Baldwin Cárdenas Fernández - Colombia más competitiva
←
→
Page content transcription
If your browser does not render page correctly, please read the page content below
Baldwin Cárdenas Fernández
ABOUT THE TASC PLATFORM The Thinking Ahead on Societal Change (TASC) Platform is an open forum where policymakers, businesses, researchers and civil society can come together to tackle some of the biggest universal challenges of the future. By providing a place for people to share perspectives, connect visions and develop new solutions, we can bring the unknowns of the future into focus and act on them today. As an independent body supported by the Government of Switzerland and embedded in the Center for Trade and Economic Integration (CTEI) of the Graduate Institute of International and Development Studies, Geneva, we leverage the latest thinking in the academic world and embrace a culture of openness, critical thinking, and global cooperation. ACKNOWLEDGEMENTS The authors would like to thank Ricardo Hausman and members of the Internationalization Mission for discussions, inspiration and guidance. We thanks SwissContact for financial support of the project “CTEI TASC Platform Telemigration”. We also thank Francisco Fernandez for research assistance. 27 February 2021
CONTENTS
SUMMARY 04
1. INTRODUCTION AND MOTIVATION 05
2. HOW MANY COLOMBIANS COULD POTENTIALLY WORK ONLINE? 08
2.1. The Dingel-Neiman method adapted to Colombia
2.1. Results from the Dingle-Nieman method applied to Colombia
3. HOW COMPETITIVE WOULD COLOMBIAN WORKERS BE IN THE US? 14
3.1. Disaggregate results for wage gaps
4. WHAT IS STOPPING THE ARBITRAGE? 16
4.1. CAGE Framework
4.2. Digital access from home
4.3. Regulatory barriers
5. FOCUS ON FREELANCING 19
5.1. A look at data from 3 important platforms
5.1.1. Discussion of the chosen platform and data gathered
5.1.2. What jobs are most in demand?
5.1.3. Broader analysis of the job postings – demand for freelancers
5.1.4. Analysis of the location of freelancers
5.1.5. Supply of freelancers: Most frequent job titles and skills offered
5.1.6. Supply and quality of Colombian freelancers
5.2. Qualitative evidence: Interviews with employers and freelancers
5.2.1. Interviews with employers
5.2.2.Stories from the workers perspective
6. OTHER ORGANISATIONAL FORMS USED IN DEVELOPING NATIONS 29
6.1. Examples in Colombia
6.2. Observations
6.3. Principal Success Factors
7. HOW OTHER NATIONS HAVE DONE IT: COSTA RICA AND PHILIPPINES 32
7.1. Costa Rica
7.2. The Philippines
7.3. Argentina
8. CONCLUDING REMARKS 34
8.1. Some policy ideas
8.1.1. Promoting Colombian to freelance
9. REFERENCES 35
SUMMARY
Telemigration is simply working from home when home is
abroad. Will telemigration be an important aspect of Colombia’s
internationalisation in the coming years? This paper makes the
case that the answer is ‘yes’. The argument is founded on three
facts. First, about 21% of Colombians have jobs that are
‘teleworkable’ (and thus potentially tradeable, given digital
technology). Second, Colombian wages/salaries in teleworkable
occupations are, on average, about a tenth that of US workers in
similar occupations (and so there are potentially large cost
savings to US firms if they outsource service tasks to
Colombians). Of course, labour costs should be adjusted by
labour productivity, but we do not have information on this.
Third, very few Colombians are currently engaged in
telemigration, judging from the data we gathered from the major
freelancing platforms that provide one way to telemigrate.
Since so many Colombians could telemigrate but so few do, and
since US firms could save so much money by hiring Colombian
workers, we conjecture that service exports can play a
significant role in the internationalisation of the Colombian
economy. The paper discusses some barriers to telemigration
and offers very preliminary indications of policies that might
promote this form of export.
04
INTRODUCTION
AND MOTIVATION
1.1. How has digitech Moreover, since distance is critical
Digital transformation is
changing the world economy changed the viable of in offshoring, the low-wage–high-
at an explosive pace. One development pathways? tech combination is on offer only to
nations sufficiently near the world’s
aspect of the phenomenon The China-path is becoming more high-tech manufacturing hubs:
which is particularly pertinent challenging due to automation and North America, the EU, and North-
to the Colombia competition. East Asia. Faraway nations have
Internationalization Mission is
little chance since the offshorable
the way in which digitech is Digitally-led automation is taking manufacturing stages are limited
changing development the jobs out of manufacturing – and there are many good
journeys, or ‘pathways to thus making industrialisation a less alternatives that are closer to the
prosperity’. attractive development pathway hubs.
(Loungani et al. 2017, Hallward-
The transformation is Driemeier and Gaurav 2017). With The India-path is becoming easier
simultaneously making the few exceptions, industrialisation is due to digital technology.
China-path (manufactured no longer enabling transformations
exports) harder and the India- based on a big share of the As digitech shuts off one pathway,
path (service exports) easier, population walking out of the fields it is opening another by making
as Baldwin and Forslid (2019) and into factories. This ‘robo- remote workers less remote – a key
argue. This paper focuses on facturing’ future is not for point since remote workers are so
the India-path and suggests tomorrow, but it is coming. When much cheaper. Digitech has
that the export of office contemplating massive enabled excellent
services, ‘telemigration’, infrastructure and industrial telecommunications, radically
could be an important aspect investments, governments should improved machine translation, and
of Colombia’s keep in mind that robots will allowed the rise of internet
internationalisation. eventually take the ‘manu’ out of platforms that do for international
manufacturing. trade in services what eBay and
Alibaba have done for international
Moreover, the nature of trade in goods. Each of these is
international competition has facilitating international price
shifted. What Baldwin (2006, 2016) arbitrage in the service sector (in
calls the ‘second unbundling’ the form of service-sector
(industrial offshoring to developing offshoring).
nations teamed with knowhow
transfers) transfigured comparative 1.2. The pandemic has
advantage by allowing producers in radically accelerated this
some developing nations to trend in four ways.
combine low wages with advanced
technology. The resulting high- An epic number of people have
tech–low-wage combinations been let go by rich-nation service
undermined the competitiveness of firms (the potential employers of
other developing nations who are Colombian telemigrants).
competing the traditional way –
with low wages and low-tech.
