1

Evaluating Innovation and Labour Market Relationships: The Case of Italy

Luca PIERONI - Fabrizio POMPEI

Department of Economics, Finance and Statistics

University of Perugia.

ABSTRACT In this paper the link between labour market flexibility and innovation is analysed, paying particular attention to

the different technological regimes of economic activities and the different geographical areas of the Italian

economy. A dynamic panel data specification is used to assess the endogenous relationship between patents,

included as a proxy of the innovation, and job turnover and wages which represent labour market indicators. The

results show that higher job turnover only has a significant and negative impact on patent activities in regional

sectors of Northern Italy, while a positive and significant impact of blue and white collar wages has been

generally found.

Key words: Labour market flexibility, Innovation, Dynamic panel data, Endogeneous relationship J.E.L. : R12; J40; O31

Acknowledgement: We are grateful to both the participants of the Italian Association for the Study of Comparative Economic Systems (AISSEC) Conference, Naples, 27-28 February, 2004, and those of the 94th Applied Econometric Association Conference, Naples, 01-02 June, 2006, for their suggestions. All errors are our own.

2

1. Introduction

Traditionally, most of the economic debate concerning technological progress and

employment relied on the classical compensation theory. It currently concentrates on the

impact that different compositions of process and product innovations have on labour-saving

and make questionable the counter-balancing mechanisms, resulting from prices and new

demand, that absorb unemployment (Vivarelli, 1995; Vivarelli and Pianta, 2000; Piva and

Vivarelli, 2005).

New approaches analysing the same relationship, such as the skill-biased technological

changes theory, are concentrated on evaluating the impact of last wave innovations on the

wages and skills of the workforce (Bound and Johnson, 1992; Berman et al., 1994; Johnson,

1997, Mortensen and Pissarides, 1999; Mincer, 2003). Within this context, theoretical and

empirical results show the magnitude of the shift of the relative demand for skilled labour,

yielding a new equilibrium characterized by a higher relative wages and a higher quota of

skilled employment. Therefore, wage inequality and the need to relax the firing and hiring

restrictions in the labour market have been seen as a direct effect of higher innovation

activities.

Despite the dominance of investigations dealing with the unidirectional impact of innovations

towards the labour market, there are some fields of theoretical investigation where particular

market segmentations (Doeringer and Piore 1971; Osterman 1982), or complementarities

between investments in innovation activities and the demand for skilled labour (Acemoglu,

1997a; 1997b; 2002), or innovative milieux (Keeble and Wilkinson, 1999; Lawson and

Lorenz, 1999) have been recognised as determining an endogenous character of the labour

market/innovation relationship. However, very few detailed empirical investigations, both on

the direct impact of labour flexibility on the accumulation of skills and innovative

performances and on their likely endogeneity, have been performed (Capello, 1999; Bassanini

and Ernst, 2002; Michie and Sheehan, 2003).

3

In other terms, the feedback from the effects that employment conditions and the flexibility

levels of labour market have on innovation, has not been studied in depth. This appears

striking given that the European and Italian economic policy debate has been particularly

animated in recent years, both regarding labour market flexibility and productivity questions

(Treu, 1992; Bertola and Rogerson, 1997; Costabile and Papagni, 1998; Zimmermann, 2005).

The lack of flexibility has often been identified as a determinant of a pathological

unemployment rate and has been recognized as hindering investments in innovations.

Nonetheless, the possibility that a circular causality is at play between labour flexibility and

innovation, reflecting on the long-term innovative performances of the economic systems, has

been in great measure neglected.

The present paper attempts to take a step forward by analysing labour market flexibility,

represented by labour mobility and wages, to determine whether it influenced the innovation

activities of Italian industries and regions in the nineties.

Firstly, we consider the endogenous character of the labour flexibility/innovation relationship,

by means of a dynamic model, paying attention to the likelihood of circular causality.

Secondly, the same relationship is assumed to be strongly context-dependent. In other terms,

we take into account both the specific technological context at the sectoral level (Malerba and

Orsenigo, 1996; 1997) and the different regional development patterns (Cooke et al.,1997;

Capello, 1999; Keeble and Wilkinson, 1999; Lawson and Lorenz, 1999). Moreover, with

respect to other surveys concerning the Italian case and showing very similar aims (Capello,

1999), our investigation not only takes into account the endogeneity problem but also includes

all Italian regions and manufacturing industries1.

The remainder of the paper is organized as follows. In section 2 we develop the conceptual

framework supporting the empirical analysis. Section 3 focuses on the variables implemented

in the econometric model and presents some descriptive statistics. Details on the econometric

specifications and a brief discussion on the Dynamic Panel Data estimator are reported in

4

section 4. Finally, in section 5, the estimated results are discussed, while final considerations

are given in section 6.

2. The conceptual framework of empirical analysis

The question of labour market flexibility has been widely investigated, but many points

remain controversial. Thus, most authors recognised that the term “flexibility” can assume

different meanings, depending on the context of the political debate or on the theoretical point

of view of the analysis. For example, Piore (1986, 2004) highlighted that since the 1980s a

different way of interpreting the flexibility of labour has become rooted in North American

and European business communities. Other large surveys stressed several dimensions of

flexibility according to different schools of economic thought: e.g. institutionalist vs neo-

classical theories (Creedy and Whitfield 1988) or post-fordist vs managerialist views

(Brewster et al. 1997).

Moreover, labour flexibility can be discussed in different ways, depending on the elements of

the economic system and on the nexus taken into consideration, e.g. labour flexibility and

unemployment, labour flexibility and innovation, labour flexibility and the firm’s

performance.

In order to provide theoretical support for the current empirical analysis, in the next sub-

sections we limit our survey by using only those conceptual tools that are useful to explore the

labour flexibility and innovation nexus, without neglecting the particular views with which

this relationship has been implicitely or explicitely treated by some of the main schools of

economic thought.

2.1 Labour flexibility and innovation according to the institutionalist view

Undoubtely theories of internal and dual labour markets, stemming from an institutionalist

view, constituted an important challenge to the wage competition model used by the

5

traditional neo-classical school. By studying the level of inter-firm labour mobility, Doeringer

and Piore (1971) stressed the presence of local labour markets where low mobility results

from the efforts of employers to reduce turnover in order to preserve skill-specificity. These

skills are only useful in a small range of jobs and show a high complementarity with other

specific resources of the firm. The consequence is that firms draw a distinction between

incumbents and otherwise similar workers outside the firm. Therefore, skill-specificity is seen

to promote the restriction of the lower job classifications into an internal labour market and

higher mobility occurs within the firm rather than between firms. It is worth noticing that in

this case also the reverse causal nexus holds: an internal labour market protects the

accumulation of skill-specificity and favours incremental innovations within the firm.

This early fordist view has been modified because decentralisation of the productive structure

occurred in the most developed countries during the 1970s and 1980s. The interpretation of

these processes in fact relied on the shift from mass production to flexible specialisation

systems (Piore and Sabel 1984; Tolliday and Zeitlin 1986; Lash and Urry 1987).

Consequently, the segmentation of the labour market into a primary sector, where a more

stable skilled labour force operates, and secondary sector, characterised by unskilled workers,

lower wage levels and higher job turnover rates, has also been seen as occurring within large

firms and as favouring the de-verticalisation processes (Osterman 1982)2.

Focusing on the micro-level, Atkinson (1985a; 1985b; 1986) identified three different

dimensions of the flexible firm: a numerical flexibility, which is the ability of firms to change

the number of people they employ; functional flexibility, as the ability to vary the amount of

labour that firms use, without resorting to the external labour market; wage flexibility, that

represents the ability of pay and payment systems to respond to labour market conditions and

to reward and encourage improved performance.

The dimensions mentioned above also characterize the regional level of the analysis, once the

decentralisation process occurs. For example Brusco (1982), have stressed that the

6

outsourcing of the secondary sector from large firms has generated local Small and Medium

sized Enterprise systems (SMEs). Actually the same author, by outlining the workings of a

case study representing the Italian North-East industrial districts model, well highlighted the

heterogeneity of the secondary sector where, besides home-workers and other kinds of

subcontractors, highly skilled workers operate.

Therefore, this secondary labour market sector, mainly made up of small firms, very often

shares the primary sector’s advanced technologies, innovative capacities and, at least in

periods of expansion, the secondary sector returns flexibility in the use of labour to the entire

productive structure. The link between the primary and secondary sector generates flexibility

and entrepreneurship that, in turn, produce higher rates of growth. This virtuous cycle pushes

up family incomes, so enabling them to increase their education and accumulation of skills.

Relying on previous results, the studies realized within institutionalist and evolutionary

paradigms throughout the 1990s notably pointed out the role played by labour mobility in

SME systems.

