover, this paper contributes to an ongoing debate on the use of O*NET data
in developing countries (Lo Bello et al., 2019). Specifically, we have used the Occupational
information network (O*NET) and the Skills Toward Employability and Productivity survey
1
(STEP) to adapt different international work-from-home and proximity measures according
to the Colombian context. We find that the results from the O*NET and STEP lead to
similar statistical estimates. This fact encourages the cautionary complimentary use of
O*NET as secondary data for analysis in developing countries.
The document is divided into four parts. Section 2 presents the data used. Section 3
presents the analysis at the national level and to describe which household and worker char-
acteristics are more likely associated with teleworking. In this section, we also evidence the
different results obtained from O*NET and the from the STEP index constructed. Section
4 is devoted to analysing at the national level the effect of proximity and face-to-face inter-
actions. Analysing the latter dimension is important because the affected occupations (and
populations) might see a considerable change in demand, induced by changes in consumer
behaviour to avoid contagion. Section 5 presents the conclusion and present some remarks
relevant for public policy.
2 Data
Since 2006, the Colombian statistics office of Colombia (DANE) has conducted a monthly
cross-sectional household survey (GEIH, by its initials in Spanish) to measure the charac-
teristics of the Colombian workforce. The GEIH with a total sample size of approximately
23,000 households monthly is nationally representative and the main source for official labour
market information in Colombia. In this survey, people are asked about their current level
of education and occupation, among other characteristics. The main variables that we used
in this analysis are reported in table A1.
However, one caveat of the GEIH is that the occupational classification used is the Na-
tional Classification of Occupations 1970 (SOC 1970) at 2-digit level, a national occupation
classification based on the qualification level and the sector1. As mentioned by C´ardenas
(2020) the use of such outdated classification might affect any statistical analysis due to
1Sector is the closest translation for ‘performance area’.
2
labour market changes. For instance, a set of occupations might have emerged during the
last decades and the SOC 1970 might not have the proper categories to group these occupa-
tions.
One might try to use crosswalks and re-code the SOC-1970 to use a more up-to-date
occupational classification such as the International Standard Classification of Occupations
2008 (ISCO-08) provided by the International Labour Organization (ILO). However, the use
of crosswalks between SOC-1970 and ISCO-08 is limited. For instance, the SOC-1970 at
2-digit has the following occupational group: ‘Building keeper, doorman, sacristan, cleaner,
window cleaner, chimney sweep’(code 55). These group of occupations belong to different
occupational groups in ISCO-08. For instance, a Building keeper in ISCO-08 is ‘Security
guards’ (code 54 at 2-digit level), while a cleaner is coded as ‘Domestic cleaners’ (code 91).
Consequently, any attempt to re-code SOC 1970 to ISCO-08 occupational groups might have
considerable inaccuracies. These measurement errors are relatively relevant for the estima-
tion of teleworkable or proximity occupations since one occupation in the same occupational
group might require, for instance, high proximity with others while the opposite can be true
for another occupation in the same occupational group at 2-digit level2.
Thus, we used the raw text responses of the GEIH to reclassify the information to ISCO-
08 and SOC-ONET at 4-digits and 6-digits level, respectively. To do so, we train a machine
learning algorithm over 2.2 million of observations from a posted vacancy database, which
has been pre-classified and validated (C´ardenas, 2020). This codification at 4-digit level is
relevant because it allows us to merge with more precision different international classifica-
tions to the Colombian data.
In contrast, using the ISCO at 1, 2 or 3 digit level and international classifications (e.g.
O*NET) might be considerably imprecise in categorising teleworkable or high proximity
occupations. For instance, there is the following category group at 3-digit level ISCO ‘client
information workers’ (code 422). This occupational category has these subgroups ‘Telephone
2To the extent of our knowledge, there is not an official crosswalk between SOC 1970 and ISCO - 08 due
to the difficulties in using crosswalks between this two classifications.
