Last year, the Software Sustainability Institute conducted a survey of Research Software Engineers (RSEs) to learn more about them and their work conditions. The RSE community has grown from a concept born at an Institute event to an international phenomenon. It's important to learn more about this community so that our campaigning, and that of our international partners, continues to help RSEs gain the recognition they deserve for their huge contribution to research.
We began surveying RSEs in 2016, in 2017 we also surveyed Canadian RSEs and last year we added four further countries. Our thanks to our partners: Scott Henwood (Canada), Stephan Janosch and Martin Hammitzsch (Germany), Ben van Werkhoven and Tom Bakker (Netherlands), Anelda van der Walt (South Africa) and Daniel Katz and Sandra Gesing (USA).
Diffusion
In total, we analysed 841 responses across the five countries listed below.
| Countries | Number of analysed responses | Link to analysis | Link to data |
|---|---|---|---|
| Canada | 110 | Jupyter notebook | Public data |
| Germany | 325 | Jupyter notebook | Public data |
| Netherlands | 77 | Jupyter notebook | Public data |
| United Kingdom | 253 | Jupyter notebook | Public data |
| United States of America | 164 | Jupyter notebook | Public data |
| South Africa | 22 | Jupyter notebook | Public data |
Composition of the survey
The base questions for the survey were tailored to meet the requirements of each country. They covered ten subjects:
- Demographics: traditional social and economic questions, such as gender, age, salary and education.
- Coding: how much code do RSEs write, how often, and for whom.
- Employment: questions about where RSEs work and in which disciplines.
- Current contract: understanding stability of employment by questioning the type of employment contract RSEs receive.
- Previous employment: understanding routes into the profession the reasons for choosing it.
- Collaboration and training: who RSEs work with, how many people they work with, and the training they conduct.
- Publications: do RSEs contribute to publications and are they acknowledged?
- Sustainability: testing, bus factor, technical handover.
- Job satisfaction: what do RSEs think about their job and their career?
- Network: how do RSEs meet and gain representation?
A quick insight into the results
Results differ across countries, but there are a few broad themes that appear common for RSEs around the world (obviously, we'll make some generalisations here, but just go directly to the data as described in the linking to the data section below.
Gender, age and qualifications
RSEs are mainly male and generally found in the 25 to 44 years old age bracket. The majority of them hold a doctorate, except for in Germany where just under half of RSEs have a PhD.
| Countries | Gender | Age | Qualification |
|---|---|---|---|
| Canada | N/A | N/A | Doctorate (45%) |
| Germany | Male (83%) | 25 – 44 (84%) | Doctorate (48%) |
| Netherlands | Male (63%) | 25 – 44 (84%) | Doctorate (56%) |
| United Kingdom | Male (84%) | 25 – 44 (75%) | Doctorate (67%) |
| United States of America | Male (82%) | 25 – 44 (69%) | Doctorate (60%) |
| South Africa | Male (92%) | 25 – 44 (76%) | Doctorate (68%) |
Discipline of study and work
RSEs are most likely to have a degree in: Computer sciences, Physics and Astronomy, or Biology.
| Country | First discipline | Second discipline | Third discipline |
|---|---|---|---|
| Canada | Information technology (25%) | Biomedical engineering (7%) | Artificial intelligence (5%) |
| Germany | Physics and astronomy (26%) | Computer sciences (17%) | Biology (11%) |
| Netherlands | Computer sciences (20%) | Physics and astronomy (18%) | Biology (12%) |
| United Kingdom | Computer sciences (27%) | Physics and astronomy (27%) | Biology (8%) |
| United States of America | Computer sciences (25%) | Physics and astronomy (15%) | Biology (15%) |
| South Africa | Physics and astronomy (55%) | Other (9%) | Electrical & Electronic Engineering (5%) |
RSEs tend to work in the fields of Computer sciences, Biology, and Physics and Astronomy.
| Country | First discipline | Second discipline | Third discipline |
|---|---|---|---|
| Canada | Information technology (21%) | Artificial intelligence (8%) | Biomedical engineering (7%) |
| Germany | Computer science (18%) | Physics and astronomy (14%) | Biology (14%) |
| Netherlands | Computer science (20%) | Biology (10%) | Physics and astronomy (10%) |
| United Kingdom | Computer science (15%) | Biology (11%) | Physics and astronomy (10%) |
| United States of America | Computer science (21%) | Biology (12%) | Physics and astronomy (9%) |
| South Africa | Physics and astronomy (29%) | CS (12%) | Mathematics (10%) |
However, a closer look at the results shows an interdisciplinary perspective to RSEs work. As shown in this further analysis of the UK results, up to 63% of UK RSEs are working in more than one discipline.
