Abstract
The paper elaborates an agent based simulation model (ABM) to explore the endogenous long-term dynamics of knowledge externalities. ABMs, as a form of artificial cliometrics, allow the analysis of the effects of the reactivity of firms caught in out-of-equilibrium conditions conditional on the levels of endogenous knowledge externalities stemming from the levels of knowledge connectivity of the system. The simulation results confirm the powerful effects of endogenous knowledge externalities. At the micro-level, the reactions of firms caught in out-of-equilibrium conditions yield successful effects in the form of productivity enhancing innovations, only in the presence of high levels of knowledge connectivity and strong pecuniary knowledge externalities. At the meso-level, the introduction of innovations changes the structural characteristics of the system in terms of knowledge connectivity that affect the availability of knowledge externalities. Endogenous centrifugal and centripetal forces continually reshape the structure of the system and its knowledge connectivity. At the macro system level, an out-of-equilibrium process leads to a step-wise increase in productivity combined with non-linear patterns of output growth characterized by significant oscillations typical of the long waves in Schumpeterian business cycles.
Similar content being viewed by others
Notes
We define the equilibrium cost of knowledge as the cost of a standard divisible input, traded in a competitive market, that can be fully appropriated, wear and tear because has a clear exhaustibility, and is used to produce an output that is traded in a competitive market.
NK models assume the reverse, defining density of the components in the landscape, and their knowledge to be exogenous (Levinthal 1997).
See “Appendix A” for detailed presentation of the pseudo code of the model, “Appendix B” for the parameters of the model, “Appendix C” for the parameters of the simulation.
Note that the system is analytically consistent. Naming \(\prod \) the profit of a generic enterprise and D the dividend it will pay to its shareholders, and remembering Eqs. 1, 2 and 5 it is possible to write the following equations:
$$\begin{aligned} \hbox {D}_{\mathrm{i}} = \prod \nolimits _{\mathrm{i}} = \hbox {pO}_{\mathrm{i}} - \hbox {W}_{\mathrm{i}} \end{aligned}$$(15)where D could be less than zero if a loss had to be reintegrated. The amount of dividends paid to the whole systems is:
$$\begin{aligned} \hbox {D} = \sum \hbox {D}_{\mathrm{i}.} \end{aligned}$$(16)At the aggregate level the system could be resumed as follows:
$$\begin{aligned} \hbox {Y} = \sum \hbox {W}_{\mathrm{i}} + \sum \hbox {D}_{\mathrm{i}.} \end{aligned}$$(17)By specifying \(\hbox {D}_{\mathrm{i}}\) using Eq. (16) it is possible to obtain:
$$\begin{aligned} \hbox {Y} = \sum \hbox {W}_{\mathrm{i}} + \sum \hbox {pO}_{\mathrm{i}} - \sum \hbox {W}_{\mathrm{i}.} \end{aligned}$$(18)By operating simple compensations Eq. (18) becomes:
$$\begin{aligned} \hbox {Y} = \sum \hbox {pO}_{\mathrm{i}.} \end{aligned}$$(19)Recalling expression (2) it is evident that the whole system can reach equilibrium and the amount of money into the system remains always constant.