3
switchboard operators’ (code 4222) and ‘Hotel receptionists’ (code 4224). The former group
might tend to be a teleworkable occupation, while the second group not. Thus, any attempt
to use the ISCO at 2 or 3-digit level might lead to incorrectly assign attributes to different
worker groups.
We combined the Colombian household survey with other sources at 4-digit level to de-
termine the proportion of workers susceptible to distress due to the COVID-19 pandemic.
We use different criteria: (i) the extend for which occupation is teleworkable or not, de-
scribes what share of the population could not be affected by social distance measures; (ii)
a proximity index that describes how close a person must be to other individuals to perform
his job, so the demand could drop because of the virus; (iii) a face-to-face indicator that
shows if the occupation has frequent face-to-face interaction, and could be affected because
of the spreading of the virus.
In particular, to measures to what extent an occupation is teleworkable, we use the
construction by Dingel and Neiman (2020), which is constructed in selected dimensions of
the work context and work activities from O*NET. Using O*NET in developing countries
has been of concern in research given the differences in working realities in the developed and
developing world, the differences in technology adoption, occupation licensing and legislation,
the informal prevalence of work in developing countries among the many criticisms.
To address this issue, we followed (Saltiel, 2020). We construct a teleworkable indicator
using the STEP survey3, with all the countries pooled4. The STEP survey contains informa-
tion on the skills and work activities in low and middle-skill countries. The final result is a
teleworkable indicator at the 4-digit occupational level. We compare the obtained results in
both indicators, and both agree in the classification for 84.3% of total workers in Colombia.
We report in the next section the estimates for each of the teleworkable indicators. Con-
3The index is composed by the following dimensions: (i) Lifted Anything more than 50 Pounds; (ii)
Time/contact involved with non-coworkers/customers; (iii) Repair electronic equipment (already generated);
(iv) Operate/work with heavy machines (already generated); (v) Use a computer at work. The results are
then aggregated to ISCO-08 3-digit level and merged with the Colombian household survey.
4We use all of the 16 countries, excluding China and Albania. We do not consider only the information
on Colombia because the sample size is small (1930 observations).
4
sidering that technology penetration is lower in Colombia, the estimates are an upper limit
(an optimistic scenario) for telework-compatible occupations. The number reported is also
in line with the country level reported in Dingel and Neiman (2020).
To measure workers that are at higher risk of contagion because their tasks involve close
proximity with others could be affected by the virus, we calculate a proximity index following
Leibovici et al. (2020) and Mongey and Weinberg (2020). The proximity index indicates if
an occupation to be performed needs to be at a proximity of 1.5 o less (less than an arm of
distance). This information is contained in category 4.C.2.a.3 from O*NET (work context).
We present the results form this index only since there is not a comparable measure in the
STEP survey. Nevertheless, a similar disclaimer as above can be made, and we consider that
the estimates are a lower bound due to cultural differences and institutional arrangements
(occupational licensing and health and safety regulations). Identify the share of occupations
that need proximity is important to determine potential changes in both demand and supply.
In demand, since people will be reluctant to buy goods and services in which the personal
proximity is large in order to avoid the contagion. The changes can also be induced by the
supply of labour to those occupations since workers trying to avoid risk for them, and their
families decide to avoid such employment relationships.
Finally, we also construct a face-to-face index following Avdiu and Nayyar (2020). This
measure indicates if the person to perform the job need face-to-face contact. This information
also comes from the O*NET taxonomy5. Taking as an input such index, we construct
an indicator function, assigning the value of 1 if the value is over 75th percentile of the
distribution of the index.
5The construction is based in the work context 4.C.1.a.2.l, that relates to the frequency in which a
person needs face to face discussion to perform its job. A higher value means that face to face interactions
are more frequent.
5
3 Teleworking and employment type
3.1 National characterization
The effects of social distancing measures are not homogeneous across sectors and occupations.