Supplier/customer relationships, spin-off from universities or other firms, and the inter-firm

mobility of workers have been recognised as the main mechanisms for knowledge

transmission and learning in innovative milieux (Keeble and Wilkinson 1999). In particular,

most innovative activities realized in these regions are based on collective learning, that is, the

creation of an increasing base of common knowledge among individuals enabling them to co-

ordinate their actions in the resolution of technological and organizational problems (Lorenz

1996, Lawson and Lorenz 1999, Capello 1999). Given that the sharing of largely tacit

knowledge promotes the re-combination of the region’s diverse resources, the mobility of

highly skilled personnel in the local labour market guarantees a suitable technological transfer

across firms.

7

The higher mobility of labour supporting collective learning has also been found to be a

crucial determinant in the development of some European High-Tecnology Clusters of recent

years (Keeble and Wilkinson 2000; Longhi and Keeble 2000; Camagni and Capello 2000).

2.2 Labour market and innovation in the neo-classical perspective of the last decades

There are also neo-classical lines of research worth noting which distinguish themselves from

simple wage competition models and focus on job turnover and wage levels from a different

point of view. In this context Labour Turnover (Stiglitz, 1974; Arnott and Stiglitz, 1985;

Arnott et al., 1988) and Job-Search theories (Mortensen and Pissarides, 1997; 1999) aim to

analyse unemployment variability as the result of imbalances between flows into and out of

the job market. It is necessary to remark that in the Labour Turnover framework, innovation

is only tacitly considered while the focus is on the labour mobility-wage structure. Low wages

cause a costly high mobility of labour that, in turn, negatively affects labour costs,

productivity and human capital accumulation of workers. On the other hand, if efficiency-

wage considerations emerge to solve this problem and labour market rules make layoffs

prohibitively expensive, labour mobility decreases in the short term, but rises in the long term.

Firms cannot lay workers off, go bankrupt and an increase in the unemployment level occurs.

In Job-Search theories, the labour market/innovation relationship is explicitly discussed.

According to these theories, job security reduces job destruction. The incentive to create new

jobs in response to the need to change products and production processes is reduced. For this

reason over-restrictive market rules inhibit an efficient reallocation of labour and hinder

innovative activities.

An extension of the Job-Search models was carried out by Acemoglu (1997a; 1997b). Within

this view, when complementarities between workforce skills and technology choice are taken

into account (i.e. an economy with endogenous technology choices), a deregulated labour

market is no longer the best solution. If the turnover rate increases, the firm does not invest in

8

new technology (or R&D) and on-job training for workers, because the additional return on

training, or gains stemming from acquired knowledge in R&D activities will benefit the

worker who will probably soon leave the firm. On the other hand, if workers do not expect

firms to invest in new technology (or R&D), their wages cannot be adequately high and they

do not invest in human capital accumulation. Thus, life-time employment relationships are

important factors contributing to technological changes.

The wage level can play an important role to stimulate innovation as a result of the

performance of innovative and highly profitable firms. But it is not difficult to consider the

equally important reverse direction of the causality. Thus, there are other branches of

literature, within the neo-classical paradigm, underlining that when wages are kept above their

market-clearing level, regulative interventions (minimum wages, union power, normative

traditions) and efficiency are involved (Shapiro and Stiglitz, 1984; Stiglitz and Greenwald,

1995). The disparate contributions to the signalling/incentive literature have been synthesized

within the efficiency wage models (Akerlof and Yellen 1986), which explain why firms find it

unprofitable to reduce wages when there is high unemployment. In brief, wage cuts are said to

harm productivity and, therefore, while they would reduce total labour costs, they may

increase labour costs per efficiency unit.

Finally, it must be mentioned that also for according to some evolutionary and istitutionalist

views, higher level wages exert a direct and positive effect on the active participation of the

workforce in the learning process, enhancing loyalty and commitment, and stimulate

practitioners into developing informal relationships, sharing information and accelerating the

emergence of tacit knowledge (Kleinknecht, 1998; Antonelli, 1999; Kitson et al., 2000).

2.3 The basic hypothesis supporting empirical analysis

The previous discussion leads one to believe that the relationship between labour market

flexibility and innovation is not so straightforward and raises at least four questions:

9

a) To what extent do higher wage levels improve innovation?

b) Is it still possible to find internal labour markets, essentially coinciding within the

firm’s boundary, where the skill specificity that supports innovation is protected by

low inter-firm mobility?

c) Is higher labour inter-firm mobility, characterising the SME contexts and favoured by

less labour market regulation, always the result of an effective balance of interests by

individual producers (embedding in network relations versus loss of proprietary

knowledge) or does it hinder, in some situations, innovative activities?

d) Does the existence of complementarieties between highly skilled workers and

technological choices of employers somehow force us to take into account the

endogenous character of labour flexibility/innovation relationship?

Indeed some empirical works have found that the impact of labour market regulation on

innovation shows different outcomes and reveals a strong context-dependent influence. For

example, Kleinknecht (1998), focusing on the Dutch case, pointed out that the extension of a

policy of restricted wage increases to all the economy, negatively affected the improvement of

labour productivity and innovation in dynamic and hi-tech sectors. He draws important

conclusions, that we will take into account, regarding the limited short-run success of policies

concentrated on overly modest wage increases, downward wage flexibility and various other

attempts to remove labour market rigidities. In fact, in the long run these schemes discourage

productivity growth, product innovation and all other innovative performances of the

economic system.

Bassanini and Ernst (2002) carried out a comparative survey among OECD countries, where

the impact of product and labour market regulations on innovation is highlighted by

distinguishing between different technological intensities of industries. Michie and Sheehan

(2003), using a survey of UK firms, explicitly investigated firms’ use of various flexible work

practices, and the innovative activities of those firms, within the various industrial relation

10

systems. Capello (1999) focuses on three Italian high technology milieux by considering the

different impact of the labour force turnover on process innovations, product innovations and

radical innovations.

In the current paper, we address the four questions mentioned above by referring to the whole

Italian economy in the 1990s. More precisely, we start from the consideration that both the

manner of organizing innovative activities and the geographical contexts are essential when

we explore the labour flexibility/innovation relationship. Therefore, the endogenous character

of this link will be analysed by distinguishing between the different technological regimes of

industries and among the different territorial patterns of development shown by Italian

regions.

As far as the technological regime of an industry is concerned, Malerba and Orsenigo

(1996;1997), relying on empirical works defined it as a combination of technological

opportunities, appropriability conditions, knowledge accumulation characteristics and base

knowledge. The analysis of the organization of innovative activities led the same authors to

identify the classical Schumpeterian sectoral patterns by means of four indicators: i)

localisation of innovative activities; ii) size of innovative firms; iii) permanence in the

hierarchy of innovators; iv) new entry of innovators.

The Schumpeter Mark I pattern (SMI), defined as a creative destruction regime, shows low

concentration of innovative activities at the firm level, instability in the hierarchy of

innovators and higher new entry of small business in innovation activities. Within this context

knowledge spillovers among firms and collective learning are relevant. Therefore the

cumulative process regarding the knowledge that supports innovation occurs at the territorial

level and not at the firm level. The traditional low-tech branches (food industry; textile,

garment and footwear; wood and furniture; non metallic mineral products and metallic

products) are highly correlated to this pattern.

11

Conversely, Schumpeter Mark II (SMII), defining the creative accumulation regime, is

reported in the same empirical analysis as the pattern where the concentration of innovative

activities involves large corporations; the latter show permanence at the top of the innovators’

classification and are eventually less threatened by new innovators. The accumulation of

knowledge, which is more codified in nature, is supported by R&D investments and basically

occurs at the firm level. In this case there is a good correspondance between these sectors and

the so-called hi-tech industries (machinery, electrical equipment, television, office machinery,

medical components, motor vehicles, transport equipment).

In order to enforce our hypothesis, we include as a unit of analysis the Regional Innovation

System (RIS) concept. The RIS is developed within the theoretical context of the National

System of Innovation (NIS), where parallel technological changes in work organization and

production are accompanied by cultural changes or changes in habits and routines (Lundvall,

1993; Cooke et al.,1997; Asheim and Coenen, 2005). The shift from NIS to RIS concerns the

extent of the systemic character of the geographical and administrative area considered, as

well as the territorial range of the knowledge spillover. If the tacit character of knowledge is

recognized as playing a key role in innovation, the latter cannot be easily shared and applied

outside its territory of generation (Amin and Wilkinson, 1999; Antonelli, 2005). This

geographical stickiness of knowledge diffusion and learning process is only one of the main

characteristics of RIS. Within it, firms, other economic agents and local institutions co-evolve

and contribute to shape a specific political-administrative body. Thus, the RIS becomes an

institutional repository of a certain negotiated, evolving, social order that establishes routines,

norms and values by which actors may come to trust each other collectively (Cooke et

al.,1997). Different institutional settings will be likely to give rise to distinctive conventions

or forms of collective social order, leading to the establishment of different kinds of

organization of innovative activities, but also favouring different micro-constitutional

regulations that affect the labour market.

