EXAMPLE on Maestro Itais 2011 Paper20

ITAIS 2011

Conference

LUISS - ROMA

08/10/2011

10:00 – 11:00

Exploring the impact of innovation Exploring the impact of innovation policies in economic environments with policies in economic environments with

self-regulating agents in multi-level self-regulating agents in multi-level complex systemscomplex systems

EXAMPLE

Francesco Niglia Dimitri Gagliardi Cinzia Battistella

Head of ICT Unit Research Fellow Post-Doc Researcher

Innova S.p.A. Manchester University of Udine

Institute of

Innovation Research

ITAIS 2011 – PAPER #20 - EXAMPLE Roma, 08/10/2011

Acknowledgements

The authors gratefully acknowledge fundings from the Regione Puglia under POR Puglia 2007-2013, Asse I – linea di Intervento 1.1 – Azione 1.1.2 “Aiuti agli investimenti in Ricerca per le PMI”. The results presented in this paper are based on the research and development activities of the project MAESTRO. Usual disclaimers apply


Contents

The main question(s) behind the EXAMPLE idea

The positioning

The model adopted

Rules and procedures of the model

Some results

Conclusions


Main question leading the research

How to support policy-makers in defining the actions to improve the performances of a well-defined socio-economic context?

How to estimate the impact of these actions ?


Theoretical context 1/2

The evaluation of the impact of innovation policies on economic systems has been at the centre of the research policy debates for several decades.

• The rationale assumed has often been that of market failure

• Instruments and techniques used to carry out innovation policy impact studies have also been those of classical econometrics and/or qualitative analysis (Fahrenkrog et al., 2002; Shapira and Khulmann, 2003).

• These approaches have based their impact assessments “exercises” on the application of linear (or quasi- linear) theory in deterministic environments


Theoretical context 2/2

Alternative rationales for policy making are only beginning to emerge, for example • the system’s failure approach (Metcalfe, 2005;

Marzucchi, 2011) • the approaches of complexity and networks

(Santa Fe institute, Benjamin and Greene, 2009; Cross et al, 2009; Buisseret et al, 1995, Metcalfe, 1995; Georghiou and Rossner, 200; Georghiou, 2002).


EXAMPLE selected context

Problem: The complexity of innovation policy evaluation and the challenges of modelling the system of innovation

Approach: A modelling strategy based on agent–based modelling in complex systems. • the theoretical foundation of our a-b model is

the Belief-Desire-Intention (BDI) architecture, introduced by Bratman (Bratman 1987)

• The concepts of belief, desire and intention as mental attitudes, to mimic human actions are captured by informational, motivational, and deliberate attitudes of agents respectively


Regional context

In the Region of Puglia, Italy there are some 245 thousand companies and organisations, mostly SMEs, engaged in the agro-food industry constituting the 15% of the total national agro-food sector (ISTAT, 2010 – Data 2007)

The regional agro-food system has been modelled on the actual data available from official databases (i.e. Eurostat and ISTAT). This formalisation has been used to evaluate the variables pertaining to each agent and to collate the time series against which we benchmarked our simulations.


Model consolidation

Equation System reduced

Estimation of the coefficients evaluated in Structural Equation Model (alternatively VAR)

Test of significance R2, t-test for each coefficient, F-stat and ANOVA


Our model

18 agents, 4 categoriesSystem indicatorsI1 Revenues”I2 “Stock”I3 “N Companies”I4 “Employment ”I5 “Sustainability”EquationsΔQ = f (αΔM, βΔR) + ε ΔNa=f (γΔM,δΔR,ηΔI)+ ε ΔI = KeynesianVariablesM = StocksR = RevenuesQ = CompaniesNa = EmployeesI = Gross investments


Inside JADEX

Based on JADEX technology to implement BDI


The ghost agent

Introduced to "smooth" the differences between the modelled (calculated) system and the real economic dynamics that occur in such environments.

Minimise the difference between Result (expected) and Result (official).

The macro-level is the following• Result (simulation) + Result (buffer agent) = Result

(expected) • Result (buffer agent) = Result (official) – Result (simulation)

• N = Variables (i as index)• M = Indicators (j as index) • A, B = normalisation coefficients

• AgentBUFFER = ∑(i=1,N) Ai*Vi + ∑(j=1,M) Bj*Ij


Results – step by step


Results – policy evaluation


Further considerations

Policy • easy-to-use platform for the evaluation of

alternative policy scenarios. • it has given encouraging preliminary results. • appropriate to describe and analyse non-

deterministic dynamics of complex systems • Statistically sound (testable)

Technology • it is easily scalable• needs a model for each economic system• allows validation even with sparse data


contacts

Cinzia Battistella find us also [email protected]

Dimitri [email protected]

Francesco [email protected]

EXAMPLE on Maestro Itais 2011 Paper20

Documents