The incidence of computer use on work tasks and activities, the technological level of different
sectors of the economy, and the modernization of sectors are factors that can increase or
alleviate such effects. In this section, we describe results for those occupations that are
compatible with teleworking.
When we take into consideration the results constructed using O*NET (Table 1), one-
fifth of the jobs in Colombia can potentially be done from home. The situation nevertheless
is not as good when considering the informal population, and just the 13% of the informal
population can perform his job from home. These shares are below Dingel and Neiman
(2020) estimates for the United States (37%). The reasons for the differences can be found
in the different sector composition and the different occupational structure. The percent-
age, however, is in line with Latin American countries (Bolivia 15%, Chile 25%) and other
developing economies.
When we take into consideration the results constructed from the STEP survey, the
percentage of jobs that can be performed at home is similar to the O*NET based estimation:
around 22% of works can be done from home. The behaviour of the informal share is also
similar (10%), and is considerably below the formal share (48%)6.
If the results are constructed using the O*NET or ISCO-08 classification, why they differ
between countries? The answer is because the occupational composition in each sector is
different in each country, and the share of each sector varies among regions. For example7,
regions that have a larger share of occupations in financial services have higher shares of
teleworkable occupations (Antioquia, Cundinamarca and Valle), while the regions in which
6When we compare the two indexes, they coincide in the classification for 84.3% of the occupations. See
table A2 in the appendix.
7The following results are based on the index construct using O*NET.
6
Table 1: Share of occupations compatibles with telework by population
O*NET STEP
National Formal Informal National Formal Informal
Teleworkable compatible
occupations
19.7% 35.4% 13.4% 22.7% 48.8% 10.3%
Not teleworkable occupa-
tions
80.3% 64.6% 86.6% 77.3% 51.2% 89.7%
Source: DANE-GEIH 2016 - 2019. Own calculations.
the agricultural sector prevalence the share is lower (see Figure A1). The sector composition
also has different impacts on the informal population. According to Figure 1, the sectors
in which the highest proportion of employment is concentrated in occupations that are
not compatible with teleworking (agriculture, construction, manufacture). The opposite
being also true, where the sectors more compatible with telework do not represent a large
proportion of employment (financial services). This makes that the adverse effects of social
distance measures affect in a different manner the population. The characteristics of the
population in each of the sectors reflects this, since the average age, level of education,
income and the firm’s size changes within sectors.
Table 2: Share of workers in occupations compatible with telework by sector
Sector Total nacional Formal Informal
Agriculture, hunting and forestry 2.1% 13.0% 1.3%
Mining and quarrying 16.7% 24.7% 5.4%
Manufacturing 15.4% 25.0% 6.4%
Electricity, gas and water supply 33.6% 35.0% 44.0%
Construction 8.1% 19.5% 2.8%
Wholesale, retail trade, hotels and restaurants 19.7% 24.0% 18.1%
Transport, storage and communications 13.7% 28.8% 8.0%
Financial intermediation 64.1% 62.3% 76.7%
Real estate, renting and business activities 27.7% 38.8% 11.9%
Community, social and personal service activities 37.1% 49.8% 17.9%
Source: DANE-GEIH 2016 - 2019. Own calculations.
7
Figure 1: Correlation between telework share and worker characteristics by industry
(a) (b)
(c) (d)
(a) Share of teleworkable occupations and share of workers with low income by sector. (b) Share of telework-
able occupations and share of workers with higher degree education. (c) Share of teleworkable occupations
and share of workers older than 50 years old. (d) Share of teleworkable occupations and share of workers
with full-time contracts.
Source: DANE-GEIH 2016 - 2019.