12

Within this conceptual framework, the hypothesis regarding the endogenous relationship

between numerical flexibility (or labour mobility) and innovative activity, can be

differentiated. The numerical flexibility of the labour market can affect the innovative

activities of industries and/or of regions in different ways.

In hi-tech industries, where most of the science based and scale intensive sectors are included,

a SMII pattern structuring the innovative activities is probably working. In this case it is

expected that lower job turnover does not hinder the generation of innovation and/or its

adoption. Knowledge accumulation at the firm level generates a strong incentive to use the

firm’s internal labour market (functional flexibility). The tenure of the workforce allows not

only a simple “learning by doing” process within the firm, but also guarantees a possible co-

evolution among tangible assets, the firm’s core competences and the workers’ skills3.

On the other hand, high turnover rates provide support for the flow of knowledge across small

firms within traditional sectors (low R&D intensity industries), where a creative destruction

pattern (SMI) is probably operating.

The different systems of governance acting at the regional level and stemming from the

evolution of different socio-economical development patterns (Papagni, 1995; Cooke et al.,

1997) could also affect the joint behaviour of labour flexibility and innovative activities. For

example, aside from the technological regimes of a particular industry, higher labour

flexibility could exert a different impact in Southern Italian regions, where the problem of the

adjustement of wages and mobility of labour is deemed to be more severe with respect to the

North of Italy (Faini, 1997).

These arguments provide a theoretical framework to carry out an empirical analysis where

some aspects of labour flexibility and innovative activities are detected.

13

3. Data sources and variables

The empirical results of the relationship between labour market indicators and innovation that

will be presented in the following section concern the manufacturing sectors of Italian

industry over the period 1990-19964. The regional level is taken into account by means of

NUTS 2 statistical units.

As far as the variables are concerned, we chose patent per capita as dependent variable. It

describes innovative activities that have occurred within a specific regional sector of industry.

Patents are a measure of innovative output and are quite “popular” among innovation

scholars, even though they are not inconvenience free (Malerba and Orsenigo, 2000;

Jacobsson and Philipson, 1996; Griliches, 1990). For example, the propensity to patent can

vary across sectors and products (or production processess), according to institutional and

structural characteristics concerning the appropriability of innovations (Malerba and Torrisi,

2000). These characteristics contribute to making the specific technological regime of the

sectors, but at the same time, could severely bias the relationships to investigate.

However, it is worth noting that with respect to other indicators, such as R&D expenditures,

patents often account for informal technological activity, evaluating the amount of innovative

activity of medium and small firms (Malerba and Torrisi, 2000; Ferrari et al., 2002).

Moreover, the patent data used in the present analysis stem from the CRENOS databank and

refer to European Patent Office (EPO) applications. This indicator should be particularly

effective in taking into account potentially high remunerative innovations, which for this

reason are patented abroad (Paci and Usai, 2000). Finally, these patent data, initially classified

by means of the International Patent Classifications (IPC)5, have been converted to the

manufacturing industry, by means of the Yale Technology Concordance, in order to obtain

coherent data with the ATECO91 classification (Paci and Usai, 2000).

As far as labour mobility (or numerical flexibility) is concerned, we chose the gross job

turnover rate. Actually, there is little agreement on using gross job turnover (or job

14

reallocation) as a proxy for numerical flexibility, i.e. less hiring and firing restrictions (Bertola

and Rogerson, 1997; Contini et al., 1996; Boeri, 1996; 1999). In comparative analyses

between European countries and the US, both Bertola and Rogerson (1997) and Boeri (1999)

criticize the use of turnover rate to prove the negligible differences found in flexibility terms.

Conversely, they claim that high wage compression (coming from collective bargaining) and

high rigidity, regarding hiring and firing in the workforce, produce high European and Italian

turnover rates without the presence of real labour market flexibility. We try to take into

account this objection by introducing the wage levels into the model as explanatory variables.

Job turnover also depends on the business cycle (Schivardi 1998). We have taken into account

the overall impact of the business cycle upon innovation/labour market relationships by

introducing temporal dummies in the econometric specification.

In line with the aforementioned literature, we refer to gross job turnover as the sum of job

creation and job destruction that has occurred at the firm level and has been measured by

means of surveys carried out by the National Institution of Social Security (NISS), that

identifies the movement of employment positions across firms6.

More precisely, the average job creation occurring in the regional sector is

( )2/)( 1,,,,

1,,,,,,

,−

−

+

−=∑

tjitji

ftjiftjif

ji NN

EEC (1)

where 1,,,,,, −− tjiftjif EE is the positive difference between jobs registered in firm f, belonging

to region j and sector i, over the yearly period (t and t-1);

2/)( 1,,,, −+ tjitji NN is the average number of firms belonging to region j and sector i, in which

the growth of jobs occurred.

In the same way, the average job destruction of the regional sector is

2/)( 1,,,,

1,,,,,,

,−

−

+

−=∑

tjitji

ftjiftjif

ji NN

EED (2)

15

where 1,,,,,, −− tjiftjif EE is the negative difference, taken in absolute value, between jobs

registered in firm f, belonging to region j and sector i, over the yearly period (t and t-1).

Thus, the average gross job turnover in region j and sector i is simply

jijiji DCGJT ,,, += (3)

Also wage levels have been drawn from the NISS databank. The breakdown to sectoral and

regional level provides yearly average gross real wages7.

The NISS databank allows us to differentiate between the wages of blue and white collars. All

manual labour is included in blue collars, whereas employees in administrative and clerical

positions, technicians, cadres and executives are considered white collars. The simple

distinction in these two categories of workers, accompanied by lack of more detailed data, is

very often used as a proxy of respectively unskilled and skilled labour (Piva et al., 2005). In

our case the information about wage levels can be used as a proxy for skill levels within the

blue collar and white collar groups. Therefore, since the white collar category includes

researchers and other R&D personnel, we can assess whether the efficiency wage effect on

patent activities is only concentrated in this worker group or, conversely, it involves also

high-skilled manual workers.

In order to differentiate the territorial context corresponding to different models of

industrialization we use, as interaction dummies, the classical five geographical macro-areas

(North-West, North-East, Centre, South and the Islands).

As far as the technological context is concerned, we group 10 manufacturing sectors

according to OECD classification, which is used to identify hi-tech/low-tech industries

(Hatzichrnoglou, 1997)8. This method takes into account both the level of technology specific

to the sector (measured by the ratio of R&D expenditure to value added) and the technology

embodied in purchases of intermediate and capital goods. It also corresponds to the Italian

classification of the R&D intensity reported by ISTAT (2001) in the Community Innovation

Survey.

16

It is worth noting that this R&D intensity classification of industries is not trouble-free. The

first limitation concerns the role played by research in the innovation: of course R&D is an

important determinant but it is not the only one, e.g. licences, strategic cooperation between

companies, informal learning and collective learning are other important sources.

Moreover, in the sectoral approach, R&D intensity can be skewed because all research in each

sector is attributed to the principal activity of the firms making up the sector9.

Finally, we must keep in mind that in our case OECD classification of sectors is only an

unrefined proxy of the technological regimes discussed in the previous section. In fact, there

is not perfect correspondence between R&D intensity classification (hi-tech/low-tech) and

Schumpeterian patterns classification (SMI and SMII) provided by Malerba and Orsenigo

(1996) 10. Nevertheless, the need to enforce our hypothesis on the influence of technological

context with assumptions concerning the innovative behaviour of enterpreuners, led us to

deem positively the trade-off between this necessity and the risk of producing an analysis that

was too biased.

The size of the sample sums up to 1,400 observations (20 regions NUTS 2 times 10 sectors

times 7 years). In Table 1 some descriptive statistics on patents and labour market indicators

can be observed. More precisely, the reported data summarize information in the profile of

industries, taking into account summary statistics for regions and years.

As far as the patent activities are concerned, we standardised the number of patent

applications with respect to the population. In any case, the whole absolute number of Italian

patent applications changed from 2,237 in 1990 to 2,069 in 1996, and the average value in

those seven years was 2,212. The standardised values of patent applications reported by sector

in Table 1 show an overall higher inter-industry variability and provides suggestions for both

different appropriability conditions and knowledge accumulation characteristics.

The level of patent activities in some low-tech industries is not completely negligible: for

example, 3.18 patents per million inhabitants in the wood and furniture sector, and 3.27 in the

17

metal products sector are levels comparable with a high-tech sector such as that of motor

vehicles (3.21). Indeed, during the nineties, there were four mature sectors (wood-furniture,

textile, non metallic mineral products and metal products) in which Italy showed international

specialisation in terms of patent demand (Ferrari et al., 2002). There are also economic

activities where the firms’ territorial location in industrial districts plays a key role. Taking

into account this stylized fact, we carried out an analysis restricted to these four sectors trying

to evaluate the influence of industrial districts in the innovation-labour market relationship.