05The figures in America are in the Office space in Europe and is based on the fact the unit-labour
tens of millions. Since sunk firing America got more expensive due cost is lower in Colombia than it is
costs are important, firms’ trade-off to social distancing and other anti- in advanced economies in many
between local hiring and offshoring contagion requirements. sectors. That is, workers – even
is very different once the domestic after adjusting for productivity –
worker has already been let go. This won’t last forever, but it is cost less in Colombia. In the goods
likely to be true during much of the sector, however, lower labour costs
A large share of service firms in recovery phase. This also shifts the are necessary but not sufficient for
Europe and America have learned trade-off between local competitiveness. Manufacturing is
to work with remote teams (Kilic employment and offshoring in the complex and subject to massive
and Marin 2020). direction of more telemigration. agglomeration economies that
Finally, corporations piled on debt make it hard to break into. As a
Estimates are that from 40-70% of that means they’ll be under intense result of these powerful
workers in the US and Europe cost-cutting pressure going agglomeration forces, almost 80%
worked from home due to the forwards. of all manufacturing is in three
pandemic (Berg et al. 2020, Dingel regional hubs (37% in East Asia, and
and Nieman 2020). Rich-nation 1.3. Telemigration as an 19% each in North America and the
firms have invested in hardware, attractive addition to EU).
bandwidth, collaborative software export strategies
(and training), and secure online Several features of services Services are quite different in that
databases, and managers have suggest that the India-path may be many service sectors are subject to
learned how to manage virtual an attractive route for nations, like much less intensive agglomeration
teams. Quite simply, the Great Colombia, who are not forces (Loungani et al. 2017, and
Lockdown shifted where we do geographically close to the Baldwin and Forslid 2019).
office work, and how we do it. industrial hubs. But before turning Moreover, the means of
Employees have had to learn to to these merits, it’s worth pointing ‘transportation’ and the ‘logistics’ of
work remotely, their managers have out how the China-path and India- services are – thanks to digital
had to learn how to manage path are fundamentally similar. The technology – easier to get right.
remotely. Everyone has had to underlying source of comparative
invest in the hardware, software, advantage is the same for both To put it differently, trade in goods
and training that make it possible to (Figure 1.1). is a veil for trade in labour services,
create and deliver value from but the veil – setting up factories
remote locations to customers who Comparative advantage is really all and transport infrastructure,
have had to learn how to buy and about the relative costs of establishing an industrial base, and
use services that are delivered by production. For most developing breaking into foreign markets – is a
dispersed teams. nations, their relative cost not a simple thing. Industry is a
advantages stem from relatively classic chicken-and-egg case
Anything that makes it easier to abundant labour or natural whereby a nation’s industry can be
telework domestically tends to resources. Leaving resource competitive if it has a sufficiently
facilitate telemigration. sectors aside, the dominant cost large industrial base, but getting
advantage of a nation like Colombia the base is hard without the
US firms will realise that they can competitiveness.
get some of the remote work done
at much lower cost by hiring Figure 1.1: Service and goods export tap into the same comparative advantage
workers sitting in low-wage
countries. While it is certainly true
that foreign talent working online is
unlikely to be as good as domestic
talent working in the office, the
foreign talent may be a whole lot
cheaper. This is just service-sector
globalisation, so the deeper point is
Covid-linked changes are lowering
the technical barriers to this type of
globalisation.
Note: Exports happen when things are cheaper in the exporting nation; for most things the sources of
the cost-edge are wages that are lower even when adjusted for productivity. Exploiting this edge in
goods (manufactured exports) requires complex organisation and is subject to important bottlenecks
and large ‘minimum efficiency scales’. Exploiting it in services is not trivial but involves much
complexity and coordination. Source: Based on Baldwin and Forslid (2019).
06When it comes to services, by Finding that Colombians could do outsourcing, KPO). In this paper, we
contrast, the veil is much lighter. work for foreign clients (i.e., export focus on the easiest option in terms
Digital technology allows service their services) and that their current of fixed costs and other entry
providers in one nation to sell to wages suggest that they would be barriers – online freelancing
services buyers in another nation cost competitive if they did so, do platforms like Upwork.
without elaborate investments. The not mean that these new exports
trend has been there for many will actually appear. The implicit The section also presents examples
years, making remote workers less cost arbitrage has been apparent of such firms that are operating in
remote, so to speak (Kilic and Marin for years – and indeed some foreign Colombia and draws some
2020), but the pandemic firms are already exploiting it, but in observations from the examples.
accelerated it. Covid-19 has made a small way. The hope is that the
it even easier to sell services at Covid-accelerated digital The last section, before our
distance and thus made the veil transformation of the US service concluding remarks, briefly
even lighter. sector will open the door to broader presents three case studies of
arbitrage and the service exports countries that have made a success
1.4. Outline of the paper such arbitrage would create. of service-sector exporting: Costa
Rica, the Philippines, and (less so)
Sections two and three use labour Section six considers non- Argentina.
market data to make the prima freelancing vehicles for
facie case that Colombia has a telemigrating. Traditional trade in
comparative advantage in service services often involves large
exports. The fourth section uses companies that either send the
the CAGE distance framework to skills and talent abroad (say, hotel
organise thinking about why chains and financial sector firms) or
Colombia’s service exports are so bring the customers to Colombia
modest to date. (tourism, transport services, etc.).
Telemigration is meant to refer to
Freelancing is the focus of the fifth something quite different. It is
section. It looks at how many Colombians working for foreign
Colombians work in service-sector companies while sitting in
jobs that could be done remotely Colombia. There are many ways in
with current technology. We which this exporting could be done
identify these workers by using the including Shared Services Centres
Dingle-Neiman method to single set up by multinationals, or
out occupations whose job- specialised Colombia-based firms
descriptions suggest that ‘work that sell services to foreign
from home’ is possible, i.e., they are customers (e.g., call centres, and
already working in jobs that are business processing outsourcing,
‘teleworkable’. We argue that BPO, and knowledge processing
Colombian service workers are
likely to be competitive in the US,
by comparing US-Colombia wage
gaps by occupation. The size of the
gap provides a rough estimate of
the gains that US-based firms could
realise by outsourcing service tasks
to Colombia-based workers.