8
To provide a more detailed characterisation of the workers in work-from-home occupa-
tions, we followed Mongey and Weinberg (2020) who proposed the following approach:
yT W,i =βXi+(1)
Where yT W,i takes values of 1 if, for instance, the worker’s income is below the national
average (low labour market income) and zero otherwise; or it takes a value of 1 if the worker
has a college degree and zero otherwise. Xindicates if the person works in a work-from-home
occupation (X= 1, and zero otherwise). is the sample mean. Consequently, βshows the
fraction of workers who are in work-from-home occupation by different population groups.
Figure 2 plots the corresponding coefficients. For instance, at a national level, the fraction
of people in work-from-home occupations with low labour income is 22.8 percentage points
lower than the fraction of people in non-teleworkable occupations with a low labour income.
This result suggests that the share of low labour income workers is relatively higher for non-
teleworkable occupations. In contrast, the fraction of workers in work-from-home occupations
with an internet connection at home is almost 30 percentage points higher than the fraction
of people in non-teleworkable occupations with an internet connection at home. Overall,
these results show that vulnerable workers (e.g. Low income, low educated workers, etc.)
tend to have a lower share in work-from-home occupations.
9
Figure 2: Characteristics of workers in teleworkable occupations, by job type (for-
mal/informal)
Source: DANE-GEIH 2016 - 2019. Own calculations.
10
3.2 Households and workers characteristics
To characterise the populations that are under more stress due to social distance measures
is one of the main objectives of the present document. We contribute with evidence that
helps as input to design policies to tackle the possible effects in such populations.
Large urban areas concentrate on the largest share of economic activity and employment
in Colombia. In developing countries, agglomeration economies in urban areas are highly
correlated with informal activities and jobs. One of the advantages of using the house-
hold survey is that we can distinguish the effect that the different dimensions have, and
characterise the specific for the informal population. Informality is measured only at the
urban level, so is not possible to assess if economic activities in rural areas are more or less
associated to telework, but we can measure other characteristics for the informal population.
In order to calculate the likelihood to work in work from home occupations, we run the
following regression:
1T W,i =βXi+DR,i +DS,i +(2)
Where 1T W,i is the teleworkable indicator function, Xis a set of covariates that charac-
terise the worker and its household and Dkis a set of regional and sectorial dummies. In
order to identify the effect in the informal population, we run the regression in three different
subsets of the database: We run the first regression at the national. The second regression
with the formal workers and the last containing the informal workers.
The econometric estimation follows Saltiel (2020), in order to compare his findings for a
developing country, and our estimates from O*NET for Colombia. This choice is different
from the model presented by Mongey and Weinberg (2020), in which the observables are in
the left-hand side, and the work-from-home index is the covariate.
As in the aggregate level, the characteristics form the formal and informal population
change are heterogeneous in the population. Even if in general the results are significative
11
there are no statistical differences between the formal and informal population, with a few
exceptions. The most important is the considerable difference in the likelihood of having
a work-from-home depending on the college degree. In the national sample, the likelihood
to have a teleworkable job is around 20%. This value is in line with the findings of Saltiel
(2020) in which the estimates for the college graduate are between 20% and 30%, and are
less than the results for the U.S economy, by almost 10% (Dingel and Neiman, 2020). When
we consider the informal population, the likelihood is still positive and significant but it is
one quarter (6%) of the estimated value for the national level. Another variable for which
the effect in the informal population differs is regarding part-time employment. While part-
time employment for the formal has a large contribution to the likelihood of being employed
(20%), for the informal population, it has no importance. Another interesting finding is that
being older, if in the process of an informal job, lowers the likelihood of having a work from
home job, while the opposite is valid for a formal worker.
From the worker characteristics analysed, being an external immigrant8, having a low
labour income (income below the minimum wage), being male decreases the chances of
being in a teleworkable occupation. There is an associated penalty of being an immigrant at
the national level, but for the informal market in the case of internal migrant, the difference
disappears.
From the household characteristics, we observe that renting, living in a low-income house-
hold and having at least one children at home decrease the chances of having a teleworkable
job. The effects are small but significant. A big increase is found in households that have
internet access, for which the increase is around 10%9.