18

Table 1 - Summary statistics by industry (average 1990-1996) Patents per million Inhabitants Turnover Sum Dvst Min Max Mean Dvst Min Max Food, beverages and tabacco 0.36 0.03 0.00 0.21 4.45 0.70 3.49 6.92 Textile products, Wear industry, Leather industry; Luggage, handbags and footwear

1.24 0.10 0.00 0.52 5.61 0.90 3.48 9.96

Wood, Forniture and other manufacturing 3.18 0.17 0.00 0.61 4.76 1.94 3.41 25.13

Paper, printing and publishing 0.62 0.04 0.00 0.19 4.51 0.70 2.53 6.31 Coke and refined petroleum products, Chemical products and synthetic fibres, Plastic products

11.03 0.66 0.00 3.10 7.29 2.12 4.47 14.52

Non metallic mineral products 0.68 0.04 0.00 0.32 5.41 1.04 3.58 10.15 Fabricated and structural metal products 3.27 0.18 0.00 0.60 5.75 1.47 3.94 11.71

Machinery, electrical equipment, television, office machinery, Medical components and Instruments for measuring

32.15 1.74 0.00 6.27 6.35 1.51 3.75 13.79

Motor vehicles, Transport equipment 3.21 0.25 0.00 1.38 14.97 17.92 2.00 121.67

Building 0.12 0.01 0.00 0.04 5.33 0.84 3.55 7.52

Blue collar wages White collar wages Mean Dvst Min Max Mean Dvst Min Max Food, beverages and tabacco 28506 2890 22720 34663 35011 5231 25088 47479 Textile products, Wear industry, Leather industry; Luggage, handbags and footwear

23240 2271 18959 28060 27767 6243 16176 41131

Wood, Forniture and other manufacturing 24757 2317 19751 29143 30604 3873 22116 38461

Paper, printing and publishing 27375 3113 21509 36566 32078 5036 19809 44581 Coke and refined petroleum products, Chemical products and synthetic fibres, Plastic products

24964 2652 15832 30302 32534 4230 24601 46942

Non metallic mineral products 27804 2738 22059 34099 34027 4581 24636 47990 Fabricated and structural metal products 28005 3155 22324 34485 33788 5701 23314 46046

Machinery, electrical equipment, television, office machinery, Medical components and Instruments for measuring

25421 3085 19007 34004 32306 4508 23422 45972

Motor vehicles, Transport equipment 26771 3333 12896 33657 30915 8248 7829 45987

Building 30916 2172 25570 35019 35147 2978 28153 42635

Concerning labour market indicators, higher average turnover rates were found in hi-tech

industries and they were probably the outcome of the severe reorganization processes that

took place in these industries in those years. These processes were accompanied by high

standard deviation, signalling strong differences among regions.

19

It is worth noting that higher wage levels, mainly within the blue collar group, did not occur

in the hi-tech sectors, although it did in some low-tech ones. Finally, the geographical

concentration reported in empirical studies: about 56% of the demands for patents are by

firms situated in the Northern Italy (Ferrari et al., 2002). This fact underlines the importance

of traditional historical factors that concern different models of industrialization.

4 Models and Estimations

The hypothesis that innovation activities are influenced by the wages or labour mobility

indicators has been widely supported by other micro-econometric works (Chennells and Van

Reenen, 1997; Flaig and Stadler, 1994; Mohnen et al.,1986).

In this work a dynamic panel data has been carried out in order to estimate the

aforementioned relationship and, simultaneously, to test the persistent role of the firm’s

behaviour in innovation.

The estimation strategy uses a two-way static panel data approach as a first step. In the formal

way, the static panel data specification takes the following structure:

tititi xy ,',, μβ += (4)

where tiy , is the dependent variable measuring the innovation activity, ',tix is the 1 ×K vector

of explanatory variables and β is a K × 1 vector of parameters. It is assumed that the

error ti ,μ follows a two-way error component model:

tititi ,, νλμμ ++= (5)

where ),0( 2, vti IID σν −

In particular iμ denotes the individual-specific residual differing across sectors, while tλ year-

period effects is assumed to be fixed parameters estimated as coefficients of time dummies for

20

each year in the sample. This can be justified by Italian macroeconomic cyclical fluctuations

concerning the down-turn in the 1990-1996 period.

As recalled above, to measure the relationships between innovation activity and labour market

indicators, two facts should be considered. Firstly, innovation processes are generally

characterized by cumulative effects; thus, it is interesting to specify and test the existence of

persistent behaviours in the innovation process by a dynamic econometric model. Secondly,

the innovation process depends on some relevant explicative proxies of the labour market that

are not strictly exogenous, such that the unidirectional causality relationship could be

questionable.

Arellano and Bond (1991) gave an answer to the first problem by developing a difference

GMM estimator that treats model (4) as a system of dynamic equations, one for each time

period, in which the equations differ only in their instruments, moment condition sets and

endogeneity problems. The following equation describes the dynamic specification:

)()()( 1,,'

1,',2,1,, −−−− −+−+−=Δ tititititititi xxyyy ννβϑ (6)

Since tiy , is a function of iμ , the lagged dependent variable 1, −tiy is also a function of iμ .

Hence, 1, −tiy , a right-hand regressor in (6), is correlated with the error term, leading the OLS

estimator to be biased and inconsistent. Moreover, the fixed effect estimator is biased and

potentially inconsistent even if ti,ν is serially uncorrelated, since 1, −tiy is correlated with

residuals (Baltagi, 2001).

Finally, the transformed equation (6) uses instrumental variables to estimate parameters11 in a

GMM framework, in order to obtain consistent estimates if there is no second order serial

correlation among errors. In particular, the assumption that the idiosyncratic error term in

equation levels is not autocorrelated has two testable implications in the first-differenced

equation: disturbances will exhibit negative and significant first-order serial correlations and

zero second- or higher -order serial correlations.

21

In the Arellano-Bond estimator, Sargan’s test for over-identifying restrictions and a robust

version of the first step of the Arellano-Bond estimation are included to verify the adequacy

of the model specification and the robustness of estimated parameters.

The benchmark specification used to estimate the dynamic relationship between innovation

activity and the labour market, and written for simplicity in levels, is:

titititi xyy ,',1,, μβθ ++= − (7)

where ti,μ follows, as in equation (5), a two-way error component model. Again, iμ denotes

the individual-specific residual. A sector with a major propensity to patent is likely to have

larger innovations year after year so that we can expect to have a large iμ .

The variable tiy , denotes the value of innovation activity at time t (with t = 0, . . . , 7),

belonging to the sectoral group i12.

According to the conceptual framework explained in section 2.3, we expect to find

statistically significant relationships among explanatory variables of job turnover and wage

levels in the innovation activity.

As far as turnover is concerned, the explorative nature of the analysis leads us to suppose that

an overall negative sign could support the predictions of internal labour market theory and the

insights of Acemoglu’s model (1997a), in which the high mobility of labour hinders

respectively the accumulation of skills within firms, but also the innovation investments of

firms and human capital investments of workers before hiring. Conversely, if the result does

not appear statistically significant, a technological or geographical differentiation is needed in

order to explore the same hypotheses in different contexts. With a technological regime

differentiation, we expect that a higher turnover rate affects the innovative activity of the

SMII technological regime negatively, given that knowledge and competences are

accumulated at the firm level and firms benefit from the tenure of the workforce. The opposite

should happen in the SMI regime (proxied by low-tech sectors), where the creative

destruction Schumpeterian pattern holds.

22

After a geographical differentiation, we expect the prediction of Acemoglu’s model and the

internal labour market theory to be confirmed in the macro-area, where both innovative

activities and hi-tech industries are more concentrated, that is in Northern of Italy (Ferrari et

al. 2002).

According to efficiency wages theory and to Kleinknecth’s suggestions (1998), wage levels

are expected to have positive and significant parameter signs.

The explanatory variables on the right hand side of (7), also include one immediate lag of the

value of the innovation activity. Since the data are a collection of sectoral information,

dynamic components control cumulative effects of innovation activities within regional

sectors. In this case, we do not have an a priori idea concerning the expected sign of these

effects.