072. HOW MANY
COLOMBIANS COULD
POTENTIALLY WORK
ONLINE?
The basis of such differences has for the Colombian service sector.
This section looks at what
long been left to judgement and This sets the baseline expectations
might be called the ‘export
received wisdom, but it was always for how much service exports could
potential‘ of the Colombian
founded on the idea that things expand.
service sector. Many services
cross borders much more easily
are non-tradeable – for
than workers. Services that require The Dingel-Neiman procedure uses
example, most government
people to be face-to-face responses to surveys undertaken
services, and household and
(haircutting) or face-to-things by the US Department of Labour
personal services. Other
(household cleaning) or are only that were designed to understand
services, by contrast, are
used by people located in one the nature of American jobs. US
highly tradeable, say financial
nation (issuing local driver’s workers, whose jobs are classified
services and international
licences) were categorised as non- very finely (into more than 1,000
transportation services.
traded. Things, like financial occupations) answered questions
services, that could be done by about the nature of their jobs.
correspondence (in the old days), Dingel and Neiman used the
or digitally in recent decades, were answer to specific questions to
classified as potentially tradeable. qualify particular occupations are
‘teleworkable’ or not. For instance,
A step towards quantification came if the respondent answered yes to
in the 2000s, in the height of the these questions, their occupation
offshoring debate in the US: Alan was classed as not teleworkable
Blinder developed a method to (i.e., cannot be performed from
determine which US occupations home): “Use email less than once
were ‘offshore-able’. His ballpark per month,” “performing general
figure for the US was that 30-40% physical activities is very
were at threat. These numbers important.”
would have been alarming in any
case, but he added to it by making Once they had a list of occupations
the classic mistake of assuming the that could be done from home,
future would arrive faster than it Dingel-Neiman uses the US Bureau
actually did. Call it Blinder’s blunder of Labour Statistics (BLS) data to
– a trap that this paper will strive to determine how many workers were
avoid. in teleworkable occupations. The
BLS data also contains a wealth of
2.1. The Dingel-Neiman information about the workers
method adapted to including educational attainment,
Colombia gender, location, and wage/salary.
Their bottom-line aggregate figure
We use the latest techniques – a
was that 37% of jobs are
procedure developed by Dingel and
teleworkable (at the high end of
Neiman (2020) – to identify which
Blinder’s guesstimate from 2006).
Colombian services are tradeable
and thus potentially exportable. In a
Bringing this methodology to
sense, we are creating a new,
Colombia is not a simple matter.
customised definition of ‘tradeable’
08In Colombia, there are major their job from home. That is 21% of and crosswalks imperfections. The
barriers (e.g., informality) to using the labour supply. For comparison, percentages in this presentation
this methodology to calculate the Dingle and Neiman (2020) estimate assume that missing observations
share of teleworkable jobs directly. that 37% of US workers have jobs behave as observed values; in
The Colombian household survey that can plausibly be performed at other words, the shares are
classifies occupations according to home. We take this figure as calculated over observed data.
a 1970s classification that is indicative of the amount of work Third, 2019 data is used in all
outdated and difficult to match with that might be performed online for exercises to avoid temporary
the International Standard clients abroad, i.e., exported (Table effects caused by Covid-19 on the
Classification of Occupations 2.1). occupations.
(ISCO). However, Cárdenas (2020)
recodes the raw data from the To make this aggregate figure more To highlight the enormous
household survey at the level of useful for policy purposes, we variability across broad categories
ISCO 4 digits. Using Cárdenas break this down by occupation. As of jobs, we present the figures in a
(2020) and Dingel and Neiman Table 2.2 shows, the most chart form (Figure 2.1). To address
(2020) recoded in ISCO, we were teleworkable occupations fall in the the common belief that
able to identify the number of ‘professionals‘ category, followed telemigration is something that only
workers that perform occupations closely by clerks and technicians – workers with advanced education
that are compatible with work from all with shares that are 50% or can undertake, the chart also
home. We also included workers higher. At the other end of the shows the share of teleworkable
who are unemployed but have spectrum are occupations that jobs in the broad categories that
experience in such occupations. clearly involve workers being in the require tertiary education.
physical presence of someone or
2.2. Results from the something. For example, the share The results are striking. In some
Dingle-Nieman method of teleworkable jobs in the types of jobs, telemigration is really
applied to Colombia elementary occupations, plant and not an option. In the primary
machine operators, and craft sectors, workers have to be
According to this procedure,
workers is below 5%. Senior physically together with things like
around 3,700,000 Colombian
officials and managers are in land and machines, and the same is
workers, or about 20% of the
between with a share near the US true for plant and machine
employed population, are currently
figure. operators, elementary occupations,
working in occupations that could
and craft and related trades
be performed from home (Gran
There are a few provisos to these workers. Other professions, like the
encuesta integrada de hogares
calculations. First, the crosswalks military, are non-tradeable by
[GEIH] 2019). Since we are
used in this exercise are sensitive nature (at least in times of peace).
addressing potential outcomes, we
to the level of development, and
add to this figure around 740,000
the abilities required to perform The teleworkable jobs are
unemployed people who have
certain occupations are not concentrated in what might be
experience, or are looking for a job,
necessarily the same in developed loosely called office jobs or desk
in those teleworkable occupations.
and developing countries, jobs. Among technicians and
This is 33% of the unemployed.
particularly among occupations that associate professionals, 50% of
require low levels of education. Colombian workers have jobs that
Taken together, we estimate that
Second, missing information arises are potentially appropriate for
around 4,451,303 Colombian
from raw observations in the GEIH telemigration. Only 65% of these
workers could perform
that could not be coded as ISCO, jobs also require higher education.
Table 2.1: How many Colombians work in teleworkable occupations?
US Colombia
Employed Unemployed with
Dingel-
(Cárdenas- experience or Total
Neiman
Montaña) looking for jobs
% of occupations compatible with telework 37% 19.7% 32.6% 21%
Non-compatible with telework 15,107,253 1,530,841 16,638,094
Compatible with telework 3,710,847 740,455 4,451,303
Not identified 3,469,180 343,663 3,812,843
Working-age population 22,287,281 2,614,959 24,902,240
Source: Dingel & Neiman (2020), Cárdenas & Montana (2020) & authors calculations using GEIH (2019). *Shares over observed data.
09Table 2.2: Jobs that can be performed from home among labour force (employed + unemployed) and over occupation (ISCO 1D)
Total % Compatible
Compatible Missing (including (excluding
missings) missing)
Professionals 1,471,792 283,441 2,593,180 64%
Science and technology professionals 525,668 150,446 1,277,718 47%
Law professionals 296,353 65,322 466,837 74%
Teaching and education professionals 649,771 67,673 848,624 83%
Clerks 726,505 366,950 1,764,917 52%
Technicians and associate professionals 644,738 267,199 1,553,034 50%
Legislators and managers 312,626 119,056 828,633 44%
Managers 260,124 99,076 734,303 41%
Legislators 52,502 19,981 94,331 71%
Service workers and shop and market sales
1,006,005 235,404 5,801,152 18%
workers
Craft and related trades workers 102,814 358,187 3,054,462 4%
Plant and machine operators and assemblers 63,371 149,376 2,177,527 3%
Elementary occupations 121,926 535,198 4,525,231 3%
Skilled agricultural and fishery workers 1,301 63,435 1,168,836 0%
Military forces 224 1,422 2,094 33%
Missing 0 1,433.174 1,433.174
Total 4,451,302 3,812,843 24,902,239 100%
Source: Authors’ calculation, details available upon request.
Figure 2.1: Shares of teleworkable jobs by broad occupation and education level
Professionals
Clerks
Technicians and associate professionals
Legislators and senior offcials managers
Service workers and shop and market sales workers
Skilled agricultural and fishery workers
Craft and related trades workers
Plant and machine operators and assemblers
Elementary occupations
Military forces
Teleworkable Teleworkable and tertiary education
Source: Authors’ calculation, details available upon request
10The highest share of teleworkable Map 2.1: Teleworkable jobs by region 1
jobs is among occupations classed
as professionals, and here, 88% of
Colombians with teleworkable jobs
in this category also have education
beyond high school.
The importance of higher education
is not universal. In the categories
for service workers and shop and
market sales workers, only 24% of
the teleworkable occupations are
associated with tertiary education.
The geographic distribution of
teleworkable jobs is highly skewed
towards urban areas, as might be
expected (Map 2.1).
Turning to the nature of the
workers that have teleworkable
Map 2.1: Teleworkable jobs by region
jobs, we see that overall, the
possible export possibilities are
heavily aligned with education Figure 2.2: Jobs that can be performed from home among labour force (employed +
attainment (Figure 2.2). A number unemployed) by education level
of aspects are worth highlighting.
Note that a beyond-high-school
education is a good indicator of
teleworkability for some types of
jobs but not for others. In the
professionals category, 90% of
teleworkable jobs require higher
education. The opposite is true in
‘service workers and shop and
market sales’ workers. In four of the
broad categories, higher education
is associated with teleworkability
but not in the rest. Source: Authors’ calculation, details available upon request.
The opportunities are also not
evenly spread by age but Figure 2.3: Jobs that can be performed from home among labour force (employed +
telemigration will not, according to unemployed) by age
these suggestive calculations, be
only for those under 40. Around
55% of the potential is for workers
over 40 years old (Figure 2.3).
As far as the gender breakdown is
concerned, we found that on
average, i.e., considering the whole
labour force, women have
professions with higher
Source: Authors’ calculation, details available upon request.
1 The high level of teleworkable activities in Guaviare is related to the number of sales assistants.
11Table 2.4: Jobs that can be performed from home among employed by formality status and ISCO 1D
Informal Formal Total
% / total jobs
% Works performable % Works performable % Works performable
from home from home from home
Managers 30% 61% 44%
Professionals 59% 66% 64%
Technicians and associate professionals 46% 53% 50%
Clerical support workers 54% 51% 52%
Services and sales workers 19% 15% 18%
Skilled agricultural, forestry and fishery 0% 1% 0%
Craft and related trades workers 2% 10% 4%
Plant and machine operators and assemblers 1% 7% 3%
Elementary occupations 2% 6% 3%
Total 14% 33% 19%
compatibility with remote work than Two more slices of our calculations Table 2.5 shows the breakdown by
men. Overall, about 16% of men shed light on the sorts of jobs that sectors of the economy where the
have teleworkable jobs while about are plausibly offshorable and thus various occupations are listed. The
26% of women have such open to exporting. Table 2.4 shows highest shares are in the high-end
occupations. Workers with informal that there is not a dominate pattern service sectors such as Financial
jobs are about half as likely to have in teleworkability in terms of and Information and
teleworkable jobs (about 11% formality or informality of the job, Communication. Among the service
compared to 20% for the whole although there is evidence of a sectors, the largest number of jobs,
workforce). This observation is difference for managers . It is also however, are in the Commercial
important since women and young worth noting that teleworkability is sector – almost a million.
workers have been especially hard twice as high for formal jobs than it
hit by Covid-19. is for informal jobs.