References
Axtell R (2005) The complexity of exchange. Econ J 115:F193–F210
Anderson PW, Arrow KJ, Pines D (eds) (1988) The economy as an evolving complex system. Addison Wesley, Redwood
Antonelli C (2008) Localized technological change. Towards the economics of complexity. Routledge, London
Antonelli C (ed) (2011) Handbook on the economic complexity of technological change. Edward Elgar, Cheltenham
Antonelli C (2015) Innovation as a creative response. A reappraisal of the Schumpeterian legacy. Hist Econ Ideas 23:99–118
Antonelli C, Colombelli A (2015a) External and internal knowledge in the knowledge generation function. Ind Innov 22:273–298
Antonelli C, Colombelli A (2015b) The cost of knowledge. Int J Prod Econ 168:290–302
Antonelli C, David PA (eds) (2016) The economics of knowledge and knowledge driven economy. Routledge, London
Antonelli C, Ferraris G (2011) Innovation as an emerging system property: an agent based model. J Artif Soc Soc Simul 14 (2) no page
Antonelli C, Ferraris G (2017) The Marshallian and Schumpeterian microfoundations of evolutionary complexity. An agent based simulation model. In: Cantner U, Pyka A (eds) Foundations of economic change - Behaviour, interaction and aggregate outcomes. Springer, Berlin (Forthcoming)
Arrow KJ (1969) Classificatory notes on the production and transmission of technical knowledge. Am Econ Rev 59:29–35
Arthur WB (2014) Complexity and the economy. Oxford University Press, New York
Arthur WB, Durlauf SN, Lane DA (eds) (1997) The economy as an evolving complex system II. Addison-Wesley, Redwood City
Barabasi LA (2010) Bursts. The hidden pattern behind everything we do. Dutton, New York
Beaudry C, Breschi S (2003) Are firms in clusters really more innovative? Econ Innov New Technol 12:325–342
Bischi GI, Dawid H, Kopel M (2003) Gaining the competitive edge using internal and external spillovers: a dynamic analysis. J Econ Dyn Control 27:2171–2193
Bresnahan T, Gambardella A, Saxenian A (2001) Old economy inputs for new economy outputs: cluster formation in the new Silicon Valleys. Ind Corp Chang 10:835–860
Cohen WM, Levinthal DA (1990) Absorptive capacity: a new perspective on learning and innovation. Adm Sci Q 35:128–152
Cowan R, Jonard N (2004) Network structure and the diffusion of knowledge. J Econ Dyn Control 28:1557–1575
David PA (1993) Knowledge, property and the system dynamics of technological change. In: Summers L, Shah S (eds) Proceedings of the World Bank annual conference on development economics. World Bank, Washington
David PA (2007) Path dependence: a foundational concept for historical social science. Cliometrica J Hist Econ Econ Hist 1:91–114
Ferraris G (2006) Tempo, casualità e sincronizzazione degli eventi nei modelli. In: Terna P, Boero R, Morini M, Sonnessa M (eds) Modelli per la complessità - La simulazione ad agenti in economia. Il Mulino, Bologna
Fleming L (2001) Recombinant uncertainty in technological search. Manag Sci 47(1):117–132
Fleming L, Sorenson O (2001) Technology as a complex adaptive system: evidence from patent data. Res Policy 30:1019–1039
Griffith R, Redding S, Van Reenen J (2003) R&D and absorptive capacity: theory and empirical evidence. Scand J Econ 105:99–118
Griliches Z (1979) Issues in assessing the contribution of research and development to productivity growth. Bell J Econ 10:92–116
Griliches Z (1992) The search for R&D spillovers. Scand J Econ 94:29–47
Guiso L, Schivardi F (2007) Spillovers in industrial districts. Econ J 117(516):68–93
Krugman P (1994) Complex landscapes in economic geography. Am Econ Rev 84:412–417
Lane DA (2002) Complexity and local interactions: towards a theory of industrial districts. In: Quadrio Curzio A, Fortis M (eds) Complexity and industrial clusters: dynamics and models in theory and practice. Physica, New York, pp 65–82
Lane DA, Pumain D, Leeuw SE, West G (eds) (2009) Complexity perspectives in innovation and social change. Springer, Berlin
Levinthal DA (1997) Adaptation on rugged landscapes. Manag Sci 43(7):934–950
Lundvall B (1988) Innovation as an interactive process: from user-producer interaction to the national system of innovation. In: Dosi G et al (eds) Technical change and economic theory. Frances Pinter, London, pp 349–369
Mansfield E, Schwartz M, Wagner M (1981) Imitation costs and patents: an empirical study. Econ J 91:907–918
March JC (1991) Exploration and exploitation in organizing learning. Organ Sci 2:71–87
Martin R, Sunley P (2012) Forms of emergence and the evolution of economic landscapes. J Econ Behav Organ 82:338–351
Metcalfe JS (2002) Knowledge of growth and the growth of knowledge. J Evol Econ 12:3–16
Miller JH, Page SE (2007) Complex adaptive systems. Princeton University Press, Princeton