The assumption of strict exogeneity of labour flexibility variables is not assertable (see

par.2), since the variables could be predetermined or endogeneous, leading to a mis-

specification of the true relationship between labour market indicators and innovation. For

this reason, in order to obtain the best rationale for data, we specify wage levels (both for

white and blue collars) as a predetermined variable, including the possibility that the

unforecastable errors in the innovation activity (at time t) might affect future changes in wage

levels. Moreover, the possibility of a causal relationship between innovations and job

turnover, is questionable if we consider an economy with endogenous technology choice. In

the empirical part endogenous behaviours of the job turnover is assessed, non-rejecting

specifications that depicts the circular causality. From an econometric point of view, we

remark that lagged levels of endogenous variables are available to serve as instruments, while

the different characterization of the job turnover and wage levels as endogenous and

predetermined variables, respectively, reduce the likely multicollinearity when the same

labour market indicators are considered “exogenous”.

23

Summing up, the specification in equation (7) is used as a maintained hypothesis with the job

turnover variable included as an endogenous variable and wage levels as a predetermined

variable, also when we distinguish between hi-tech from low-tech technological intensity

levels and macro-geographical areas. Finally, to evaluate different impacts on innovations

when the statistical parameters of labour market indicators are not significant, interaction

dummies as well as restricted samples are included, aiming to specify restricted hypothesis

over the impact of labour flexibility indicators.

5. Results

The static panel data estimation of specification (4) confirms the statistical significance of the

time-dummy parameters, stressing the need for testing dynamic panel data 13. Indeed, as

previously mentioned, problems concerning the statistical serial correlation as well as the

presence of endogeneity among labour market indicators and innovation activity could be

solved simultaneously by taking into account models specified dynamically. The estimation of

the baseline specification of equation (7) by the Arellano and Bond estimator (1991) is shown

in Table 2.

24

Table 2 – Estimation of baseline specifications

The two columns report separate estimated results of different groups of workers, blue and

white collar respectively, using a mix of statistics for one-step and two-step procedures and

controlling for heteroscedasticity in data. In particular, the two-step Arellano-Bond estimator

is implemented to obtain consistence of the Sargan test since this test is over-rejected in a one-

step framework, while one-step estimations, corrected for heteroschedasticity, are used for

inference on the coefficients.

The estimated parameters in column 1 of Table 2 suggest that only blue collar wages have a

meaningful impact on patent performances, taken at the regional level. More precisely, the

higher wages of blue collars seem to improve innovative activities, whereas neither job

Dependent Variable: Patents (1) (2)

Patents (t-1) -0.1828 -0.1944 (-1.18) (-1.21)

Turnover 0.0007 0.0008 (0.61) (0.72)

Blue collars wages 0.0161 (2.42)

White collar wages 0.0004 (2.41)

Time Dummy 1993 -0.0418 -0.0427 (-2.80) (-2.91)

Time Dummy 1994 -0.0494 -0.0665 (-1.78) (-2.40)

Time Dummy 1995 -0.0601 -0.0799 (-1.43) (-1.93) Time Dummy 1996 -0.1038 -0.1255 (-1.93) (-2.38)

Constant 0.0066 0.0177 (0.47) (1.50)

Arellano Bond test Ho: non- autocorrelation (first order)

z=-2.09 (0.036)

z=-1.89 (0.059)

Arellano Bond test Ho: non- autocorrelation (second order)

z=-1.03 (0.304)

z=-1.10 (0.271)

Sargan test (Prob>χ2) (0.6009) (0.5432)

z value in brackets

25

turnover nor the cumulative effect of technology (the lagged dependent variable) play a role in

this general specification. In the second column, where we replace blue collar wage levels

with the white collar ones, the same result holds; we remark that the positive impact on

innovative activities of the latter is slightly less stressed. Moreover, the significant influence of

temporal dummies, with a negative sign, underlines the role played by cyclical fluctuations.

Probably the downturn period linked with the sample that has characterized the Italian

business cycle, negatively affected R&D investment levels that, in turn, discouraged patent

activities14.

The significant inference of the dynamic specification is supported by the p-value of the

Sargan test (0.60 and 0.54 respectively), non-rejecting the included instruments. Confirming

the validity of the dynamic panel data specification, the first-order no-autocorrelation is

rejected at the usual five percent level, while a second or higher autocorrelation order is

rejected.

An interaction dummy has been included in the model in Table 3, in order to test the

sensitivity of job turnover to the geographical differentiation.

26

Table 3 - Estimation by territorial differentiation

Once again, both the first and second autocorrelation tests are coherent with a dynamic

specification of the panel data in each equation reported below, as well as with Sargan tests.

In first column of Table 3, where the specification includes blue collar wages as the

predetermined variable, job turnover exerts a significant and negative impact in the North-

Dependent Variable: Patents (1) (2) (3)

Sub sample North-West and

North-East

Patents (t-1) -0.1833 -0.1954 -0.0486 (-1.19) (-1.22) (-0.43)

Turnover 0.0002 0.0022 -0.0029 (0.24) (0.85) (-1.94)

Blue collars wages 0.0172 0.0435 (2.48) (2.89)

White collar wages 0.0051 (2.84) NorthWest*turnover -0.0021 -0.0036 (-2.46) -(1.42) NorthEast*turnover -0.0020 -0.0037 (-2.17) (-1.37) Centre*turnover 0.0349 0.0025 (0.36) (0.25) South*turnover 0.000001 -0.0034 (0.00) -(1.11) Time Dummy 1993 -0.0425 -0.0438 -0.0713

(-2.79) (-2.94) (-2.32)

Time Dummy 1994 -0.0466 -0.0649 -0.0980 (-1.69) (-2.36) (-1.57)

Time Dummy 1995 -0.0601 -0.0812 -0.1075 (-1.43) (-1.96) (-1.12)

Time Dummy 1996 -0.1032 -0.1253 -0.2482 (-1.90) (-2.36) (-2.02) Constant 0.0056 0.0169 -0.0012 (0.40) (1.44) (-0.04)

Arellano Bond test Ho: non- autocorrelation (first order)

z=-2.07 (0.038)

z=-1.85 (0.064)

z=-2.37 (0.018)

Arellano Bond test Ho: non- autocorrelation (second order)

z=-1.07 (0.285)

z=-1.15 (0.248)

z=-1.27 (0.202)

Sargan test (Prob>χ2) (0.6194) (0.6008) (0.2222)

z value in brackets

27

West and North-East of the country. Conversely, the same geographical interaction dummies

lack statistical significance when we replace white collar wages with the blue collar ones

(column 2). The significance of the results obtained for parameters in the North-West and

North-East regions is increased by the estimation of the equation in column 3 with a sample

restricted to these areas. As expected, the conditional estimation shows a negative and

statistically significant parameter for job turnover, while the robustness of the blue collars’

parameter is remarkable with respect to the unconditional estimation of column 1 (column 3).

As mentioned in section 3, the patent demand is mainly localized in these areas. Therefore,

this finding is not negligible and provides support for insights stemming from internal labour

markets theory and more recent views summarized in Acemoglu (1997a), in which higher

inter-firm mobility increases hiring costs, while uncertainty about the tenure of job relations

hinders accumulation of specific skills by firms and negatively affects innovation activities. It

is also worth remarking on the crucial role played by higher blue collar wages: patent

activities benefit more from informal knowledge accumulation favoured by incentive effects

operating upon the skilled manual labour force.

The impact of job turnover on innovation activities is not clarified by the technological

differentiation of industries (Table 4).

28

Table 4 - Estimation by technological intensity of industries The remarkable outcome of these estimations is the different behaviour of the wages of each

category of workers. In hi-tech industries, only the blue collar wage levels influence

innovative activities, acting as a sort of binding factor (column 2). Probably in this context the

problem was not the lack of research, but the following set-up of the product or process to

patent, carried out by qualified blue collars. Conversely, in low-tech sectors the pecuniary

incentive for white collars was the real binding factor (column 4), as signalled by the

significance of the positive coefficient of this category. Statistically, almost all specification

tests are significant. Only in low-tech industries, where blue collar wages are considered as

the predetermined variable (column 3), could the Sargan test be questionable (p-

Hi-Tech sectors Low-Tech sectors Dependent Variable: Patents (1) (2) (3) (4)

Patents (t-1) -0.8761 -0.1250 -0.097 -0.1167 (-0.65) (-0.83) (-1.08) (-1.39)

Turnover -0.0001 0.0006 -0.0021 -0.0004 (-0.10) (0.58) (-0.45) (-0.11)

Blue collars wages 0.0232 0.0123 (2.51) (1.59)

White collar wages 0.0004 0.0003 (1.18) (2.19)

Time Dummy 1993 -0.1482 -0.1320 -0.0086 -0.0115 (-2.65) (-2.60) (-1.12) (-1.71)

Time Dummy 1994 -0.1779 -0.2001 -0.0028 -0.1506 (-1.96) (-2.17) (-0.27) (-1.75)

Time Dummy 1995 -0.2561 -0.2718 0.0061 -0.0093

(-1.89) (-1.98) (0.39) (-0.85)

Time Dummy 1996 -0.4006 -0.4128 -0.0019 -0.0189 (-2.34) (-2.43) (-0.09) (-1.19)

Constant 0.06413 0.0766 -0.0114 -0.0026 (1.77) (2.12) (-1.08) (-0.57)

Arellano Bond test Ho: non- autocorrelation (first order)

z=-2.57 (0.010)

z=-2.20 (0.028)

z=-3.04 (0.002)

z=-3.23 (0.001)

Arellano Bond test Ho: non- autocorrelation (second order)

z=-0.64 (0.521)

z=-0.79 (0.428)

z=-1.28 (0.199)

z=-1.43 (0.152)

Sargan test (Prob>χ2) (0.2354) (0.4070) (0.0785) (0.5179)

z value in brackets

29

value=0.0785). However, since the p-value is greater than the usual critical value we accept

valid instruments in the estimation.

