London Social?

Tracing Social Discourse in the UK Startup Ecosystem

Team Members

Jeroen de Vos, Natacha Berbers, Ana Garza

Contents

Introduction

The methodology employed in this research draws on digital methods, which refers to "repurposing online devices (such as Google searches, Facebook and Wikipedia) for social and political research that would often have been otherwise improbable" (Rogers 2014). Specifically, to this case, a Twitter-driven method is used to map the local ecosystem with the intention of investigating the topics and actors that are influencing the online social entrepreneurs’ debate. Moreover, this research builds on earlier research by De Vos (2016), of which the preliminary results are published on www.startupecosystem.nl.

Research Questions

• RQ 1: How to map the social entrepreneurs'' discourse?

• RQ 2: To what extent do the issues associated with social entrepreneurs change over time?

• RQ 3: How is the discourse of social innovation positioned in the larger debate of startups?

• RQ 4: To what extent can we compare the discourse surrounding social entrepreneurs in London?

As this is exploratory research, aimed to provide a frame for further research, we did not have any specific expectations or hypotheses for our findings.

Initial Data Sets

1. London-based startups

Using the website http://london.startups-list.com/ as a starting point, we harvested all outgoing links and scraped the links for twitter accounts and up to 1200 tweets per user were recorded, in a backwards fashion. When a user had more than 1200 tweets, only the newest tweets were scraped. In our database, half of the users have tweeted more than 1200 times, leading to recency bias and a bias towards less active users.

This original scraping led to 1.580.399 tweets, tweeted by 1.318 distinct users, with the first tweet dating back to 2007. Since 2010 the tweet volume has increased exponentially, but the distinct users only increased in a linear fashion. This can partially be explained by the fact that users who tweet a lot have been throttled.

This database will be used as the starting point for exploring the London startup ecosystem, to subsequently zoom in on the social entrepreneurs driven sub-debate.

2. Netherlands-based startups

To make a comparison with Dutch social entrepreneurs, we also used a second dataset. This dataset has been provided by the Dealroom.co, an aggregator of startup information. The list of 'early growth' and 'seed' companies was used with TCAT to scrape their accounts in similar fashion to the collection of the London-based startups. The TCAT user-based query listed 598 Dutch startups and scraping each account scraped up to 3200 tweets resulted in a database with 474.613 tweets. For more information on the previous research based on this database, see www.startupecosystem.nl.

1. How to map the social entrepreneurs' discourse?

Through a general twitter and google search, we were able to determine what hashtags were used to refer to social entrepreneurs and we decided on using the most popular one, socent, as our starting point. Employing a co-hashtag analysis, we identified which hashtags were most often used in conjunction with socent. The graph below shows the hashtags most often co-appearing with #socent, such as socinn (social innovation), mhealth (mental health) or impact.

To further analyse the issues associated with socent, we conducted a so-called second-level co-hashtag analysis. All relevant hashtags that were listed in the co-hashtag analysis described above, were included in a query used to filter the entire 'London startup' dataset. To ensure that we were indeed capturing those hashtags used by or associated with social entrepreneurs, we analysed the most frequently used hashtags and confirmed that they refer to issues and topics social entrepreneurs concern themselves with. We cleaned out generic terms such as London, startup and startups, and visualised the network with Gephi.

The hashtags included in the query are listed below.

[#2014nt100 OR #2030now OR #3bl OR #careerchange OR #charity OR #clearlysocialangels OR #crowdbacker OR #crowdfunding OR #digitalhealth OR #dv OR #edtech OR #education OR OR #ff OR #funding OR #futureisfemale OR #impact OR #impinv OR #investinhubbub OR #ldntechweek OR #mentalhealth OR #mhealth OR #peacehackbey OR #refugees OR #sec2016 OR #skollwf OR #socent OR #social OR #socialenterprise OR #socialgood OR #socimp OR #socinn OR #socinv OR #t4gbuddyapp OR #tcdisrupt OR #techforgood OR #theventure OR #thinkbigcrowdfund OR #vaw OR #women OR #womenintech]

The following network is based on the result of the search above. It would include any of the 1.4M tweets from London startups that include any of the hashtags above and sheds light on the immediate topics surrounding the 'socent' topic. In other words, the socent hashtag has been used as an entry point into the sub-discussion on social entrepreneurs.