Mokyr J (1990) Punctuated equilibria and technological progress. American Economic Review P&P 80, pp 350–354
Ostrom E (2010) Beyond markets and states: polycentric governance of complex economic systems. Am Econ Rev 100:641–672
Ostrom E, Hess C (eds) (2006) Understanding knowledge as a commons: from theory to practice. MIT Press, Cambridge
Page SE (2011) Diversity and complexity. Princeton University Press, Princeton
Reichman J (2000) Of green tulips and legal kudzu: Repackaging rights in subpatentable invention, Vanderbilt Law Review 53, pp 17–43. Reprinted In: Dreyfuss R, Zimmerman D (eds), Expanding the boundaries of intellectual property, Oxford University Press (2001) Oxford, pp 23–54
Romer PM (1990) Endogeneous technological change. J Polit Econ 98:S71–S102
Rosser JB (ed) (2004) Complexity in economics. Methodology interacting agents and microeconomic models. Edward Elgar, Cheltenham
Schumpeter JA (1942) Capitalism socialism and democracy. Harper and Row, New York
Schumpeter JA (1947) The creative response in economic history. J Econ Hist 7:149–159
Sorenson O, Rivkin JW, Fleming L (2006) Complexity, networks and knowledge flow. Res Policy 35(7):994–1017
Terna P (2009) The epidemics of innovation. Playing around with an agent-based model. Econ Innov New Technol 18:707–728
Weitzman ML (1996) Hybridizing growth theory. Am Econ Rev 86:207–212
Zhang J (2003) Growing Silicon Valley on a landscape. J Evol Econ 13:529–548
Acknowledgements
The authors acknowledge the financial support from the European Commission DG Research under Grant Number 266959 for the ‘Policy Incentives for the Creation of Knowledge: Methods and Evidence’ (PICK-ME) project, as part of the Cooperation Programme/Theme 8/Socio-economic Sciences and Humanities (SSH), of the Collegio Carlo Alberto, and the University of Torino project IPER.
Author information
Authors and Affiliations
Corresponding author
Appendices
Appendix A: The pseudo code of the model
(Parameter of the simulation are written in italic bold)
Appendix B: Parameters of the model and set up for the different simulations
In order to control the simulations and allow configuration of a wide set of different scenarios, a reach set of parameters has been provided to bias both the behaviour of agents (firms) and the structure of the economic at system macro level and common macro level too. In the simulations of to this paper few of the available parameters vary, their number has been set to support further evolutions of the research, so many parameters have had the same value for all the simulations presented in the paper, related to those parameter no sensitivity analysis, neither specific simulations have been done due to the fact their values were always the same for each simulations or scenario. Models based on the Swarm protocol distinguish two different object devoted to control the simulation: the Observer that is charged to collect and report the results emerging during the simulation the Model that is charge to build all the objects to populate the model and schedule the activity of those ones. Both Observer and Model give the possibility to specify customized parameters,
The observer uses a first set of two parameter to determine the output shape and update:
-
displayFrequency set up the interval, in model steps, between each refresh of the graphs. Because the presented simulation was devoted to study the dynamic of the system, this parameter has been set to 1 (i.e. graphs are redrawn at each simulation step) to fully report variations in the observed quantities.
-
zoomFactor influence the shape of the graphs produced and updated during the simulation run. Its value is usually set to 2, all the simulations used that value.
The model has been provided a wider set of parameters to make the configuration of different scenario very easy. In details:
-
randomSeed: (any natural number in the interval ]0,\(\infty \)[) it is used to initialize the random seed generator. Useful both to vary the random distributions as well as to ensure the possibility to replicate an experiment with the same random number distribution.
-
Agents: (any natural number in the interval ]0,\(\infty \)[) determines the number of firms that will be put in the simulated economy. The maximum number of firms allowed for a simulation depends on the memory and processing power of the computer used for the simulation.
-
Commons: (any natural number in the interval ]0,\(\infty \)[) specifies how many commons will be generate and used into the simulation. The maximum number of firms allowed for a simulation depends on the memory and processing power of the computer used for the simulation.