It must be remarked at this stage, that the estimated parameters concerning the persistence and

cumulative character of patent activity levels is not significant in all specifications. On the one

hand, this could mean that a general difficulty to systematically make innovations both at the

regional and sectoral level exists, but, on the other hand, the same result could simply indicate

that there was only an occasional propensity to patent radical innovations that randomly

occurred in the Italian productive systems in the nineties.

The last estimation concerns four mature sectors (textile, wood and furniture, non metallic

mineral products and metal products) quoted both for their relevant contributions to

technological specialisation in patent terms and for their plentiful supply of qualified workers

(Ferrari et al., 2002). The patent stocks and flows obtained in these branches have been

relevant in Italy compared with other OECD countries and have contributed to the

technological specialisation in low-tech sectors. Within this context, we have explored labour

market-innovation relationships differentiating between the presence (or absence) of industrial

districts in at least one of the four sectors, taken at the regional level. The results are

illustrated in Table 5. Firstly, we can observe that job turnover is neither sensitive to particular

low-tech sectors nor significant to district effects, as shown by the non-significant values of

the respective estimated coefficients. Moreover, in the sample characterized by regions that

include industrial districts (column 2), it is worth noting the negative sign of the lagged

innovation variable, as well as the positive impact of white collar wages which are both

statistically significant. According to the previous result concerning the estimations for low-

tech sectors, only the latter exert a positive impact on patents. However, the parameter size

indicates that white collar wages play a more important role in the industrial districts relative

with the aforementioned four sectors as compared to the whole low-tech sector group.

30

Table 5 - Low-tech sectors that displayed patents specialisations

Finally, the negative impact of lagged dependent variable highlights that patent activities

follow a cycle within the industrial districts of “Made in Italy”. In this sectoral and

geographical context, it is known that patent activities depend on the skills of a few firms or

in some cases, only to one. Thus, since the same leaders could be the producers of patents,

their flow follows the periodicity of research efforts and patent achievement within each

industrial district, so that this behaviour does not spread over the local productive systems and

an accumulation process does not occur.

Regions with districts Regions without districts

Dependent Variable: Patents (1) (2) (3) (4)

Patents (t-1) -0.1282 -0.2091 -0.1625 -0.1294 (-1.72) (-2.59) (-1.33) (-0.98)

Turnover -0.0054 -0.0028 -0.0013 0.0013 (-0.53) (-0.33) (1.15) (0.97)

Blue collars wages -0.0004 0.0001 (-0.38) (0.34)

White collar wages 0.0006 0.0001 (2.42) (0.50)

Time Dummy 1993 -0.0310 -0.0312 0.1172 0.0116 (-2.00) (-2.10) (1.43) (1.33)

Time Dummy 1994 -0.0103 -0.0456 0.0056 0.0046 (-0.43) (-2.25) (0.60) (0.48)

Time Dummy 1995 0.0026 -0.0304 -0.0007 -0.0015

(-0.09) (-1.21) (-0.06) (-0.11)

Time Dummy 1996 -0.1592 -0.0457 -0.0015 -0.0022 (-0.41) (-1.24) (-0.08) (-0.11) Constant 0.0089 -0.0873 0.00002 -0.0005 (0.62) (-0.88) (0.00) (0.10)

Arellano Bond test Ho: non- autocorrelation (first order)

z=-2.90 (0.004)

z=-2.50 (0.012)

z=-1.82 (0.068)

z=-1.78 (0.074)

Arellano Bond test Ho: non- autocorrelation (second order)

z=-1.74 (0.081)

z=-1.57 (0.116)

z=-0.01 (0.991)

z=0.19 (0.849)

Sargan test (Prob>χ2) (0.722) (0.625) (0.978) (0.985)

z value in brackets

31

6. Conclusions

In this paper, we have investigated the links between labour market flexibility and the

innovative activities of the Italian economy, from both the point of view of the technological

context and the geographical one. According to the theoretical literature that stresses the

importance of complementarities between technological choices of entrepreneurs and human

capital investments of workers and more generally recognizes a circularity in the causal

nexus, we have tested dynamic specifications in order to account for the likely endogeneity of

labour market indicators with innovation.

Despite the fact that we undoubtely faced difficulties in dealing with variables that were not

inconvenience-free, some findings are worth noting at least to open the way to further

investigations.

In almost all specifications, both blue and white collar wage levels have shown a positive

impact on patent activities. This means that where efficiency wages considerations emerge

and a distribution of wealth policy favours wage increases, we find innovative activities to be

more intensive. Therefore, in the Italian economy of the nineties, strategies, that stimulated

labour of better quality and incentives, that improved collaboration of personnel within firms

could result in successful innovative performances.

On the contrary, the gross job turnover, taken as indicator of labour market mobility, has not

shown an overall statistical significance. Nonetheless, the results obtained through the

geographic differentiation are not negligible: in regions where patent activity is more

significant (the North-West and North-East of the Italy), labour mobility exerts a negative

impact on innovation.

Undoubtely this finding needs to be more thoroughly investigated. At this stage we can only

conjecture that intensive patent activity could occur within internal labour markets coinciding

with large and medium firms of Northern Italy, where lower inter-firm mobility protects the

32

accumulation of firm-specific skills and/or favours the simultaneous choice of technology

investments by employers and human capital investments by workers.

Finally, a general result concerns the non-significant impact of the past patent activities on the

present ones. This lack of persistence indicates that Italian firms probably use European

patents only to protect radical innovations that randomly occur in the regional and sectoral

systems of production.

33

Footnotes

1 In Capello’s paper (1999) a very refined proxy for innovative activity and labour mobility

has been used. On the other hand, the limit of this empirical analysis is that it is restricted to

a sample of three Italian High-Tech clusters located only in three different provinces.

2The institutionalist literature did not clearly define the differences between the internal

labour market theory (internal vs. external labour markets) and the dual labour market one

(primary vs. secondary labour markets). Both cases mainly focused on the macro-economic

level, even though the dual labour market view is better fit to analyse labour segmentation

in the primary and secondary sector that occurs within firms and allows de-verticalisation

processes (Guidetti, 1995).

3 The crucial role played by the co-evolution of tangibile (capital, natural resources, etc.) and

intangible (competencies, reputation, etc.) resources within corporations is examined within

the resource-based view and other fields of strategic management theory (Prahalad and

Hamel, 1990; Teece and Pisano, 1994; Teece, 2000).

4 Technical problems, faced by the National Institution of Social Security, in updating and

releasing specific data on the labour market, constrained us to limit our analysis to this

period. Unfortunately more updated, data coming from other sources, are not as suitable and

reliable as NISS data (for a review on statistical sources concerning the Italian labour

market and its flexibility see Contini 2002).

5 A system that categorizes invention by product or process.

6 It must be remarked that the NISS data used in the current analysis concern the firm and not

the single worker as observation unit. This information allows us to considerably simplify

the framework of worker flows. In this way we avoid taking into account the personnel

movements among subsidiaries or plants belonging to the same firm and only consider the

inter-firm mobility.

34

7 In other terms, we used pre-tax wages including basic wage, overtime wage, bonuses,

allowances and subsidies only paid by the employers.

8 More precisely, we redefined only 2 classes, aggregating high and medium-high technology

sectors in hi-tech, and low and medium-low technology sectors in low-tech. It must be

remarked that some two-digit sectors have been aggregated in order to resolve matching

problems between the patents dataset and the labour market’s variables dataset. We

obtained 10 industries from this aggregation process.

Therefore, hi-tech includes: 1)Coke and Refined Petroleum Products, Chemical Products

and Synthetic Fibres, Plastic Products; 2)Machinery, Electrical Equipment, Television,

Office machinery, Medical Components and Mesuring Instruments; 3)Motor Vehicles,

Transport Equipment.

Low-tech embodies: 4) Food, Beverages and Tobacco; 5)Textile Products, the Garment

Industry, the Leather Industry, Luggage, Handbags and Footwear; 6)Paper, Printing and

Publishing; 7)Wood, Furniture and Other Manufacturing; 8)Non-Metallic Mineral Products;

9)Fabricated and Structural Metal Products; 10)Building.