In this sphere, we see five general themes emerge in the hashtags; funding, marketing efforts associated with social media, health, education and women. Funding is located at the centre of the sphere. Particular forms of funding form a bridge with the issue of health, education and technology. Funding seems to be central in the sphere concerning all social innovation sub-issues. This might be explained by the fact funding is central to any business model (social or not), nevertheless, it is surprising that crowdfunding is an even bigger node than the ‘funding’ node. Perhaps social entrepreneurs more strongly tap into alternative forms of funding, as they have a difficult time convincing traditional investors to invest in a company which does not have financial gains as their primary business goal.

The themes health, education and women can be interpreted as the social issues that are central to London social entrepreneurs. Health is often associated with technology and mental health, indicating that social entrepreneurs see technology as a tool to solve certain health issues. Likewise, education is associated with technology as well. Surprisingly, the theme ‘women’ also emerges. Hashtags such as womeninstem, womenintech, vaw (violence against women), dv (domestic violence) and empowerwomen, can all be found in the sphere. In some cases, hashtags indicate that social entrepreneurs are trying to solve an issue, such as violence against women. In other cases, the presence (or need for a presence) of women is emphasised, such as #womenintech.

While it is interesting to note what we ARE seeing, it is equally important to note what we ARE NOT seeing. Within the London startup sphere, there is relatively little mention of social entrepreneurs. We suspect, that similar to what has previously been observed in the Netherlands (De Vos 2016), social entrepreneurs shy away from using the term social entrepreneur because it can hinder their funding efforts. In addition, in the Dutch social entrepreneur sphere, there was a large emphasis on themes such as sharingeconomy, circulaireconomy and agriculture.

In the general startup-sphere, the topic of financial technology (FinTech) is popular, but it is completely missing in the social entrepreneur debate. We expected FinTech to be closely related to social entrepreneurs as financial payment services solve many issues social entrepreneurs in developmental countries are facing. Perhaps entrepreneurs in FinTech do not acknowledge that they are creating these products to solve a social problem, as this would hinder their funding efforts. Instead, they might emphasise the large untapped demand for financial payment services in less developed parts of the world, where the general banking system is not functioning optimally. In other words, they might veil their social entrepreneurship under the guise of being a traditional entrepreneur, solely concerned with financial gains.

2. To what extent do the issues associated with social entrepreneurs change over time?

To investigate to what extent issues associated with social entrepreneurs changes over time, we conducted a hashtag frequency with the term ‘socent.’ We choose a relatively short time frame, 2 years, as this would allow us to see if there is an annual rhythm to the issues. Moreover, this timeframe provides a solid body of tweets and as tweets were dropped in a backwards fashion, a short timeframe would allow for the most unbiased dataset. We divided this period up into four six-month periods.

The graph above indicates that the discourse is not a fixed conversation, instead topic change over time in a dynamic fashion. It seems to be driven by a set of topics such as crowdfunding, education, edtech and social media. In addition, events seem to be an important driver of the social entrepreneurial discourse. Issues such as tcd disrupt, LDNTechWeek or SEC2016 all refer to events.

3. How is the discourse of social innovation positioned in the larger debate of startups?

To contextualise the network described in our first research question, we want to position this conversation in the larger debate of startups in general. We ran a co-hashtag analysis without the use of any queries to investigate what social themes link most to the general startup sphere.

We then manually tagged all social enterprise-specific hashtags (see RQ 1) in the co-hashtag network of the entire startup database (no query used). Not many hashtags were present in the general startup sphere. We can see that only tip of the iceberg of social-specific topics is reflected in the larger debate. While social entrepreneurs aim to solve large variety of social problem, both generic tags 'health' and 'education' one of the few social topics that were present in the general startup sphere.

In the London startup space, the yellow and pink coloured nodes represent health and education respectively. Both issues tie in the debate from different angles, and thus social innovation is not thematically clustered.