-
InitialWealth: (any real number in the interval ]0,\(\infty \)[) specifies the initial endowment of workers they’re going to offer into the market to buy the first productive cycle’s output.
-
StartingProductivity: (any real number in the interval ]0, \(\infty \)[) indicates the upper limit for the interval used to assign each agent an initial productivity, by randomly tossing, for each, a different real number into the interval ]0,startingProductivity[.
-
StartingFactor: (any real number in the interval ]1,\(\infty \)[) specifies the quantity of work units the enterprises will demand on the market and employ for production in the first simulated production cycle.
-
SuccessesThreshold: (any natural number in the interval ]0,\(\infty \)[) specifies how much consecutive successes have to be piled before starting a trial for innovation. A success is achieved every time the own profit of the agent is greater than the average common’s one + a tolerance percentage.
-
FailuresThreshold: (any natural number in the interval ]0,\(\infty \)[),\(\infty \)[) specifies how much consecutive failures have to be piled before starting a trial for innovation. A failure is suffered every time the own profit of the agent is less than the average common’s one—a tolerance percentage.
-
IprDuration: (any natural number in the interval ]0,\(\infty \)[) specifies the number of production cycle the patent rights protect each innovation, during this time the innovation is hidden to the other agents.
-
RevampTime: (any natural number in the interval ]0,\(\infty \)[) specifies the number of production cycle after an agent is gone out of business for having another one keep its place. The name of the parameter is due to the fact that a new agents is only the revamp of the old one, with productivity equal to the average common’s one.
-
FactorUp: (any real number in the interval ]1,\(\infty \)[) is the number used to multiply the previous demand for factor to determine the actual one, by the firms that achieved a profit. Because the base assumption is that profitable firms will expand the production this parameter have to be set at a value greater than one, but close to one; for instance setting factorUp to two would means that enterprises that had a profit will double their demand for factor (work) for the next production cycle.FactorDown: (any real number in the interval ]0,1[) is the number used to multiply the previous demand for factor by the enterprises that just suffered a loss; this parameter has to be set close to one too, even less than one.
-
PotentialPerStep: (any real number in the interval ]0,1[) represents the fraction of the common’s productivity that each agent accumulate in each production cycle due to experience. This parameter has to be set accordingly with the following “Productivity upgrade”, that measure the amount of accumulated experience needed to enhance technology of one unit.
-
ProductivityUpgrade: (any real number in the interval ]0,\(\infty \)[) represents the minimum quantity of accumulated potential that could be transformed in a unit of technological enhancement. Note that transformation can be performed for this amount of cumulated experience at a time only, and gives one unit of technological enhancement.
-
Tolerance: (any real number in the interval ]0,1[), defines the symmetric interval around the average common’s profit used by each firm to decide if take actions to improve its technology. Unless the result (either profit or loss) of an enterprise in a certain production cycle was less than [average common’s results * (1-tolerance)] or was grater than [average common’s results * (1+tolerance)] no improvement on the technological level are tried.
-
MaxLoss: (any real number in the interval ]0,\(\infty \)[), it measures the maximum loss an enterprises can cumulate before going out of business. The meaning is in that if cumulated results of an enterprises reach a negative amount less than (−1)maxLoss it goes immediately out of business and will be replaced, into the same common, by another one after revampTime production cycles.
-
TransformationCost: (any real number in the interval [0,\(\infty \)[), it measures the amount of work unit an enterprise has to demand on the market to perform a transformation of accumulated experience in technological enhancement.
-
MovingCost: (any real number in the interval [0,\(\infty \)[), it measures the amount of work unit an enterprise has to demand on the market to move from the actual common to another one.
-
SpilloverMinProb: (any real number in the interval [0,1]), is the success base probability assigned to a generic spillover action, the effective probability of success for a spillover action is computed as: (1-spilloverMinProb)(1-delta)+spilloverMinProb), where delta measures the distance between productivity after and before the spillover action.
