Spatial perspectives on niche empowerment: an agent-based model

ABSTRACT

This paper aims at increasing our understanding of the role of space dimensions in niche evolution dynamics. To accomplish this goal, an agent-based model is developed, locating agents in geographical space. The model is then used to show how local niches built on geographical proximity perform differently from global niches built on relative proximity, in terms of the timing of niche creation, the velocity of niche maturation, average profit, environmental uncertainty and knowledge exchange. It concludes that the spatial dimension of niches should be considered in the development of policies to promote sustainable transitions.

KEYWORDS:

• innovation niche
• geographical proximity
• relative proximity
• agent-based modelling
• policy intervention

JEL:

• O31
• O33
• O38
• Q55

DISCLOSURE STATEMENT

No potential conflict of interest was reported by the authors.

Notes

1. We distinguish between local and extra-local space. The former is considered a geographically delimited area, in which knowledge can be exchanged through face-to-face interaction or word of mouth, while the latter represents a broader space of interaction, outside the local dimension.

2. The base assumption is that firms producing with the regime technology operate under perfect competition; thus, every firm has zero extra profits.

3. This opportunity is extended to all agents, while in the EINM it is limited to only those with a high level of expectation.

4. Agent position is fixed randomly during the initialization phase and does not vary during the simulation, as it represents the physical location of niche firms. This is substantially different from the EINM, which views the grid as a social space and considers the movement of agents on the grid representative of the meeting chance among them. Thus, our extension aims at restoring a more appropriate space modelling, as the grid reflects the true physical delimited space.

5. Since the knowledge difference varies in the range [0, 100], the parameters of f are fixed so its maximum corresponds to the medium value of this range (50). Thus, Γ increases in the range [0, 50] and decreases in the rest of the range, when cognitive distance becomes excessive.

6. To define ${\text{β}}_{lbd}$ we used the same parameter employed by Lopolito et al. (2013), who fixed it at 0.025.

7. This is a standard procedure for ABM simulations. A limitation of this procedure is that the calculation of averages can prevent the determination of tipping points (i.e., points of inflection in the simulation). To overcome this limitation, we screened all simulation results.

8. Network knowledge and power are the sum of all knowledge and power accumulated by the network agents during the simulation timeframe. Network knowledge differs from knowledge exchanged locally and/or globally, since the latter represents only a fraction of the knowledge exchanged, while the former represents the total amount of knowledge pertaining to the network.

9. Results of longer simulations are available from the authors upon request.