12 2 Informality is defined using the social security contributions criteria.3
Table 2.5: Jobs that can be performed from home among labour force (employed + unemployed) by DANE 1D sector
% work Number of jobs Number of Total
DANE Title performable performable jobs not (including
from home from home performable missing)
from home
Financial 63% 162,465.80 95,579.80 316,763.50
Information and communication 56% 140,223.10 109,761.40 323,212.60
Public administration 54% 1,097,299.80 941,745.50 2,529,626.30
Professional, scientific and technique 31% 351,745.40 773,155.90 1,381,835.20
Real estate 30% 765,427.30 1,758,248.30 282,711.60
Electricity, gas and water distribution 25% 34,805.00 106,098.40 197,180.10
Commercial 25% 946,735.10 2,893,159.80 4,250,235.80
Mining 17% 18,777.00 93,460.20 196,204.00
Industry and manufactories 15% 296,144.90 1,646,000.10 2,503,549.00
Other 13% 234,409.40 1,511,012.20 2,062,864.20
Transportation and storage 10% 140,046.60 1,262,656.70 1,545,165.40
Construction 9% 114,175.10 1,222,405.40 1,521,267.80
Hotel and accommodation 6% 89,017.80 1,337,195.00 1,655,384.30
Agriculture 2% 60,030.20 2,887,615.30 3,521,280.40
Total 4,451,302.5 16,638,094 22,287,280
3 CIIU rev 4. 12 groups 133. HOW COMPETITIVE
WOULD COLOMBIAN
WORKERS BE IN THE
US?
In this section, we turn to quantifying the wage gap between Colombian and US workers in various
occupations. Occupational Employment Statistics (OES) is a good source for wages in the US. The
difference in hourly wages (across sectors) between Colombia and US for jobs that can be performed
from home is estimated using the GEIH (2019) and the OES (2020) 4 . Specifically, Colombian hourly
income is converted from Colombian peso (COP) to USD using the average annual exchange rate of
COP 3,281 per USD 1 for 2019.
On average, a worker in Colombia service-sector tasks from US to now. Table 3.1 shows that the
earns USD 2.2 per hour working on offices to Colombian service ratio ranges from almost 15 times in
occupations that are teleworkable, workers would have the potential to the ‘legislators, senior officials and
while the average wage for such be cost saving. managers‘ category to about 12
occupations in the US is USD 25, times in the elementary
i.e., more than 11 times higher. Of occupations (which are not very
course, the same occupation 3.1. Disaggregate results teleworkable, as we saw).
performed in the US and Colombia for wage gaps
may be very different, but the very
The aggregate averages hide
large gap suggests that outsourcing
important differences that we turn
Table 3.1: Wage ratios for various occupations, US versus Colombia
Hourly wage USD (median) – Wage ratio
teleworkable jobs Wage ratio (US to
ISCO title
(US to pesos) Colombian
Colombia (GEIH) USA (OES) PPP)
Legislators, senior officials and managers 3,52 51,4 14,6 6
Professionals 3,52 33,1 9,4 3,9
Technicians and associate professionals 1,76 24,4 13,9 5,7
Clerks 1,47 17,4 11,8 4,9
Service workers and shop and market sales
1 12,9 12,9 5,3
workers
Craft and related trades workers 1,37 17,2 12,6 5,2
Elementary occupations 1,16 13,5 11,6 4,6
Weighted average (ISCO 1D level) 2,2 25 11,8 4,9
Median 1,7 30 17,6 7,3
Notes: Source: GEIH (2019) and OES (2020). The difference in hourly wages (across sectors) between Colombia and the US for jobs that can be
performed from home is estimated using the GEIH and the OES. Colombian wages are converted from COP to USD using the average of the annual
exchange rate of 2019 (COP 3,281 per USD 1) and a PPP ratio of (COP 1,349 per USD 1) using the average of the PPP exchange rate from the World
Bank’s WDI for 2019. Colombian hourly wages are estimated by dividing annual labour income by 52 times the reported weekly working hours.
4 Note Colombian monthly income includes monthly income in kind; plant and machine operators and assemblers do not appear in the results
14 because very few workers with these occupations can work from home.The job classification in this table is To address this, we selected the Colombia. Table 3.2 shows that for
highly aggregated, but it establishes sectors with the largest median a selection of occupations at the
a ballpark figure of US wages being hourly wages for jobs that can be four-digit level, the ballpark
10-15 times higher than Colombian performed from home and compare estimate is not too bad.
wages. Does the ballpark estimate them to the respective median
hold for more finely defined hourly wage in the same sector in
occupations?
Table 3.2: Wage ratios for selected highly teleworkable occupations, US versus Colombia
Hourly wage USD (median) –
teleworkable jobs Wage ratio
ISCO 1 ISCO title (US to
Colombia (GEIH) USA (OES) Colombian)
2144 Mechanical Engineers 3.7 50.3 13.7
2152 Electronics Engineers 5.1 50.5 9.8
1212 Human Resource Managers 3 51.7 17.2
2611 Lawyers 4.9 52.7 10.8
1211 Finance Managers 6.2 54.2 8.7
1221 Sales and Marketing Managers 7 55 7.8
Communications Technology Services
1330 7.3 59 8.1
Managers
Source: Authors’ calculation, details available upon request.
154. WHAT IS
STOPPING
THE ARBITRAGE?
4.1. CAGE Framework appreciate readily since they are
When there is supply and
negotiated in detail in every trade
there is demand but there are When it comes to trade in services, agreement.
few transactions, something physical distance is not much of an
must be wrong, or more issue (apart from time zones), and Physical distance, by contrast, is of
precisely, something must be there are rarely tariffs or other little import once the service
hindering the transactions. taxes to hinder the trade. This is supplier has a computer and good
The first instinct of not to say the trade is unhindered. internet access. Traditional barriers
economists is to look for Ghemawat (2007) has developed a – tariffs and quotas – are also
barriers to transacting – broader conceptualisation of mostly non-existent for many types
tariffs, quotas, regulations, distance called the CAGE distance of service exports that are
and the like. But it is also framework that points to Cultural, delivered across borders (so-called
possible that the supply does Administrative, Geographic and Mode 1 services). The WTO’s e-
not match the expectations of Economic differences across commerce moratorium has banned
the demanders. Logically, nations. countries from imposing customs
there can be demand-side
duties on electronic transmissions
problems, but in the case of As with all these frameworks for since 1998. This is supported to a
the internationalisation of organising complicated things, its large extent by the practical
Colombia’s service providers, merit is mostly in helping difficulty officials would have in
this is of second-order policymakers consider a broad collecting such duties. Moreover,
interest. Of course, the range of issues and realise that WTO advance economy members
factors all intertwine, and it is there is no silver bullet. There is no committed to imposing no barriers
the combination of difficulties one trigger to pull. on Mode 1 services in many areas
that really matters – see
back in 1994 – in part because they
Dubuque (2021b, c), but here When looking at service exports, never believed that developing
we focus on barriers using CAGE helps encourage a shift in nations would be competitive in
the well-known CAGE mindset away from what is these sectors.