4. To what extent can we compare the discourse surrounding social entrepreneurs in London and the Netherlands?

1. General startup ecosystem compared

Now that we have positioned the social entrepreneur debate within the London startup debate, we can further understand the dynamics of the UK social entrepreneurs, by comparing it with another startup ecosystem. We choose to combine the UK social entrepreneurs network with the Dutch social entrepreneur network for several reasons. First, the UK was the first country to integrate and recognise the general business model employed by social entrepreneurs and provide a legal framework for them to operate it, and contrastingly, the Netherlands was relatively late in recognising this business model and to this date, has not provided a legal framework for it. As such, it would be interesting to see the connections between these two eco-spheres. Second, its geographic closeness would allow for many connections, and the presence of connections or lack thereof signals how inward-focused both spheres.

The graph above represents both co-hashtag analyses of the UK social entrepreneur network (left-hand side) and the Dutch social entrepreneurs network (right-hand side). One general observation we can make is the low level of overlap in the discourse. Relatively little nodes are connected to both networks. The little nodes that do connect the two networks are related to education and education technology. Further research will have to show why specifically education and education technology are linked in the network, while issues such as mental health do not overlap. One explanatory factor for the low level of overlap is the language barrier. As some Dutch social entrepreneurs are focused on solving a local issue, they might only tweet in Dutch and as a consequence, use different, Dutch, hashtags. This would prevent many connections.

Another way to compare the two ecosystems is to focus on the user side; how many actors are actors in both spheres. Similar to the method used to produce the graph above, we appended the two data-sets and limited both to the same two-year time-frame. We then selected the top 500 most mentioned users in both networks, which normalised the comparison. When we compare the overlap in actors, we can see that both ecosystems have way more actors in common than they have shared themes. In this network every node is a Twitter account, and an edge (link) is created with every mention in a tweet showing their affiliation. The actor most present in both spheres is 3dHubs, a social startup with Dutch roots, who recently expanded to New York and Amsterdam.

A surprising finding is the location of two national newspaper within the user networks. The Volkskrant, a Dutch newspaper, was more closely related to the UK user mention network. Likewise, the Guardian, a British newspaper, is more closely linked to the Dutch user mention network. A possible explanation is that Dutch social startup might actively target the Guardian in their tweets.

2. Specific social startup ecosystem compared

Another way to find the unique and common features of the Dutch and UK social entrepreneur networks is to compare the issue space of both networks. Using the exact same method as we employed for our first research question, a second-level co-hashtag analysis, we plotted the Dutch social entrepreneur ecosystem. We choose to use the same term [socent] as the starting point for our Dutch analysis as socent is a term also used in the Dutch ecosystem. All relevant hashtags that were associated with socent were included in a query and we ran the second-level hashtag analysis for the Dutch ecosphere.

The hashtags included in this query are listed below.

[#africa OR #buurt OR #chaldal OR #deeleconomie OR #digitaldev OR #education OR #financialinclusion OR #globaldev OR #groei OR #ict4ag OR #iedereenwinst OR #impact OR #impinv OR #intdev OR #klussen OR #participatie OR #sdgs OR #socent OR #socents OR #social OR #socinn OR #sustainability OR #wasmachine OR #zonne]

The London-based socent sphere is based on the list of 1200 startup actors initially, whereas the Netherlands-based socent sphere is based of 598 Dutch startups. Twice as many users are active in the London social entrepreneur ecosystem, than in the Dutch ecosystem. Consequently, the British sphere is denser. The Dutch ecosystem has a mixture of English and Dutch hashtags, which reflects the trend of Dutch users that tweet in Dutch but include English-language hashtags.

These two spheres express the discussion around and by local social startups. Whereas social innovation in London seems to be built around topics education, emancipation/empowerment and health, the Dutch sphere shows a different image. Besides the general marketing terms (socialmedia) and brand discussions (socialbusiness), the deeleconomie [sharing economy] has quite a prominent place in the sphere. Sustainability and impact closely relate to the original query 'socent', and link to agriculture (in black) and education -the only topic that clearly overlaps.