9 For example, a significant proportion of some of the Motor Vehicle industries’ R&D

concerns electronics. Accordingly, the R&D intensity of the Motor Vehicle industry will be

overestimated, while that of electronics will be underestimated.

10 There are some branches of the mechanical and chemical sectors considered as hi-tech, but

included in SMI technological classes. However, the Italian case notably reflects Malerba

and Orsenigo’s claim regarding the fact that SMI technological classes are to be found

especially in the traditional low-tech sectors, whereas most of chemical and electronic

technologies are characterized by the SMII model (Malerba and Orsenigo, 1996, p.463).

Moreover, Pieroni and Pompei (2006) found a high correlation between hi-tech/SMII and

low-tech/SMI in carrying out an analysis concerning the Italian context that was very

similar to this one.

35

11 It is known that valid instruments are 2, −tiy and lagged values of ',tix .

12 Obviously, the sectors are taken from the regional level.

13 In order to save space, the results of static model (4) are not reported. The estimated results,

the full data set and the program carried out with package STATA 8 are available upon

request to the authors.

14 We could not directly control for R&D investments by including them on the right side of

the econometric specification, because of the lack of a suitable breakdown of R&D data

involving both a sectoral and regional profile. For this reason we think that temporal

dummies also capture the influence that R&D investment flows exert on patent activities.

36

Bibliography

Acemoglu, D. 1997a. Technology, unemployment and efficiency, European Economic

Review, vol.41, pp.525-533.

Acemoglu, D. 1997b. Training and innovation in an imperfect labour market, Review of

Economic Studies, vol.64, pp. 445-464.

Acemoglu, D. 2002. Technical change, inequality, and the labour market, Journal of

economic literature, vol.XL, pp. 7-72.

Akerlof, G. and Yellen, J. 1986. Efficiency wage models of the labor market, Cambridge,

C.U.P.

Amin, A. and Wilkinson, F. 1999. Learning, proximity and industrial performance: an

introduction, Cambridge Journal of Economics, vol.23, pp.121-125.

Antonelli, C. 1999. The microdynamics of technological changes, London, Routledge.

Antonelli, C. 2005. Models of knowledge and systems of governance, Journal of Institutional

Economics, vol.1, pp.51-73,

Arellano, M. and Bond, S. 1991. Some test of specification for panel data: Monte Carlo

evidence and an application to employment application, Review of Economic Studies, vol.58,

pp.277-297.

Arnott R. J. and Stiglitz, J. E. 1985. Labor turnover, wage structures and moral hazard,: the

inefficiency of competitive markets, Journal of Labor Economics, vol. 3, no. 4, pp.434-462.

Arnott, R. J., Hosios, H.J. and Stiglitz, J. E. 1988. Implicit contracts, labour mobility and

unemployment, The American Economic Review, Vol.78, No.5, pp. 1046-1066.

Asheim, B.T. and Coenen, L. 2005. Knowledge bases and regional innovation system:

comparing nordic clusters, Research Policy, vol.34, pp.1173-1190.

Atkinson, J. 1985a. Flexibility, uncertainty and manpower management, Institute of

Manpower Studies, report no. 89, Mimeo.

37

Atkinson, J. 1985b. Manpower strategies for flexible organisation, Personnel Management,

August, pp. 28-31.

Atkinson, J. 1986. Changing work patterns. How companies achieve flexibilities to meet new

needs, London, National Economics Development Office.

Baltagi, H.B. 2001. Econometric Panel Data, Wiley, Wiley College. Bassanini, A. And Ernst, E. 2002. Labour market institutions, product market regulation and

innovation: cross-country evidence, OCSE, Economic Department, Working Papers n. 316.

Berman, E. Bound, J. and Griliches Z. 1994. Changes in the demand for skilled labour within

US manufacturing industries, Quarterly Journal of Economics, 109(2), pp.367-398.

Bertola, G. and Rogerson, R. 1997. Institutions and labour reallocation, European Economic

Review, vol.41, pp.1147-1171.

Boeri, T. 1996. Is job turnover countercyclical?, Journal of Labor Economics, vol. 14, no. 4,

pp.603-625.

Boeri, T. 1999. Enforcement of employment security regulations, on the job-search and

unemployment duration, European Economic Review, vol. 43, pp.65-89.

Bound, J. and Johnson, G. 1992. Changes in the structure of wages in the 1980’s: an

evaluation of alternative explanations, American Economic Review, 113(4), pp.1245-79.

Brewster, C., Mayne, L. and Tregaskis O. 1997. Flexible working in Europe, Journal of

World Business, Vol. 32, No. 2, pp. 133-151.

Brusco, S. 1982. The Emilian model: productive decentralisation and social integration,

Cambridge Journal of Economics, Vol. 6, pp. 167-184.

Camagni, R. and Capello, R. 2000. The role of inter-SME networking and links in innovative

high-technology milieux, in Keeble, D. and Wilkinson, F. (eds) High-technology clusters,

networking and collective learning in Europe, Aldershot, Ashgate.

Capello R. 1999. Spatial transfer of knowledge in high technology milieux: learning versus

collective learning processes, Regional Studies, Vol. 33, No. 4, pp. 353-365.

38

Chennells, L. and Van Reenen, J. 1997. Technical change and earnings in British

establishments, Economica, pp.587-604.

Contini, B. 2002. Osservatorio sulla mobilità del lavoro in Italia, Bologna, Il Mulino.

Contini, B. Malpele, C. and Villosio, C., 1996. Wage dynamics and labour mobility in Italy,

Quaderni di Ricerca 96, n.06, Dipartimento di Economia, Università di Torino.

Cooke, P. Gomez Uranga, M. and Etxebarria, G. 1997. rRgional innovation systems:

institutional and organizational dimensions, Research Policy, vol.26, pp. 475-491.

Costabile, L., Papagni E. 1998. Dinamica dell’occupazione, salari e innovazione nell’industria

italiana, Rivista Italiana degli Economisti, No. 3, pp.363-394.

Doeringer, P. and Piore, M.J. 1971. Internal labor markets and manpower analysis, Lexington

Mass., Heat and Company.

Creedy, J. and Whitfield, K. 1988. The economic analysis of internal labour markets, Bulletin

of Economic Research, Vol. 40, No. 4, pp. 247-269.

Faini, R. 1997. Trade unions and regional development, Centro Studi d’Agliano, Working

Papers, n.115,.

Ferrari, S. Guerrieri, P. Malerba, F. Mariotti, S. and Palma, D. 2002. L’Italia nella

competizione tecnologica internazionale. Terzo rapporto, Milano, Franco Angeli.

Flaig, G. and Stadler, M. 1994. Success breeds success: the dynamics of the innovation

process, Empirical Economic, vol. 19, pp.55-68.

Griliches, Z. 1990. Patent statistics as economic indicators: a survey, Journal of Economic

Literature, Vol. XXVIII, pp.1661-1707.

Guidetti, G. 1995. Labour flexibility and segmentation: an analysis of the effects of techno-

organizational constraints, discussion paper University of Aberdeen, No. 95-10, mimeo.

Hatzichronoglou, T. 1997. Revision of the Hi-Technology sector and product classification,

OCDE STI Working Papers, no. 2.

Istat (2001), Statistiche sull’innovazione tecnologica, Roma.

39

Keeble, D. and Wilkinson, F. 1999. Collective learning and knowledge development in the

evolution of regional clusters of high technology SMEs in Europe, Regional Studies, Vol. 33,

No. 4, pp. 295-303.

Keeble, D. and Wilkinson, F. 2000. Regional clustering and collective learning: an overview,

in Keeble, D. and Wilkinson, F. (eds) High-technology clusters, networking and collective

learning in Europe, Aldershot, Ashgate.

Kitson, M. Martin R. and Wilkinson F. 2000. Labour markets, social justice and economic

efficiency, Cambridge Journal of Economics, vol.24, pp.631-641.

Kleinknecht, A. 1998. Is labour market flexibility harmful to innovation?, Cambridge Journal

of Economics, vol. 22, pp.387-396.

Jacobsson, S. and Philipson J. 1996. Sweden’s technological profile. What can R&D and

patents tell and what do they fall to tell us, Technovation, vol.26, pp.245-253.

Jonson G. 1997. Changes in earnings in equality: the role of demand shifts, Journal of

Economics Perspectives, vol.11, pp.41-54.

Lash, S. and Urry, J. 1987. The end of organised capitalism. Cambridge, Polity press.

Lawson, C. and Lorenz, E. 1999. Collective learning, tacit knowledge and regional innovative

capacity, Regional Studies, Vol. 33, No. 4, pp. 305-317.

