1: \begin{abstract}

2:   A local culture denotes a commonly shared behaviour within a

3:   cluster of firms. Similar to social norms or conventions, it is an emergent

4:   feature resulting from the firms' interaction in an economic

5:   network. 

6:   To model these dynamics, we consider a distributed agent population,

7:   representing e.g. firms or individuals. Further, we build on a continuous

8:   opinion dynamics model with bounded confidence ($\epsilon$), which assumes

9:   that two agents only interact if differences in their behaviour are

10:   less than $\epsilon$. Interaction results in more similarity of 

11:   behaviour, i.e. convergence towards a common mean. This framework is

12:   extended by two major concepts: (i) The agent's in-group consisting of

13:   acquainted interaction partners is explicitly taken into account. This

14:   leads to an effective agent behaviour reflecting that agents try to

15:   continue to interact with past partners and thus to keep sufficiently

16:   close to them. (ii) The in-group network structure changes over time,

17:   as agents can form new links to other agents with sufficiently close

18:   effective behaviour or delete links to agents no longer close in

19:   behaviour. Thus, our model provides a feedback mechanism between the

20:   agents' behaviour and their in-group structure. Studying its

21:   consequences by means of agent-based computer simulations, we find that

22:   for narrow-minded agents (low $\epsilon$) the additional feedback helps

23:   to find consensus more often, whereas for open-minded agents (high

24:   $\epsilon$) this does not hold. This counterintuitive result is

25:   explained by simulations of the network evolution.

26: 

27: \end{abstract}

28: