GlobelicsAcademy2015_HFM_catch up_Final

MODELING CATCH UP WITH

HISTORY FRIENDLY MODELS

Franco Malerba

CRIOS Bocconi University

Globelics Academy, Tampere, June 2015

• Why history friendly models (HFM) ?

• The characteristics of HFM

• The 2016 CUP book

• HFM and catch up

• HFM and public policies for catch up

• The road ahead

WHY HISTORY FRIENDLY MODELS?

VARIETY IN THE CHARACTERISTICS AND EVOLUTION OF INDUSTRIES

From the empirical cases and the historical analyses of semiconductors, computers, pharmaceuticals, aircraft, chemicals, textiles, and so many other industries it is evident that:

• The characteristics of industries differ

• The evolution of industries presents a wide variety of patterns

-

• A rich set of factors can be identified: various types of capabilities, innovative users, vertical and horizontal boundaries of firms, actors such as universities or government, specific institutions, and so on.

CAN EXISTING MODELS ADDRESS EXAMINE THESE EVOLUTIONS?

NO

A. Evolutionary models à la Nelson-Winter 1982

These models have microeconomic learning processes; selection with heterogeneous population of firms; de-strategising conjectures; processes of experimentation and imperfect trial and error

Nelson and Winter, 1982; Dosi, Kaniovski and Winter, 1999

Recognition of some stylized facts and development of an evolutionary model able to reproduce those phenomena (i.e. the relationship between innovation and concentration; diffusion curves)

Very abstract models.

Focus on some generic basic properties of industrial structure and dynamics

B. Industry life cycle models

They mainly focus on the relationship between product and process innovation, entry and firm growth, exit, industrial concentration

Basic model of industry life cycle derived from the evidence of the auto industry (Klepper, 1996)

Different industry life cycles and divergence from the standard model due to factors such as the characteristics of demand, technological discontinuities, the type of competition and innovation, pre-entry experience and first mover advantages, as in several models by Klepper and associates

HOW TO MODEL THE EVOLUTION OF DIFFERENT INDUSTRIES

AND THE VARIETY OF FACTORS AFFECTING EVOLUTION ?

• In sum, except for some versions of the standard industry life cycle model, there are no models which focus on the different characteristics and evolution of industries and on the factors that have been examined by historical analyses and case studies

• This has been the initial original motivation for the development of history-friendly models (HFMs).

• The first HFM was published in 1999 (ICC, 1999)

• A book on HFMs by Malerba, Nelson, Orsenigo and Winter is on its way

(CUP, 2016)

• Here I want to present the motivation of these models, the broad features, the building steps, the main blocks, what we have learned so far and the road ahead, and finally present an example of a HFM

WHAT ARE HFMs ?

• HFMs are agent based simulation models which aim to capture in stylised form qualitative theories about mechanisms and factors affecting innovation and industry evolution

• These mechanisms and factors are put forth by empirical research in industrial organization, business strategy and organization and by the histories of industries

• They aim at establishing a link between empirical evidence, developing stand alone simulation models and formal theory

• HFMs aim to explore whether particular mechanisms and forces built into the model can generate (explain) the patterns examined

• HFMs are guided by verbal explanations and appreciative theorizing

THE MAIN FEATURES OF HFMs

• HFMs are evolutionary models: firms are boundedly rational agents; their behavior is guided by routines; learning is a key process; heterogeneity of agents and capabilities characterizes the industry

• These models aim to analyze factors affecting the (short-run and long-run) dynamics of technology, innovation, market structure, industry architecture and industrial leadership

• HFMs are complex dynamic stochastic systems

• Non linearities are present

• Bottom up perspective: aggregate properties are emergent out of the repeated interaction among agents

THE STEPS IN HFM

• Study of the characteristics of the phenomenon under examination

(innovation, evolution of the industry...). Identification of the main features to be analyzed and of the appreciative explanatory model, and consequently also of the hypotheses to be examined and tested.

• Development of the model. Translation of the theoretical structure into the programming language (variables, methods and algorithm) of the model. Repeated process of internal verification of the correctedness of the computer implementation, and balance between accuracy of the model and explanatory power.

• Runs, calibration, analysis of the results and sensitivity analysis

Good surveys are Garavaglia, 2009 and Yoon 2009

THE EMPIRICAL VALIDATION OF HFM (1)

• It is not the purpose of history-friendly modeling to produce simulations that closely match the quantitative values observed in the histories under investigation.

• The goal is to match overall patterns in qualitative features, in particular the trend behavior of the key descriptors of industry structure and performance of a sector

• In a sense, HFMs represent also an abstraction from the specific motivating historical episode

• The goal is to feature some particular causal mechanisms that have been proposed by the appreciative theories for the empirical phenomena under examination

• So, HFMs do not attempt detailed quantitative matching to historical data, nor detailed calibration of the parameters

THE EMPIRICAL VALIDATION OF HFM (2)

• There is some common sense guidance and some basic learning from the case studies in the choice of the plausible orders of magnitude of the parameters

• Moreover some of the dimensions known as relevant are not easily measurable, for example some rules and behaviors

• Some value choices for parameters involve implicit unit choices for variables, which means that the quantitative variables are at the end somewhat arbitrary. However the relations among parameters have to be made with a view to consistency

• So the methodology is different from the one by Werker and Brenner

(2004) in which models are constructed using detailed empirical data on assumptions and on implications

RUNS THAT MATCH AND THAT DO NOT MATCH

THE QUALITATIVE FEATURES OF THE HISTORICAL PATTERNS

• Once the model is developed and the basic runs done, HFMs test the possibility of different outcomes if the parameter values of some key variables are changed.

• More than developing different histories, this is a way of testing the causation mechanism which are present in the model

• For example, if the model proposes and the simulations show that one of the reason for industrial concentration is the presence of a high bandwagon effect at the demand level, would industrial concentration be lower if the bandwagon parameter is significantly smaller?

• This is particularly useful as a normative tool in order to examine different policies or socio-economic settings

INNOVATION AND INDUSTRY EVOLUTION:

HISTORY FRIENDLY MODELS

Malerba Nelson Orsenigo and Winter CUP 2016

• Part 1:

• C. 1 The subject matter

• C. 2 Methodology

• Part 2: models of industries

• C.3 Computers

• C.4 Computers and semiconducors

• C. 5 Pharmaceuticals

• Part 3: Conclusions

• Chapter 3 The computer industry (1950-1985)

Cumulativeness and increasing returns along different product trajectories.

Technological and market discontinuities. Bandwagon effects at the demand level.

Emergence and consolidation of concentration in some product segments, entry and competition in others, depending on the level of bandwagon effects.

• Chapter 4 The coevolution of the semiconductor and computer industries

(1950s-1985).

High opportunity conditions and major technological and market discontinuities in two vertically linked industries.

Specialization and vertical integration as co-determined by the dynamics of capabilities, technology and firm size in the upstream and downstream industries.

• Chapter 5 The pharmaceutical industry (from the early period to molecular biology)

Low cumulativeness of technological advances, segmented demand and presence of IPR regimes.

Generation of low level of overall concentration in the era of random screening. Low level of overall concentration also after the biotechnology revolution, but change in the composition of the industry with new actors

WHAT DID WE LEARN? (1)

From all these models, we have learned that the following factors affect the short-term and long term dynamics of innovation and market structure and more broadly the evolution of industries

• FIRMS CAPABILITIES:

Learning processes

Level, types and distribution of capabilities

• TECHNOLOGICAL REGIMES:

Technological regimes in terms of technological opportunities, knowledge accessibility, cumulativeness and appropriability

WHAT DID WE LEARN ? (2)

• DEMAND:

Homogeneous, bandwagon and network effects, segmentation, experimental customers, new demand, user-producer relationships

• INSTITUTIONS:

Universities, public policy (including competition policy and IPR)

The joint analysis of capabilities, technological regimes, demand and institutions provide a systematic analysis of the dynamic effects of the main factors affecting innovation, market structure and the evolution of industries

WHAT DID WE LEARN (3)

In addition, HFM have greatly progressed our understanding of three key issues in industrial dynamics and industry evolution:

• THE VERTICAL STRUCTURE OF PRODUCTION AND INNOVATION :

The vertical structure of production and the division of innovative labor is the result of the level and distribution of capabilities, technological regimes, demand conditions and the institutional environment. At the same time it is a major factor affecting their changes

• SELECTION: Selection processes are not totally exogenous. They emerge out of the interaction among individual firms. Exit rules are as important as entry conditions

• EFFECTS OF PUBLIC POLICIES: There are major interdependent, dynamic and unintended effects of public policies

A history-friendly model of catch up cycles

Joint work with Fabio Landini

We have developed a history-friendly simulation model of catch-up cycles, which is general enough to investigate the role of different windows of opportunities and of the speed and type of domestic firms learning

We have examined firms in three countries: a leader country and two follower countries.

Firms in both countries learn. Learning is cumulative and systemic.

Firms may sell in one country only or on both countries

Technological windows are examined. Windows open at a certain time during the evolution of an industry

History-friendly model of technology-driven catch-up cycles

1 st generation technology

2 nd generation technology

Historical Cases (I)

Mobile phone industry:

• Until the beginning of the 1990s Motorola (US) dominated the analog-based cell phone sector;

• With the emergence of digital technologies, Nokia (Finland) dethrones Motorola, which tended to stay along with analog technologies;

• In the smart phone Era, finally, Samsung (Korea) dethroned Nokia;

Two shifts in the techno-economic paradigm, causes two waves of successive catch-up.

Historical Cases (II)

Camera industry:

• Until 1959 German companies (Leica, Contax) dominated the rangefinder camera market

• The emergence of analog and digital single lens camera led

Japanese companies (Nikon, Canon) to take industrial leadership away from Germany

• In the era of compact system camera, finally, Korean companies

(Olympus, Panasonic, Samsung Elec) are the new industry leaders;

Once again, two technology shifts led to two leadership change….

History-friendly Model

Structure of the model: r

0 r

1 r

2 r

3 r

4

…

(innovation activities) firms of country A

Technology space firms of country B

(market activities) market of country A market of country B r

J firms of country C market of country C


Firm-level effects:

• Firms are born with a technology that allows them to search a restricted portion of the technology space; over time new technologies emerge and firms have the chance to adopt them;

• Firms have heterogeneous capabilities: higher capabilities allow firms to look for better techniques and improve market share;

• Demand is vertically segmented: the higher the firm’s capabilities and technical merit, the higher the product quality, the higher the market share;

• Firms have a competitive advantage in serving their national market.


Country-level effects:

• System of innovation: the higher the average value of the technique in a country, the higher the probability that firms of that country find better techniques;

• Lock-in effect: the probability that a firms of a given country perceives a new technology decreases with the country’s market share using the old technology;

• Learning: firms improve their capabilities over time; the growth of a firm’s capabilities is faster the higher the average level of capabilities in the firm’s country

Simulation Dynamics

1 st Gen. technology r

0 r

1 r

2 r

3 r

4

…






1 st Gen. technology r

0 r

1 r

2 r

3 r

4

…






2 nd gen. technology

1 st gen. technology

Technology space r

0 r

1 r

2 r

3 r

4

…

(innovation activities) Window of opp. 1 firms of country A firms of country B r

J firms of country C

(market activities) market of country A market of country B market of country C





0 r

1 r

2 r

3 r

4

…

(innovation activities) firms of country A firms of country B r

J firms of country C



3 rd gen. technology




0 r

1 r

2 r

3 r

4

…

(innovation activities) firms of country A r

J

Window of opp. 2 firms of country B firms of country C


Results

‘History-friendly’ runs:

• Symmetric advantage in national markets; markets of country B and

C starts relatively small (compared to A) and enlarge over time;

• Two technology shifts in periods 100 and 200 respectively; country

B enter at period 50, country C enter at period 150.