Discussion

1. Hashtag snowball sampling

Enquiring in something as intangible as a discourse is not a straightforward process. The initial actor-focussed approach lead to limited results since the concept 'social entrepreneur' seems to have quite a specific definition and clarity. Triangulating the London startup list (1200) with the membership list (900) of Social Enterprise UK led to only two results, so we had to come up with a different angle to tackle this problem. Fortunately, the 'social enterprise' is also a hashtag, which has been trending some time ago in the UK, 'socent'. Employing a actor-network like approach, we decided to follow the hashtag into a method we coined 'hashtag snowball sampling'. After the qualitative research method to get to new contacts and information by asking the interviewee whether they know someone to interview next, the hashtag 'socent' has been interrogated for its relations based on a co-hashtag analysis. The resulting hashtags would in turn be used to inform the search query for a second level snowball. At the same time, when using just one hashtag as an entry point in the field needs to be taken into that this hashtag will at the same time be the selection bias of the analysis. Additionally, the snowball sampling technique will add 'related' topics, which will become noisier with every next level scrape -to a point where noise and context are hard to distinguish. Lastly, when looking at the [socent] informed graphs, it needs to be taken into account that they are based on startup tweets only.

2. Comparability

Throughout this research we have worked with two databases: the first one scraped of 1200 London-based startups and the second of 600 Netherlands-based startups. The question to what extent the two ecosystems can be compared led to two specific insights. First of all, an actor (mention) network or discourse (hashtag) network are hard to compare in their network specificities. Although general descriptions can be given on the amount of participants, the density of the network it is hard to make any contextual claims on the basis of these differences. Do these steps allow to compare the incomparable, and can we even talk about a comparison? It turned out most productive to compare the two networks based on their distinctiveness and commonalities. Actually, the organising denominators, being hashtags or users have been taken as the means of comparison. This move has been productive in creating a better understanding of the two ecosystems, but we sometimes had a hard time to make a hypothesis to explain why certain hashtags / users take a prominent role in between the two spaces. Therefore, similarities in the two networks should more be considered as suggestions for further exploration rather than explanatory in itself.

Conclusion

When mapping the social discourse, we found that five general themes emerged (funding, marketing, health, education and women). All but one of these themes were unsurprising. The theme of women, in the form of women empowerment, or solving an issue often related to women (domestic violence) were very surprising. Further research could provide more explanations as to why this theme has emerged and whether it is a long-term trend.

We found the discussion among social entrepreneurs to be dynamic and mainly event-driven. When positioning social entrepreneurs in the larger startup debate, we found that not many social entrepreneur specific hashtags were used in the general startup debate.

Comparing the Dutch social innovation debate to the London social innovation debate, we found that there was relatively little overlap in discourse. Themes that did overlap include education and education technology, while mental health was only present in the London debate. In addition, both debates were very locally based, and issues did not have an outward focus. In contrast, more actors were active in both spheres, than overlapping issues. In further research, we could zoom in on specific users and further understand why specifically these users overlap in both networks, while the question 'what is innovation' is seen through a local lens.

In sum, our exploratory research indicates that social innovation is not central to the UK startup debate. Our visualisations showed that social entrepreneurship/enterprise only occupy a small space within the issue space of startups and are located in the periphery. This suggests that social entrepreneurs are in the periphery in the London UK startup debate but social entrepreneurs can be a core of another network.

Bibliography

De Vos, Jeroen. "Tracing the Social: A Mixed Method Approach to Startup Ecosystems". Unpublished paper. 2016.

Dees, J. Gregory, and others. The meaning of social entrepreneurship. Kauffman Center for Entrepreneurial Leadership, 1998. Google Scholar. Web. 5 feb. 2016.

Drucker, Peter. Innovation and entrepreneurship. Routledge, 2014. Print.

Leadbeater, Charles. The Rise of the Social Entrepreneur. London: Demos, 1997. Print.

Rogers, Richard, and others. “Political Research in the Digital Age”. International Public Policy Review 8.1 (2014): 73–87. Print.