Figure 3 reports the results of the estimation. The estimates identify the effects within
region and sector. When introducing the occupation dummy, the effects disappears since
the index is constructed at the occupational level.
8External immigrant refers to foreign people living in the country. In contrast, internal immigrant refers
to Colombian people who have recently (between 5 years) move to live in another city within the country.
9In Colombia, only 42.9% of workers live in a house with an internet connection (see A1), and this feature
is highly correlated with vulnerability in other economic aspects.
12
Figure 3: Likelihood of working in a teleworkable occupation, by job type (formal/informal)
Source: DANE-GEIH 2016 - 2019. Own calculations.
13
4 Proximity
With the ending of social distance measures, what will happen with people that need to
be at proximity of others to do their job? The end of social distance measures alleviates
the pressure for the occupations that can not work for home. Nevertheless, the return to
work will put the pressure on other occupations that for their execution need proximity or
face-to-face interaction. This section presents the characterisation of such populations and
tries to reply, what are the occupations that might be affected due to the high contagion
rate and the virus spread.
In order to identify the magnitude of such effects, we make use of two different indexes.
Te first one is the proximity index Mongey and Weinberg (2020), that identify if a job to
be performed need an arm or less of distance (1.5 meters or less). The second index is
constructed following Avdiu and Nayyar (2020), which present the share of jobs that need
face-to-face interaction to be performed.
Calculating the proximity index in the GEIH, we get that almost 10% of jobs are prox-
imity occupations. The figure is slightly larger for formal workers.
Table 3: Share of workers in high personal-proximity occupations
National Formal Informal
Low-proximity 89.4% 86.6% 89.2%
High-proximity 10.6% 13.4% 10.8%
Source: DANE-GEIH 2016 - 2019. Own calculations.
As in the work-from-home occupations the effect across sectors and regions (Fig. A2)
is heterogeneous (Tab. 4). The regional disparities correlate to population density and
regional sector composition. The most affected sectors are strategic sectors for employment
in Colombia: construction, tourism and social and personal services are the sectors that
strategically will be prioritized out of the quarantine. The risk of performing jobs in such
sectors is not even, a fact that should be taken into consideration. Given that the market
14
is very tight due to the impact of the virus, the outside option for workers has decreased,
pushing them to work, putting in risk their health and their family well-being. Due to the
lack of job openings in these occupations, workers will continue to maintain their job and
exposing themselves without any compensation for such behaviour, opening the opportunity
for government intervention to compensate such failures.
Table 4: Share of workers in high personal-proximity occupations by sectors
Sector Total nacional Formal Informal
Agriculture, hunting and forestry 2.5% 6.9% 2.3%
Mining and quarrying 7.9% 8.7% 2.5%
Manufacturing 6.9% 5.4% 9.0%
Electricity, gas and water supply 11.4% 10.6% 6.9%
Construction 40.4% 28.9% 44.7%
Wholesale, retail trade, hotels and restaurants 3.5% 5.4% 2.9%
Transport, storage and communications 3.1% 6.2% 1.9%
Financial intermediation 8.6% 9.6% 1.2%
Real estate, renting and business activities 14.9% 18.1% 10.0%
Community, social and personal service activities 20.1% 22.4% 26.2%
Source: DANE-GEIH 2016 - 2019. Own calculations.
Following equation 1, we estimated the fraction of workers in high proximity occupations
by different population groups. Figure 4, plots the corresponding coefficients. For instance,
at a national level, the fraction of people in high proximity occupations with a low labour
income is 7 percentage points lower than the fraction of people in non-high proximity occu-
pations with low labour income. This result suggests that the share of low labour income
workers is relatively lower in high proximity occupations. In contrast, the fraction of male
workers in high proximity occupations is almost 6 percentage points higher than the fraction
of male workers in high proximity occupations. Overall, these result show that vulnerable
workers (e.g. Low income, low educated workers, etc.) tend to have a lower share in high
proximity occupations.
