Growing up in Europe

Over the last few years I have been advising on a project called EuroCohort. The aim of the project is to set out the groundwork for a European wide cohort study. Birth Cohort studies are exceptionally powerful, especially in the fields of child development and social inequalities. Details of the project can be found here:

The only downside is that the project is dominated by British enthusiasm. Brits love cohort studies. The UK has been running cohort studies both nationally and regionally for decades. Not only that but there are several UK TV series such as 7 and up which follow individuals over time and see how they change over time. The most recent of these was child of our time, which is a great way to communicate the importance of social research, its links to medical and policy research:

It would be great to see the development of a European Birth Cohort study over the next decade. It would not only allow us to better understand the next generation of Europeans, but would also be a great way for British Social Scientists and their European Colleagues to maintain collaborations over the coming years.

Why aren’t surveys run like Uber?

The rise of companies such as Deliveroo and Uber should be considered seriously by Survey Research Infrastructures. Such disruptive models could be extended to the survey management system and lead to similarly significant reduction in costs. In this model, interviewers would register with a service and be available to service specific interview requests by survey organizers, much in the same way that the relationship between driver and passenger is brokered by Uber.

Currently survey agencies play a far larger role in this brokering and take on the overall responsibility of fieldwork implementation. However, the increasing fieldwork oversight provided in survey infrastructures such as SHARE, ESS and the GGP has winnowed this role to one in which the logical next step is to omit the survey agencies entirely. In this way, the survey infrastructure can commission and evaluate individual interviews rather than the overall fieldwork in a country, affording them greater flexibility, oversight and control.

One primary concern of such a model would be the effect on data quality. It would therefore be necessary for such a service to include an accreditation system which was able to recognize interviewers who were trusted interviewers for specific surveys and had an established track record. Such accreditation could easily be managed by existing services such as Coursera or Udemy. Furthermore, individual interviewer’s performance statistics would be stored and tracked overtime and allow survey infrastructures to quickly sort and select interviewers that are qualified to conduct the work.

This approach has many advantages. It allows survey infrastructures to commission fieldwork quickly and in very specific populations. It provides significantly more control over interviewer management and quality controls than through the current procedure of fieldwork agency brokering. It also provides survey infrastructures with enhanced control over cost management as central teams would have full control over the tradeoffs between cost, speed and quality. Over the past two decades, face to face interviews have increased dramatically in their cost and this looks set to continue. This is in part due to the increasing amount of commercial market research that is web based and thus a decline in overall demand for face to face surveys which increases the costs for survey infrastructures.

Such a service could be developed and managed by a consortium of survey infrastructures and maintained through a small service cost for each interview conducted. The service would be wholly owned by the service infrastructures and open only to accredited survey institutes that do not work on a for profit basis.

Some nice flying dots…

Me and Eugenio Paglino, a student assistant at the GGP, spent the summer creating this interactive visualization. I think it has a lot of potential for illustrating multi-state models which are really common in data such as the GGP. We were inspired by the work of Nathan Yau at Flowing data and work he did on the US time use study. What’s interesting is that the visualization doesn’t convey a specific message or finding. Its impact actually lies in it’s ability to represent the data as complex and dynamic. I think the moving dots are more ‘life like’ than in a traditional scatter plot and this means that the reader can actually empathize with specific movements and patterns. The dots are alive, and this makes the data more relatable which is essential for conveying messages to policy makers.