-
CommonsCost: (any real number in the interval [0,\(\infty \)[), is the amount used to compute the quantity of work each firms has to buy, each production cycle, to get information and manage relation into the commons it belongs to. The costs are the same for each agent into the commons.
In order to manage the simulations, some control parameters have been used:
-
Scenario: (an integer number in the interval [1,5]) specifies the scenario to be executed among: Alpha, Beta, Gamma, Iota and Theta; by choosing one of them the program automatically sets up the core parameters to configure the chosen scenario.
-
Spillover: (an integer number in the interval [0,1]) it is a simple switch that allows or stops the possibility for firms to spill knowledge from other ones into the common.
-
Reinforce: (an integer number in the interval [0,1]) used to turn on or off the effect of the common productivity on the knowledge each enterprise grows up by executing each production cycle.
Finally the old parameters: (i) focused, (ii) explorationRate and (iii) spilloverCostRate are no more used, even still present in the input form.
Appendix C: Parameters’ values for the simulations
The simulations used for the research were based upon different settings of few parameters, as described in chapter four, obtained by biasing a base configuration named alpha scenario. The full parameters setting for each scenario are reported in the following Table 6.
Appendix D: Robustness and sensitivity
1.1 Introduction—pseudo random generators
Several processes in the model are based on random events: (i) technological improvement may fail according to a probability distribution set parametrically, (ii) other firms are picked up randomly among the neighbours to observe and eventually imitate, (iii) enterprises that move decide randomly the new common to enter in, etc. The generation of pseudo random numbers has to be managed with specific care.
In order to guarantee the full independence of each agent, as well as of each environmental component, like commons, market and so on, each object has been provided an own random generator (Ferraris 2006); the control of the random distributions is based on a simple procedure:
-
(a)
The modelSwarm object (the main component that is charged to build and activate all the other ones, either agents or environmental institutions, has been provided with an own random generator (named the main generator), whose seed can be fixed by the researcher, simply supplying a value for the parameter “randomSeed”. If the value zero is specified the model tosses a random seed using the standard generator the simulation tool (for this model Swarm) provided.
Note that this generator is not used to toss random values for parameters expected to vary. In this way it is possible to fix all the random events, even with parameters that are randomly set up.
-
(b)
Each component of the model that uses random numbers is given an own random generator, made by the modelSwarm just before building the component, fed with a seed tossed by using the main generator.
Such a architecture allows both: (i) independence of each component even for humongous numbers of requests for random values, (ii) full control of the random generation. In this way the researcher is allowed to:
-
(a)
exploit the possibility to replicate a simulation with the same sequence of random numbers,
-
(b)
change randomly the sequence,
-
(c)
avoid interferences among agents and environmental institutions (or generally speaking components) even with heavy usage of random numbers.
The exploitation of the previous described architecture allowed both robustness and sensitivity tests, whose results are briefly described in the next paragraphs.
1.2 Robustness
To ensure that the results obtained from the simulations were independent from the random distributions, a simple robustness test has been performed:
-
(a)
The model has been run for one hundred times with fixed parameters values but randomly changing, each time, the seed of the main generator (recalling the architecture described in F1, this means a different random seed, each time, for the more than one thousand objects involved in each simulations).
-
(b)
For each simulation the average values of the productivity are measured at the 2000th production cycles (because each simulations employed one thousand agents, it means 2 million production cycles for each measure, that was based on several millions random tossed values for different decisions).
-
(c)
After one hundred values of productivity obtained running the simulations, the mean, variance of the values, a succession has been computed to evaluate the independence of the results from the random seeds distribution.
Table 7 resumes the results: the variance of both productivity and commons dimension is quite low and seems to confirm the simulations results are not determined by employed random distributions, neither Pearson’s r value shows correlation between results and random seeds.
Rights and permissions
About this article
Cite this article
Antonelli, C., Ferraris, G. The creative response and the endogenous dynamics of pecuniary knowledge externalities: an agent based simulation model. J Econ Interact Coord 13, 561–599 (2018). https://doi.org/10.1007/s11403-017-0194-3
Received:
Accepted:
Published:
Issue Date:
DOI: https://doi.org/10.1007/s11403-017-0194-3