Evolution of Herfindal index

Evolution of market shares

Counterfactuals

Experiment 1 – Lock-in effects

• We reduce the intensity of lock-in effects

Evolution of market shares

Evolution of Herfindal index

Counterfactuals

Experiment 2 – Size of the window:

• We reduce the size of the window -> the smaller the window, the lower the probability that a change in leadership occurs

Experiment 3 – Shape of technical landscape:

• We make the linear landscape more linear (i.e. less radical innovations) -> the less radical innovations, the lower the probability that a change in leadership occurs

Conclusion

We have developed a history-friendly simulation model of catch-up, which is general enough to investigate the role of windows of opportunities and of the speed and type of domestic firms learning

We simulate a technology-led process characterized by three cycles and two times of leadership change.

By relying on counterfactuals we test the causal effect of lock-in, size of the window, and shape of technology space.

Technological windows of opportunities do affect catch-up. When they are absent or too small, catch up is very difficult to succeed.

Learning is an important factor in catching up, particularly when new opportunities are created

Public policy and catch up: a history-friendly model

The questions: a) Which types of public policies favour catch up?

b) Is there complementarity among different types of policy?

c) Do policies have the same effect in different technological contexts (incremental vs discontinuous change) ?

Policy analysis conducted with a simulation model based on empirical cases of Lee and Malerba (2015).

Other simulation models have discussed policies but not w.r.to

catching up - Dosi et al. 2010, 2013; Malerba et al. 2001, 2008;

Dawid et al. 2012; Neugart,2008; Happe et al, 2008; Mannaro et al, 2008.

The model used in this paper

Structure of the model (Landini, Lee, Malerba, 2013) : r

0 r

1 r

2 r

3 r

4

…


Technology space firms of country B r

J

(market activities) market of country A market of country B



0 r

1 r

2 r

3 r

4

…



J




0 r

1 r

2 r

3 r

4

…



J




0 r

1 r

2 r

3 r

4

…



J




0 r

1 r

2 r

3 r

4

…



J




0 r

1 r

2 r

3 r

4

…



J


CASE 1: No technological discontinuity

r

0 r

1 r

2 r

3 r

4

…



J


No technological discontinuity r

0 r

1 r

2 r

3 r

4

…



J


CASE 2: Technological discontinuity

1 st Technology r

0 r

1 r

2 r

3 r

4

…



J


Technological discontinuity

1 st Technology r

0 r

1 r

2 r

3 r

4

…



J



2 nd Technology

1 st Technology


0 r

1 r

2 r

3 r

4

…

(innovation activities) firms of country A firms of country B r

J



2 nd Technology

1 st Technology


0 r

1 r

2 r

3 r

4

…


Lock-in effect firms of country B r

J


Types of public policy

Public policy is modelled in terms of different values of countryspecific parameters. In particular we consider five policies: a) Capability building: increase in the average level of capabilities newly-born firms are endowed with; b) Support to firms’ learning: strengthening of the link between firms’ learning and country’s average capabilities; c) Market protectionism: higher cost of export (indigenous market advantage, tariff protection and import restrictions); d) Entry support: increase in the probability that new firms enter the industry;

Simulations

The structure of the simulation is as follows:

1) First we analyse the case of policy interventions (both in isolation and in pairs) in the latecomer country without discontinuity.

2) Then, same as 2) but with a technological discontinuity.

By comparing 1 and 2 we can test how similar policies may produce similar or different effects under different technological conditions.