Following equation 2, we estimate the likelihood of working in a high personal-proximity
occupation. We observe that the individual and household characteristics effects are smaller
15
Figure 4: Characteristics of workers in high proximity occupations, by job type (for-
mal/informal)
Source: DANE-GEIH 2016 - 2019. Own calculations.
16
than in work from home occupations (see Figure 5). Moreover, the likelihood of being in a
proximity occupation is determined largely by the same set of variables as in the teleworkable
occupations. Age, education, a part-time job and internet access diminish the likelihood of
working in a high proximity occupation. Being an external migrant and male increases
the likelihood of working in a proximity occupation. The likelihood tend to be lower for
formal work which implies a heterogeneous effect and a larger effect on formal jobs. This
fact raises the attention and need to be an issue treated in public policy analysis since these
occupations will face a change in demand that will pressure the employment due to contagion
prevention. There need to be created measures to protect those employment relations, and
this can be done by imposing health and occupational regulations that give back confidence
to the customers, or by compensating directly the risk taken in working in such jobs10.
10Table A3 presents the occupations with higher proximity and a larger share of employment.
17
Figure 5: Likelihood of working in a high personal-proximity occupation, by job type (for-
mal/informal)
Source: DANE-GEIH 2016 - 2019. Own calculations.
When we take into consideration the face-to-face occupations, we find that 1 in 20 occupa-
tions is face to face at the national level. The largest share, as in the proximity occupations,
are disproportionately concentrated in the formal employment, where it affects almost 10%
of employment (see Table 5).
Table 5: Share of workers in occupations intensive in face-to-face interactions
National Formal Informal
Non-face-to-face occupations 94.4% 90.3% 96.9%
Face-to-face occupations 5.6% 9.7% 3.1%
Source: DANE-GEIH 2016 - 2019. Own calculations.
18
Moreover, the main characteristic that affects the likelihood of working in face-to-face
occupation is the educational attainment, for which the effect of the formal is the largest.
Having low labour income, being partly employed have decreases the most the likelihood of
being employed in a face-to-face occupation. As in the proximity occupations, the effects
are larger for the formal population, remarking the fact that this type of employment might
be the most affected after the social distancing measures end.
Figure 6: Likelihood of working in occupations intensive in face-to-face interactions, by job
type (formal/informal)
Source: DANE-GEIH 2016 - 2019. Own calculations.
19
5 Conclusion
The effects of social distance seemed to be heterogeneous across regions and sectors. The
occupational analysis gives more granular insights on how the effects are also different across
different occupations. We take into consideration the individual and household characteris-
tics and show evidence for which characteristics associates with more vulnerable character-
istics are most likely to be affected by the social distance measures. Especially, the effects
are stronger for workers of the informal sector.
When considering the effects after the social distance measures are taken off, the effect
on the formal employment might be higher. The individual and household characteristics
are also important but have the lowest explanatory power for proximity and face-to-face
index. Schemes to compensate for the risk of working in occupations that face more risk are
suggested in the presence of market failures.
The information presented aims to provide information for the construction of public
policy programs, that arrives at the identified population.
20
References
Avdiu, Besart and Gaurav Nayyar, “When face-to-face interactions become an occu-
pational hazard: Jobs in the time of COVID-19,” 2020.
Bello, Salvatore Lo, Maria Laura Sanchez Puerta, and Hernan Winkler,From
Ghana to America: The Skill Content of Jobs and Economic Development, World Bank,
2019.
C´ardenas, Jeisson, “Extracting value from job vacancy information,” 2020.
Dingel, Jonathan I. and Brent Neiman, “How Many Jobs Can be Done at Home?,”
Technical Report, National Bureau of Economic Research 2020.
Eslava, Marcela and Margarita Isaacs, “Vulnerabilidad del empleo a la emergencia de
COVID19,” April 2020, (Documento CEDE), 15.
