Spreading Out: The Adoption of Disruptive Technologies in Switzerland
October 2022 by Matthias Niggli
About
The analysis is part of Innoscape. The project is conducted by the Center for International Economics and Business | CIEB of the University of Basel aiming to investigate the Swiss innovation landscape from an economic perspective.
New ideas and technologies power the engine of modern economies. Some technologies have a limited impact in particular industries of the economy. But others, such as autonomous driving or cloud computing, are sometimes labeled as disruptive because they strongly transform processes within entire industries or even affect the overall economy.
In a recent study, Bloom, Hassan, Kalyani, Lerner & Tahoun (2021) define a set of 29 disruptive technologies that have been strongly emerging over the last 3 decades (e.g., Wifi, Cloud Computing, Solar Power, or Artificial Intelligence). Their findings highlight how disruptive technologies gradually spread throughout the US economy and the labor market. They diffuse from pioneering regions to the rest of the country, and from high-paying occupations to lower-paid ones. Yet, pioneering regions obtain persistently higher numbers of high-paying jobs.
To what extent have disruptive technologies spread in Switzerland? Data from JobsPikr on over 3 million Swiss job postings over the last 2 years allows us to approach this question. We assign job postings to disruptive technologies if their full text contains keywords as defined by Bloom et al. (2021). This provides an estimate of how prevalent these technologies are in Switzerland.
Click here to get a more information on the data.
The sample of job postings we use for this analysis is obtained
from JobsPikr
and consists of over 3 million Swiss job postings published by nearly 77
thousand employers. We assign job postings to so-called
Grossregionen (which correspond to
Eurostat NUTS-2 regions), according to information about the
location of the position described in the job posting provided by
JobsPickr (e.g., canton or cities).
Overall, we detect slightly over 200’000 postings with a connection to the 29 disruptive technologies from Bloom et al. (2021) in our sample (i.e., around 6% of all postings). It is noteworthy that these postings are not distributed proportionally across Swiss regions in our sample but are geographically clustered. This suggests that disruptive technologies are more prevalent in some regions compared to others.
Figure 1: Regional Specialization in Disruptive Technologies
Sources and notes: Estimated share of job
postings with a connection to disruptive technologies as defined in
Bloom et
al. (2021) among all job postings of a region. Own estimations of
the CIEB using
job postings data from
JobsPikr. Data
to replicate the plot is available on
Github.
Disruptive technologies are far more widespread in some Swiss regions compared to others
Figure 1 illustrates the aforementioned pattern. According to our data sample, the Zurich area has the largest share of job postings with a connection to disruptive technologies among all of its job postings. Almost 10% of the Zurich area’s postings in our sample have a connection to at least one of the 29 disruptive technologies. This share can be understood as the extent to which a region is specialized in disruptive technologies. Hence, our data suggests that the Région Lemanique lags the Zurich area, and Northwestern Switzerland, Central Switzerland or the Espace Mittelland are only around half as specialized in these technologies compared to Zurich. Overall, though, disruptive technologies seem to have spread in Switzerland to a similar extent as Bloom et al. (2021) report for the USA.
But how prevalent are specific technologies across the almost 77 thousand Swiss employers in our sample? What share of these employers mentions a given technology? Consider, for example, “Cloud Computing”. Almost 4400 employers mention it in at least one of their job postings, corresponding to a share of almost 6%. To evaluate the diffusion of all other disruptive technologies as well, we identically calculate employer-shares for the other technology fields. Figure 2 depicts the results for all technology fields where we detected at least 50 employers with job postings related to that technology.
Figure 2: Technology Diffusion Across Employers
Sources and notes: Estimated number of
companies publishing job postings associated with disruptive
technologies from
Bloom et
al. (2021). Own estimations of the
CIEB using job
postings data from
JobsPikr. Data
to replicate the plot is available on
Github.
The average share across technology fields is 1.6%, but several disruptive technologies, such as Electronic Gaming or Autonomous Cars, are only slightly represented among the companies with job postings in our sample. Hence, the average share is heavily driven by the four most prevalent technologies “Machine Learning & AI”, “Smart Devices”, “Cloud Computing” and, most of all, “Social Networking”. Job postings with a connection to social networks make up over 8% of all job postings in our sample – a more than 5 times higher share than the average across all disruptive technologies. This likely reflects how widespread social media and professions such as Social Media Managers have become.
For other technologies, it is not intuitively clear how widespread they have become though. For example, technologies such as Computer Vision, Search Engine, Wifi and Virtual Reality have similar, yet below-average shares of employers with job postings related to these technologies in our sample. But it could be, that some of them are much more heavily dominated by a few big employers than others. Using the so-called Herfindahl-Hirschmann index measure as a proxy, Figure 3 illustrates the concentration of job postings among employers within the ten most prevalent disruptive technologies (note that we exclude postings published by recruiting companies such as Adecco or Randstad).
Figure 3: Concentration of Job Postings Among Employers
Sources and notes: Estimated number of
companies publishing job postings associated with disruptive
technologies from
Bloom et
al. (2021). Own estimations of the
CIEB using job
postings data from
JobsPikr. Data
to replicate the plot is available on
Github.
As expected, technologies that are more prevalent across employers, are also less dominated by a few big players. However, interesting differences emerge, for example, between the aforementioned technologies Computer Vision, Wifi, Search Engine and Virtual Reality Technologies. Whereas Virtual Reality and Computer Vision showed a similar prevalence as the other two technologies in Figure 2, job postings related to Virtual Reality are clearly the most concentrated across all fields. Its postings are about 7 times as concentrated among employers compared to Search Engine and Wifi technologies, and about 1.5 times as concentrated compared to Computer Vision. Consequently, the share of jobs that are offered by the largest employers in the fields of Search Engine and Wifi technologies (e.g., Kayak, a travel search engine or Regus, an office space service company) are much smaller compared to those in Virtual Reality and Computer Vision. This is because according to our sample, there are only a few employers present in these fields. Prominent ones include the two big tech companies Meta and Apple, as well as the ETH Zurich representing academia (see also, this short overview regarding the prospects of virtual reality in Switzerland). Lesser known but also significant employers include the companies Advertima, Scandit or Netcetera.
At the other end of the spectrum, job postings seem to be a lot more dispersed across several employers for Cloud Computing (e.g., Swisscom, SAP, Atos, Salesforce) and Machine Learning (e.g., ETH Zürich, UBS, Facebook, Roche) technologies. Since these technologies are expected to further evolve and are sometimes labeled as General Purpose Technologies, their relatively substantial spread across several employers to date can be understood as an encouraging sign for the future prospect of the Swiss economy.
This
article was written using
R
Markdown and the
tidyverse R
package collection. Computations to generate the data were performed at
sciCORE
scientific computing center at the University of Basel. Data to
replicate the plots is available on
Github.