framework. important for trade in goods. When
it comes to goods, the Our original surveys of freelancers
overwhelming importance of reveal many barriers that the CAGE
geographical distance is an framework is useful in organising.
empirical fact. More or less, Likewise, when we discuss success
doubling the distance between stories in other nations, the
trade partners lowers the value of framework’s usefulness is apparent.
trade by half (Head and Mayer There is one very obvious barrier
2014). When it comes to trade in that is important in the area under
services, the dictatorship of study – communication links.
distance is powerless. Geographical
distance is a natural barrier, but
manmade barriers to trade in goods
are well-known to be important –
something that policymakers
16Table 4.1: More subtle barriers to trade: Ghemawat’s CAGE framework
Wage ratio Cultural Distance Administrative Geographic Economic
(US to Distance Distance Distance
Colombian)
Different languages Lack of colonial ties Physical distance Rich/poor differences
Other differences in
cost or quality of
Different ethnicities; lack
Lack of shared regional natural, financial,
of connective ethnic or Lack of land border
trading bloc human resources,
social networks
infrastructure, or
knowledge
Unilateral
Different religions Lack of common currency Differences in time zones
Differences in
Lack of trust Political hostility climates/disease
environments
Different values, norms,
and dispositions
Non-market/closed
Insularity economy (home bias vs. Landlocked Economic size
foreign bias)
Lack of membership in
Lack of internal
Traditionalism international Low per capita income
navigability
organisations
Bilateral
Weak institutions,
Geographic size
corruption
Geographic remoteness
Weak transportation or
communication links
Source: Wikipedia https://en.wikipedia.org/wiki/CAGE_Distance_Framework
4.2. Digital access from There are other ways of accessing 4.3. Regulatory barriers
home these essential digital tools, but the
low figures suggest that technical One of the most important
Many Colombian workers face barriers inside Colombia will need regulatory barriers that
important hindrances to to be addressed before telemigration exports could face is
telemigrating from home – even if telemigration becomes a major labour inflexibility, particularly
their job features allow it. Of the vehicle for internationalisation. concerning hourly jobs. In
labour force that could work from Colombia, it is possible to work by
home, just 54% have a computer hours, provided that the worker
and 66% have internet access earns a wage equal or higher to the
(Table 4.2). monthly minimum wage adjusted to
worked hours. However, in order to
Table 4.2: Jobs that can be performed from be formal, the worker and the
home among labour force (employed + unemployed) by digital accessibility
employer (if there is an employer)
must contribute to social security.
Job is Job is not Missing This contribution is estimated over
teleworkable teleworkable a monthly base income that cannot
be lower than the monthly minimum
Computer at home (%) 54% 26% 37% wage.
Internet access (%) 66% 39% 49%
Therefore, a person that works
Total 4,451,303 16,638,094 3,812,843 part-time at a low wage is very
Source: Authors’ calculation, details available upon request.
17likely to pay a disproportionally contributions and special provisions platforms would have incentives to
high part of their income in social for social security that create make them work full-time. Not less
security, as can be observed in distortions and perverse incentives. important is the risk of considering
Figure 4.1 5 . In order to reduce this For example, they consider that the freelancers as dependent workers,
problem, the Congress of Colombia platform should pay the Social which would lead to much higher
recently approved the Minimum Protection Floor contribution when inflexibility and social security
Social Protection Floor (Piso workers earn less than a minimum contributions. Eventually, this could
Mínimo de Protección Social). This wage. Otherwise, workers should end any telemigration exports
scheme obliges 6 workers that earn make their own contributions. As a initiatives.
less than a minimum wage – but consequence, workers would have
equal or higher than the monthly incentives to work part-time and
minimum wage adjusted to worked
hours – to contribute to a periodic Figure 4.1: Minimum contributions to social security benefits for independent workers
savings accounts scheme. These
savings are rewarded with a 20%
subsidy at retirement age.
However, workers under this
savings scheme will not be
considered formal. Workers that
earn more than a minimum wage
should pay the regular
contributions (between 12% and
14%, depending on whether they
are considered service providers or
not). It should be noted that in
Colombia, half of the workers
earned less than the minimum
wage in 2019.
Another possible source of legal
barriers that might restrict the
exports of telemigration services is
the legislation projects that are Source: Fernández and Mejía (2021).
circulating in Congress, concerning
digital platforms. These projects Figure 4.2: Minimum contributions to social security benefits for independent workers and waged
were designed with transportation workers
and delivery platforms in mind but
might easily include freelance
platforms. Most of these projects
include additional taxes and
Source: Fernández and Mejía (2021).
5 The weekly hiring contract created some years ago, but not yet used extensively, partially reduces the problem, but this is only available to cover
salaried workers. We assumed that the freelancers are independent workers.
6 The contributions under the Social Protection Floor are obligatory to workers that have several sources of income, as we assume is the case of
18 freelancers and workers under a service providers contract, and voluntary for other independent workers.5. FOCUS ON
FREELANCING
From the hiring-firms’ perspective ProFinder. A Chinese freelancing
We turn now to a specific
(the ‘importers‘), these platforms platform, Zhubajie, has more than
form of service exporting –
have dramatically lowered the fixed 16 million freelancers registered,
the hiring of Colombian
cost of hiring foreign service and it recently launched an English-
freelancers on digital
workers while at the same time language version called
platforms such as
dramatically raising the flexibility of WitMart.com. The phenomenon is
Upwork.com. These
such contracts (MGI 2016). also growing fast. A recent study
platforms – which are very
Specifically, they radically lower the estimates that the number of
much like eBay but for
international transaction costs freelance projects that are online
services rather than goods –
related to things such as search, worldwide has been expanding at
have created new ways of
employment contracting, foreign about 26% per year for the last few
offshoring service tasks by
exchange issues and risks, years (Kassi and Lehdonvirta 2017).