Why choose the RSE career path?
We asked respondents to list in priority order their reasons for choosing the RSE career. The top three responses are shown below. It is the research environment that mainly attracts RSEs, but a desire to advance research is also a major draw. Ultimately, RSEs want to work in research, but do not want to follow a conventional research career path.
| Country | First reason | Second reason | Third reason |
|---|---|---|---|
| Canada | N/A | N/A | N/A |
| Germany | Desire to work in a research environment | Freedom to choose own working practices | Desire to advance research |
| Netherlands | N/A | N/A | N/A |
| United Kingdom | Desire to work in a research environment | Desire to advance research | Opportunity to develop software |
| United States of America | Desire to advance research | Desire to work in a research environment | Freedom to choose own working practices |
| South Africa | Desire to work in a research environment | Flexible working hours | Ability to work accross disciplines |
The role of RSEs in publications
RSEs make a major contribution to publications, but they are not always acknowledged for doing so (and are named as an author even less frequently).
| Country | Software used for publication | Acknowledged in the paper |
|---|---|---|
| Canada | 76% | 70% |
| Germany | 83% | 71% |
| Netherlands | 95% | 77% |
| United Kingdom | 91% | 78% |
| United States of America | 90% | 71% |
| South Africa | 74% | 42% |
Tools and working practices
RSEs are keen users of open-source licences for their code, but less frequent users of Digital Object Identifier (DOI).
| Country | Use of open source licence | Use of DOI |
|---|---|---|
| Canada | 84% | 26% |
| Germany | 62% | 18% |
| Netherlands | 85% | 31% |
| United Kingdom | 68% | 22% |
| United States of America | 81% | 32% |
| South Africa | 44% | 11% |
We asked about testing, the bus factor of their biggest project and the presence of a technical handover plan.
| Country | Testing | Bus factor | Technical handover |
|---|---|---|---|
| Canada | N/A | N/A | N/A |
| Germany | 15% not implementing anything | 1 (57%) | Yes (19%) |
| Netherlands | 13% not implementing anything | 1 (45%) | Yes (21%) |
| United Kingdom | 10% no testing | 1 (42%) | Yes (26%) |
| United States of America | 7% no testing | 1 (39%) | Yes (18%) |
| South Africa | 24% no testing | 1 (78%) | Yes (11%) |
The percentage of RSEs not implementing any form of testing could be argued as rather low, but keeping in mind the importance of their work in producing and publishing research 10% untested code is still a problem. A similar problem is found with the bus factor. The majority response across all countries is a bus factor of 1, making research software highly vulnerable to staff absence or movement.
What languages to RSEs prefer?
There are some differences across countries, but Python is clearly the dominant language among RSEs.
| Country | First language | Second language | Third language |
|---|---|---|---|
| Canada | N/A | N/A | N/A |
| Germany | Python (18%) | C++ (10%) | Javascript (9%) |
| Netherlands | N/A | N/A | N/A |
| United Kingdom | Python (15%) | Unix Shell Script (11%) | Markup language (9%) |
| United States of America | Python (18%) | C (11%) | C++ (10%) |
| South Africa | Python (22%) | SQL (10%) | R (10%) |
Link to the data
This quick overview only scratches the surface of the data we collected. To get complete access to the available information, you can peruse the Github repository. The data is provided under a Creative Commons by attribution licence. The data and analysis are stored in Zeonodo:
We are currently working on creating the survey for 2018 and hope to run it in July. If you wish to participate, you can find out more from our contribution instructions.