Results show averages over 600 runs

1) Public policy with no technological discontinuity

Result 1: In the absence of technological discontinuity a. the best performing policies are capability-building and support to learning (as from theliterature on development).

b. Protectionism produces a positive effect but smaller than capability building or support for learning c. Entry produces a negative effect (technical change is gradual and cumulative)

1) Public policy with no technological discontinuity


Result 2: Two positive policy complementarities exist: i) protectionism and capability building; ii) protectionism and support to learning

This complementarity can be related to the capability and the national innovation system literature and to the infant industry argument





Catching up in presence of discontinuities

With a technological discontinuity, an opportunity to catch up emerges, but without policy there is no change in industrial leadership.

2) Public policies with technological discontinuity

In presence of a technological discontinuity a) Capability building and support to learning are still effective policies; b) Protectionism produces no sizeable effects (contrary to what we have seen before) c) Entry support becomes a quite effective policy instrument (contrary to what we have seen before)

2) Public policies with technological discontinuity


In presence of a technological discontinuity two policy complementarities can be quite effective: i.

Entry and Capability ii.

Entry and Learning

In a sense, these mixes combine exploration and exploitation.


Conclusions

Main findings:

1) Capability building and support for firm learning are always quite important;

2) Protectionism is relevant only when a technological discontinuity is absent;

3) On the contrary, entry support plays a role only in contexts with technological discontinuity;

4) Policy complementarities exist but they also depend on the presence of discontinuities.

In general, similar policies can perform differently depending on the technological scenario in which they are implemented.

THE ROAD AHEAD

• In sum, HFMs play a key bridging role between general and abstract theories and detailed case studies.

• Because they play this bridging role, they also provide a main message to the two camps: the theorists and the empirical scholars

• To the theorists, HFMs suggest that abstract and general modelling should take into account some degree of realism and contain empirical foundations

• To the empirical scholars, HFMs suggest some degree of formal discipline and modelling of the empirical analyses and historical works, so that rigorous and consistent explanations of industry evolution could be developed

THE ROAD AHEAD (2)

In the future, HFMs may develop along the following lines

• Expansion of industries examined

The so-called traditional industries

Agro-food

Environmental sector (Oltra Saint Jean)

Services

• Comparisons of industry evolution across countries

Comparisons among emerging countries and factors that affect the different evolution

THE ROAD AHEAD (3)

• Enrichment of the models

Institutions (Murmann and Brenner)

Entrants – de novo firms vs spin-offs (Capone)

Origin and emergence of an industry

• Inside the firm

Organization

Behavoral rules

THE ROAD AHEAD (4)

• The third one involves a comparison among HFMs that may identify “in an inductive way” some general properties and some general hypotheses that may apply to a wide range of empirical dynamic phenomena that share some common factors.

In a sense, HFMs force to recognize and model what is different and sector specific in industrial change but also what are the broad similarities and regularities in industry evolution

IF YOU ARE INTERESTED

YOU CAN DO THAT

68

GlobelicsAcademy2015_HFM_catch up_Final

MODELING CATCH UP WITH

HISTORY FRIENDLY MODELS

Franco Malerba

WHY HISTORY FRIENDLY MODELS?

WHAT ARE HFMs ?

A history-friendly model of catch up cycles

Conclusion

Public policy and catch up: a history-friendly model

CASE 1: No technological discontinuity

CASE 2: Technological discontinuity

THE ROAD AHEAD

IF YOU ARE INTERESTED

YOU CAN DO THAT

Related documents

Products

Support

GlobelicsAcademy2015_HFM_catch up_Final

MODELING CATCH UP WITH

HISTORY FRIENDLY MODELS

Franco Malerba

WHY HISTORY FRIENDLY MODELS?

WHAT ARE HFMs ?

A history-friendly model of catch up cycles

Conclusion

Public policy and catch up: a history-friendly model

CASE 1: No technological discontinuity

CASE 2: Technological discontinuity

THE ROAD AHEAD

IF YOU ARE INTERESTED

YOU CAN DO THAT

Related documents

Add this document to collection(s)

Add this document to saved

Suggest us how to improve StudyLib