making it easy for firms to
international payments, and non-
find, hire, manage, and pay
payment and non-delivery issues. This ‘online offshoring‘ is quite
foreign-based freelancers.
different from the traditional trade
From the freelancers’ perspective in services (Mode 1), and traditional
(the ‘exporters‘), these platforms service offshoring. Much Mode 1
have opened markets that were trade in services is dominated by
previously almost completely multinational firms that specialise in
closed to them (Kuek et al. 2015). high-skilled services (Mann 2017),
Online freelancing is creating many and the same holds for service
new opportunities for sufficiently offshoring (Infosys, Wipro, etc).
skilled service workers in emerging Moreover, the firms offshoring jobs
markets (ADB 2018, Kuek et al. also tend to be large, due to the
2015). fixed costs of organising and
managing offshored back-office
The phenomenon is already operations, call centres, etc. The
important. The largest of these radical reduction in the cost of
‘matchmaking‘ sites is Upwork.com hiring foreign freelancers that
with 14 million users in 180 comes with these new platforms
countries (processing over a billion seems to have changed this. While
dollars in freelancer billings, the evidence is anecdotal, the
according to information on platforms seem to have expanded
Upwork.com). There is, however, the range of tasks that can be
plenty of competition; platform economically offshored while
competition is underway. Dozens of making offshoring profitable even
start-up competitors like for micro-firms.
TaskRabbit, Fiverr, Mechanical
Turk, PeoplePerHour, and
Freelancer.com are vying for
market share. And recently,
LinkedIn – with its 450 million
registered business professionals –
has entered the race with its
195.1. A look at data from Given the above considerations, we the labour supply side. The
three important platforms have followed three criteria to information on the demand side
select the job portals to analyse the includes job titles, wages
This section looks at data scraped freelancer labour market: requested, location, skills, etc.
from three important freelancing These are collected daily over a
platforms. Traffic ranking prescribed period. Once the job
Volume of vacancies/CVs portals have been scraped, the next
5.1.1. Discussion of the chosen Website information step is to clean and organise the
platform and data gathered data.
Online job vacancies are a rich We used www.Alexa.com to
source of labour-market identify the data of job portals for Employers and job portals manage
information since these sources freelancer 7 . This search shows information according to different
can provide quick and relatively that Freelancer.com and criteria. For instance, some
low-cost data about employers’ Upwork.com are the most widely employers or websites post the
requirements. Vacancy information used job portals around the world. wage offered in dollars, others in
is publicly available on the job Moreover, these well-known pesos and so on. Furthermore, the
portals but because the information websites have a considerable wage variable may be displayed in
is on different websites, it is number of job vacancies and CVs words, ranges, or an exact number.
necessary to apply distinct and registered. For instance, Freelancer Consequently, to a certain extent, it
relatively novel techniques to reported in November 2020 around is necessary to standardise the job
gather this information. In 17,150 job postings. Likewise, both portal information to conduct a
particular, web scraping is a well- job portals contain ‘good quality‘ proper statistical analysis.
known technique to collect a information. Employers can report
massive amount of data from the detailed information about the job The standardisation process varies
internet (in this case, job portals). requirements. between websites. As mentioned
Briefly, the web-scraping before, job portals manage
algorithms automatically recognise Given that Freelancer and Upwork information according to their
certain pattern or fields in the are Australian and US platforms, criteria. Thus, some variables may
online job vacancies on the job respectively, these sources may be be available in some job portals
portals (e.g., job description, job biased to those regions. (e.g., sector), while in other
titles, wages, etc.) and download Consequently, to increase Latin websites that information may not
the corresponding information. American coverage, we have also be available or may have different
selected Workana.com, a well- categories. Moreover, the job
While web scraping yields valuable known Argentinian platform with a portals may display the labour-
information, gathering information relatively high number of job market information in different
from all the job portals is vacancies and CVs. This website languages. For instance, in most
challenging for a variety of reasons. has a well-defined structured and cases, Freelancer.com displays the
First, as every job portal has its employers can post detailed task vacancies in English or Spanish.
unique structure (HTML, JavaScript, requirements. These job portal characteristics
etc.), it is necessary to program an make standardisation challenging.
algorithm for each website that The next step is to scrape and Given that the job portals have
recognises the corresponding job organise the information for different structures and use
portal structure and collects the analysis. We use three algorithms different languages, we cleaned
relevant information. As the number to automatically collect the job and standardised the information
of job portals scraped increases, vacancies, which we think of as the within each job portal. The
the programming effort needed to labour demand side, and workers’ following presents a statistical
fit all the collected data together information, which we think of as analysis by job portal and language.
increases. Second, it is difficult to
know exactly how many job portals Table 5.1: Number of observations
by job portal (November – December 2020)
for freelancers are actually
available on the internet. Third, as
Job portal Vacancies Job seekers
jobs can be posted on more than
one site, and freelancers may offer Freelancer English 20,137
their talents on more than one site, 11,001
issues of duplication grow with the Freelancer Spanish 1,331
number of portals scraped. Fourth
Upwork 53,986 6,699
is the problem of fraudulent or fake
websites, which obviously should Workana English 751
be considered. 387
Workana Spanish 384
Source: Freelancer, Upwork and Workana. Own calculations.