Jaramillo, Iv´an, Diana Londo˜no, Paul Rodr´ıguez, and Andr´es Garc´ıa-Suaza, “La
vulnerabilidad del mercado laboral colombiano al COVID-19,” Technical Report, Univer-
sidad del Rosario 2020.
Leibovici, Fernando, Ana Maria Santacreu, and Matthew Famiglietti, “Social
Distancing and Contact-Intensive Occupations |St. Louis Fed,” 2020. Library Catalog:
www.stlouisfed.org.
Mongey, Simon and Alex Weinberg, “Characteristics of workers in low work-from-home
and high personalproximity occupations,” BFI White Paper, 2020.
Rio-Chanona, R Maria Del, Penny Mealy, Anton Pichler, Francois Lafond, and
Doyne Farmer, “Supply and demand shocks in the COVID-19 pandemic: An industry
and occupation perspective,” arXiv preprint arXiv:2004.06759, 2020.
Saltiel, Fernando, “Who Can Work From Home in Developing Countries?,” Technical
Report 2020.
21
A Figures and Tables
Figure A1: Share of workers in teleworkable occupations by region
Source: DANE-GEIH 2016 - 2019.
22
Table A1: Descriptive statistics
Variable Percentage
Workers characteristics
Adult (>50 years) 25.7 %
Male 58.3 %
Single 42.0 %
External immigrant 1.7 %
Internal immigrant 11.6 %
College degree 22.6 %
Urban zone 78.3 %
Job’s characteristics
Low labour income 42.1%
Large firm (>30 emp) 26.5%
Part-time employed 43.3%
One year or more working 71.3%
Contract 33.9%
Agriculture, hunting and forestry 16.4%
Mining and quarrying 0.9%
Manufacturing 11.8%
Electricity, gas and water supply 0.5%
Construction 6.4%
Wholesale, retail trade, hotels and restaurants 27.3%
Transport, storage and communications 8.0%
Financial intermediation 1.4%
Real estate, renting and business activities 7.8%
Community, social and personal service activities 19.6%
Household’s characteristics
Rents home 38.0%
Low household income 35.2%
Household with children 30.2%
Household with Internet 42.9%
Source: DANE-GEIH 2016 - 2019. Own calculations.
Table A2: Teleworkable index comparison
Saltiel (2020)
Non-teleworkable Teleworkable
Dingel and Neiman (2020) Non-teleworkable 70.9% 9.3%
Teleworkable 6.2% 13.4%
Source: DANE-GEIH 2016 - 2019. Own calculations.
23
Figure A2: Share of workers in high personal-proximity occupations
by region
Source: DANE-GEIH 2016 - 2019. Own calculations.
Table A3: Occupations that are performed with proximity and with a higher number of
workers
Occupations Percentage
Bricklayers and related workers 2.91%
Security guards 2.05%
Beauticians and related workers 1.25%
Personal care workers in health services and related workers 0.66%
Nursing professionals 0.59%
Health care assistants 0.50%
Hairdressers 0.37%
Home-based personal care workers 0.24%
Dentists 0.22%
Food service counter attendants 0.14%
Source: DANE-GEIH 2016 - 2019. Own calculations.
24
Agradecimientos
Esta serie de documentos de trabajo es financiada por el programa “In-
clusi´on productiva y social: programas y pol´ıticas para la promoci´on de una
econom´ıa formal”, c´odigo 60185, que conforma Colombia Cient´ıfica-Alianza
EFI, bajo el Contrato de Recuperaci´on Contingente No.FP44842-220-2018.
Acknowledgments
This working paper series is funded by the Colombia Cient´ıfica-Alianza EFI
Research Program, with code 60185 and contract number FP44842-220-2018,
funded by The World Bank through the call Scientific Ecosystems, managed
by the Colombian Ministry of Science, Technology and Innovation.
