Fig 1.
Spearman correlations among different financial and online data.
On the left panel the correlation matrix for the entire data set: all correlations between online and financial quantities are weak. This changes once we focus on specific ATECO codes, due to their peculiarities: on the right panel the same correlation matrix for sector 62 (Computer programming, consultancy and related activitie). A relatively strong correlation is present between the number of followers and the total assets, for instance. This behaviour may be due to the importance of the communications in online social networks for this sector, which increases with a firm’s resources.
Fig 2.
Validated projection of the network of firms.
The network is composed by 80 firms and 135 links. The dimension of each node is proportional to its degree, i.e. the number of connections. The various colors represent the different communities. The shape of the nodes indicate the GUO (Global Ultimate Owner): rectangles are firms owned by the state or other public bodies, diamonds are firms owned by mutual & pension funds / nominal / trust funds, circles are firms owned by individuals or families (family firms), while triangles are companies; the rest are diamond shaped.
Fig 3.
Properties of nodes in the validated network vs. the ones in the entire set.
The various distributions show that the validated accounts are those that have a greater number of friends (top left panel), use greater number of hashtags (bottom left panel) and write more messages (bottom right panel). Interestingly enough, the validated nodes are not the most popular, i.e. those with the highest number of followers (top right panel). In order to check the most popular accounts, we focused on accounts with more than 106 followers. In fact, the number of their messages is extremely limited, while their use of hashtags is extremely focused on their activities. This may be related to their strategies, to remark the exclusiveness of their products. More details can be found in the main text.
Fig 4.
Spearman correlations among different financial and online data for firms in the validated network.
On the left panel the correlation matrix for the validated network: all correlations between online and financial quantities are weak, as in the entire data set (see left panel of Fig 1), but for the one between the number of followers and the total assets (0.51). Again, the situation changes once we focus on a specific ATECO code and it is even more striking than for the entire data set (see left panel of Fig 1). In fact, the same correlation matrix for sector 62 (Computer programming, consultancy and related activitie) shows a strong correlation, for instance, between the total number of messages and the total assets (0.73) and between the number of Likes per message and the total assets (0.75). In fact, the validated accounts are those that follow a common strategy in the usage of hashtags and the importance of their appearance online increases relatively to the dimension of the company.
Fig 5.
ATECO codes of the accounts involved in the various communities in the validated network of Fig 2, 1/2.
Fig 6.
ATECO codes of the accounts involved in the various communities in the validated network of Fig 2, 2/2.
Fig 7.
Identification between themes and the community displayed in Fig 2: As it can be further observed, the themes the different groups of accounts deal with are closely related to their ATECO codes, i.e. to their sector.
Fig 8.
The top 5 frequent hashtags for the communities in the largest connected component of Fig 2.
Due to the great number of ex aequo, the top 5 most frequent hashtags is, in general, longer than 5.
Fig 9.
Frequency of accounts using the various CSR hashtags.
The frequency of accounts using CSR hashtags is higher in the greater communities, with a 100% covering in the case of “Environmental Sustainability” (in purple).
Fig 10.
Frequency of hashtags in the validated network, by community and CSR dimension.
As expected by analysing the community composition in terms of ATECO code, the attention on the various group is quite different: for instance, the ℌEnvironmental Sustainabilityℍ community (in purple) has a frequency of hashtags in the Environmental dimension that is greater than twice the value observed on the entire validated network. Similar considerations apply to the ℌRemote Workingℍ community (in sky blue) for the social dimension and the ℌMobilityℍ one (in white) for the social one.
Fig 11.
Average number of likes (left) and retweets (right) per message containing a hashtag in the environmental (ENV), social (SOC) and economic (ECON) dimensions.
Interestingly enough, the frequency of the CSR hashtags used (reported in the table of Fig 10) is not necessarily mirrored in the number of likes or retweets received on average per hashtag.
Fig 12.
The top 5 frequent hashtag for the smaller communities of the network of Fig 2.
Due to the great number of ex aequo, the top 5 most frequent hashtags is, in general, longer than 5.
Fig 13.
As expected, the various CSR dimensions used depends on the ATECO sectors covered by the community.