20 7 Alexa Internet, Inc., is a subsidiary of Amazon.com that estimates and ranks the data traffic of websites based on the browsing behaviour of internet users.Table 5.1 shows the number of standardised classifications we offering services) is even higher,
observations by job portal 8 . The used for the teleworkability namely 56.3%. The numbers are not
number of vacancies scraped from calculations. There is really no way quite as lopsided on the other two
Freelancer is 21,468 (20,137 in around this without an enormous platforms, but the category is the
English and 1,331 in Spanish), while matching exercise.10 biggest on all three by a large
the number of job seekers on this margin. Note that there is a natural
website is 11,001. Upwork 9 For the current paper, we settle for specialisation by platform in terms
vacancies and CVs (i.e., job a more descriptive approach that of skills, even though these three
seekers) are 53,986 and 6,699, suggests the sort of skills and tasks sites are broad in terms of CVs
respectively. Finally, the number of that are demanded and supplied on offered. Applications programmers
vacancies scraped from Workana is these platforms. are in second place, and taking the
1,135 (751 in English and 384 in two categories together accounts
Spanish), while the number of CVs As Table 5.2 shows, the jobs listed for between 30-40% of the jobs
on this website is 387. are highly concentrated in just six posted on all the sites.
occupations. As mentioned, these
5.1.2. What jobs are most in are not formal occupations in the The next two most popular
demand? traditional sense; they are ‘skill occupations are more creative or
The freelance sites have job bundles’ on the worker’s side and human-oriented, namely advertising
posters (employers) and job ‘task bundles’ on the employer’s and marketing professionals and
seekers (the freelancers). These side. graphic and multimedia designers.
can be thought of as indicators of These are followed by two very
labour demand (postings) and Looking at the Freelancer figures to specific skills: translators and data
labour supply (freelancers). So start with, the dominance of web scientists.
which types of occupations have and programming is clear. Of the
most job postings, i.e., have the jobs posted, 30.4% fall into the Taken together, the six categories
highest demand? A key limitation of ‘web and multimedia developers’ account for 65-70% of the jobs
our approach is that the jobs category. The corresponding figure posted, and between 60% and 90%
postings do not follow the same for the supply side (freelancers of skills offered.
Table 5.2: Share of job postings by ‘occupation’ for Freelancer, Upwork, and Workana
Share of total
Freelancer Upwork Workana
postings
Jobs Jobs Jobs Jobs Jobs
CVs CVs CVs
posted posted posted posted posted
offered offered offered
(English (Spanish (English (English (Spanish
Web and
multimedia 30.4% 30.6% 56.3% 28.0% 20.9% 31.7% 27.9% 21.5%
developers
Applications
9.7% 8.4% 10.2% 5.8% 1.9% 9.7% 3.9% 1.7%
programmers
Advertising and
marketing 6.6% 8.3% 4.0% 11.2% 8.2% 8.3% 15.1% 12.7%
professionals
Graphic and
multimedia 11.2% 11.8% 18.2% 15.9% 37.6% 13.4% 17.8% 21.1%
designer
Translator 7.3% 3.9% 3.1% 8.8% 12.6% 2.3% 1.0% 0.8%
Data scientist 5.1% 1.1% 2.5% 2.2% 1.9% 4.3% 0.3% 0.8%
Total 70.3% 64.1% 94.2% 71.9% 83.1% 69.7% 66.1% 58.6%
Source: Freelancer, Upwork and Workana. Own calculations.
8 Due to the limited time to perform the scraping and the number of queries per day allowed in each website, these numbers do not correspond to the total number of observations
available on the websites. However, as the observations were randomly scraped, the information used in this report is expected to represent the job-market dynamics of the job
portals selected.
9 Vacancies and CVs on Upwork tend to be posted only in English.
10 Unlike other job portals websites such as Computrabajo.com, Elempleo.com, etc., machine-learning occupational classifiers fail in classifying the information posted in the
freelancer websites because in these job portals people tend to include ‘noisy‘ information such people’s general descriptions or the skills possessed or demanded instead of the
job title. For instance, it is usual that in these freelancer websites people mention that they know Python, R, SQL, etc. instead of mentioning ‘data scientist’ in the job title.
Consequently, it is necessary to code a set of rules that link keywords (e.g., skills) to each job title and occupation needs and develop an algorithm that automatically classifies
the freelancer websites’ information.
215.1.3. Broader size, while there are many words websites are related to web
analysis of the job postings – with a small size. This indicates that designers, app or web developers,
demand for freelancers freelancers’ vacancies are focused marketing (logo design) and data
Critical to our labour-demand on a specific task such as management. Translation services
analysis are job titles. This variable developing websites, apps, are also highly demanded on
provides an idea of the main marketing, and other multimedia- Upwork, while social media content
characteristics of the demand for related tasks. creators are considerably in
labour. The three job portals display demand on Workana.
the job title information, so analysis The words in the word clouds are
is possible, but they do not use a by no means a full description of Most frequent job titles demanded
harmonised set of titles. the job postings. A typical job in Spanish
posting provides a text description
This section presents analyses of of the task sought. We can use the A similar pattern can be observed in
the most frequent jobs posted on frequency of common occurrences the vacancies posted in Spanish. As
the portals. Some websites display to get a more focused idea of the Figure 5.2 shows, the most
the vacancies in different sort of temporary job that the frequent words in the job titles are
languages (English, French, employers are offering. We do this ‘web‘, ‘developer‘, ‘diseño‘ (design)
German, Portuguese, and Spanish). by looking at word associations, among others. The evidence
For this document, the analysis which is something like an ad hoc suggests that similar jobs are
considers those vacancies posted job description. demanded in Spanish or English on
in English or Spanish. this job portal.
A word is associated with another
Figure 5.2: Most
Figure 5.1 shows the most frequent word if both expressions frequently frequent job titles demanded in Spanish
words that appear in the job titles appear together. As it turns out,
for each website. The size of the there are few combinations that
word indicates its frequency. For appear frequently. For instance, the
instance, the most frequent word word ‘website‘ has an association
on Freelancer is ‘website‘, followed with ‘designer‘ or ‘build‘ on
by ‘developer’, ‘app’, ‘WordPress’, Freelancer.com. This indicates that
and ‘Excel’, among others. Similar most common job titles demanded
patterns are observed in Upwork are related to graphic and
and Workana. It is important to note multimedia designers (ISCO code
that there is a high concentration of 2166). In fact, most of the job
words in job titles. There are vacancies posted on all the
relatively few words with a large
Freelancer.com
Figure 5.1: Most frequent job titles demanded in English
Upwork.com
Freelancer.com Upwork.com
Source: Freelancer, Upwork
and Workana. Own calculations
Workana
Source: Freelancer, Upwork and Workana. Own calculations
22You can also read