Longhi, C., and Keeble, D. 2000. High-technology clusters and evolutionary trends in the

1990s, in Keeble, D. and Wilkinson, F. (eds) High-technology clusters, networking and

collective learning in Europe, Aldershot, Ashgate.

Lorenz, E. 1996. Collective learning and the regional labour market, unpublished research

note, European Network on Networks, Collective Learning and RTD in regionally-clustered

high-technology SMEs.

Lundvall B. A. 1993. National Systems of Innovation, London, F. Pinter Publishers.

Malerba, F. and Orsenigo, L. 1996. Schumpeterian patterns of innovation are technology

specific, Research Policy, vol.25, pp. 451-478.

40

Malerba, F and Orsenigo L. 1997. Technological regimes and sectoral pattern of innovative

activities, Industrial and corporate change, 6, I, pp. 83-117,

Malerba, F and Orsenigo, L. 2000. Replay to letter to the editor regarding paper

“Technological entry, exit and survival: an empirical analysis of patent data”, Research

Policy, vol. 29, pp.1187-1188,.

Malerba, F. and Torrisi, S. 2000. La politica pubblica per l’innovazione, in Malerba F. (eds)

Economia dell’innovazione, Roma,Carocci editore.

Michie, J. and Sheehan, M. 2003. Labour market deregulation, “flexibility” and innovation,

Cambridge Journal of Economics, vol. 27, pp.123-143.

Mincer, J. 2003. Technology and the labour market, Review of Economics of the Household,

vol.1, pp.249-27,.

Mohnen, P.A. Nadiri, M.I. and Prucha, I.R. 1999. R&D, production structure and rates of

return in the US, Japanese and German manufacturing sectors, European Economic Review,

vol. 40, pp.381-401.

Mortensen, D.T. and Pissarides, C.A. 1997. Techonological progress Job creation and Job

destruction, Review of Economic Dynamic, vol.1, pp.733-753,

Mortensen, D.T. and Pissarides, C.A. 1999. Unemployment responses to “skill biased

technology shocks: the role of labour market policy, Economic Journal, 109, pp.242-265.

Osterman, P. 1982. Employment structure within firms, British Journal of Industrial

Relations, Vol.20, No.3, pp.349-361.

Paci, R. and Usai, S. 2000. The role of specialisation and diversity externalities in the

agglomeration of innovative activities, Rivista Italiana degli Economisti, no. 2.

Papagni, E. 1995.Sviluppo rurale e progresso tecnico nell’economia italiana, Milano, Franco

Angeli.

Pieroni, L. and Pompei, F. 2006. Localizzazione geografica delle innovazioni e mercato del

lavoro, forthcoming in Rivista di Politica Economica, Vol. Marzo Aprile.

41

Piore, M.J. and Sabel, C. 1984. The second industrial divide: possibilities for prosperity, New

York, Basic Book.

Piore, M.J. 1986. Perspectives on labor market flexibility, Industrial Relations, Vol. 25; No. 2,

pp. 146-166.

Piore, M.J. and Safford, S. 2004. Shifting Axis of Labor Market Regulation and Social

Mobilization in the United States: Reinterpreting the American experience at the turn of the

21st century, Mimeo.

Piva, M., Santarelli, E. and Vivarelli M., 2005. The skill bias effect of technological and

organisational change: evidence and policy implications , Research Policy, Vol.34, pp.65-83.

Piva M., and Vivarelli M., 2005. Innovation and employment: evidence from Italian

Prahalad,C.K. and Hamel G. 1990. The core competence of the corporation, Harvard

Business Review, vol.68, pp.79-91.

Shapiro, C. and Stiglitz, J. E. 1984. Equilibrium unemployment as a worker disciplining

device, American Economic Review, vol.LXXIV, pp.433-44.

Schivardi, F. 1998. Reallocation and learning over the business cycle, Banca d’Italia, Temi di

discussione, no.345.

Stiglitz, J. E., 1974. Alternative theories of wage determination and unemployment in LDC’s:

the labour turnover model, The Quarterly Journal of Economics, vol.88, no.2, pp.194-227.

Stiglitz, J. E. and Greenwald B.C. 1995. Labor-market adjustements and the persistence of

unemployment, American Economic Review, 85, pp. 219-225.

Teece, D.J. and Pisano, G. 1994. The dynamic capabilities of the firm: an introduction,

Industrial and Corporate Change, 3, pp.537-556,

Teece, D.J. 2000. Strategies for managing knowledge Assets: the role of firm

structure and industrial context, Long Range Planning, vol.33, no.1, pp.35-54.

42

Tolliday, S. and Zeitlin J. 1987. The automobile industry and its workers.

Between Fordism and flexibility, New York, St Martin’s Press.

Treu, T. 1992. Labour flexibility in Europe, International Labour Review,

Vol.32, No. 4-5, pp. 497-512.

Vivarelli, M. 1995. The economics of technology and employment: theory and

empirical evidence, Aldershot, Elgar.

Vivarelli, M. and Pianta, M. (eds) 2000. The employment impact of innovation. Evidence and

policy, London, Routledge.

Zimmermann, K.F. 2005. European labour mobility: challenges and potentials, De Economist,

Vol.153, No.4, pp.425-450.

QUADERNI DEL DIPARTIMENTO DI ECONOMIA, FINANZA E STATISTICA

Università degli Studi di Perugia

1 Gennaio 2005

Giuseppe CALZONI Valentina BACCHETTINI

Il concetto di competitività tra approccio classico e teorie evolutive. Caratteristiche e aspetti della sua determinazione

2 Marzo 2005 Fabrizio LUCIANI Marilena MIRONIUC

Ambiental policies in Romania. Tendencies and perspectives

3 Aprile 2005 Mirella DAMIANI Costi di agenzia e diritti di proprietà: una premessa al problema del governo societario

4 Aprile 2005 Mirella DAMIANI Proprietà, accesso e controllo: nuovi sviluppi nella teoria dell’impresa ed implicazioni di corporate governance

5 Aprile 2005 Marcello SIGNORELLI Employment and policies in Europe: a regional perspective

6 Maggio 2005 Cristiano PERUGINI Paolo POLINORI Marcello SIGNORELLI

An empirical analysis of employment and growth dynamics in the italian and polish regions

7 Maggio 2005 Cristiano PERUGINI Marcello SIGNORELLI

Employment differences, convergences and similarities in italian provinces

8 Maggio 2005 Marcello SIGNORELLI Growth and employment: comparative performance, convergences and co-movements

9 Maggio 2005 Flavio ANGELINI Stefano HERZEL

Implied volatilities of caps: a gaussian approach

10 Giugno 2005 Slawomir BUKOWSKI EMU – Fiscal challenges: conclusions for the new EU members

11 Giugno 2005 Luca PIERONI Matteo RICCIARELLI

Modelling dynamic storage function in commodity markets: theory and evidence

12 Giugno 2005 Luca PIERONI Fabrizio POMPEI

Innovations and labour market institutions: an empirical analysis of the Italian case in the middle 90’s

13 Giugno 2005 David ARISTEI Luca PIERONI

Estimating the role of government expenditure in long-run consumption

14 Giugno 2005 Luca PIERONI Fabrizio POMPEI

Investimenti diretti esteri e innovazione in Umbria

15 Giugno 2005 Carlo Andrea BOLLINO Paolo POLINORI

Il valore aggiunto su scala comunale: la Regione Umbria 2001-2003

I

16 Giugno 2005 Carlo Andrea BOLLINO Paolo POLINORI

Gli incentivi agli investimenti: un’analisi dell’efficienza industriale su scala geografica regionale e sub regionale

17 Giugno 2005 Antonella FINIZIA Riccardo MAGNANI Federico PERALI Paolo POLINORI Cristina SALVIONI

Construction and simulation of the general economic equilibrium model Meg-Ismea for the italian economy

18 Agosto 2005 Elżbieta KOMOSA Problems of financing small and medium-sized enterprises. Selected methods of financing innovative ventures

19 Settembre 2005 Barbara MROCZKOWSKA Regional policy of supporting small and medium-sized businesses

20 Ottobre 2005 Luca SCRUCCA Clustering multivariate spatial data based on local measures of spatial autocorrelation

21 Febbraio 2006 Marco BOCCACCIO Crisi del welfare e nuove proposte: il caso dell’unconditional basic income

22 Settembre 2006 Mirko ABBRITTI Andrea BOITANI Mirella DAMIANI

Unemployment, inflation and monetary policy in a dynamic New Keynesian model with hiring costs

23 Settembre 2006 Luca SCRUCCA Subset selection in dimension reduction methods

24 Ottobre 2006 Sławomir I. BUKOWSKI The Maastricht convergence criteria and economic growth in the EMU

25 Ottobre 2006 Jan L. BEDNARCZYK The concept of neutral inflation and its application to the EU economic growth analyses

26 Dicembre 2006 Fabrizio LUCIANI Sinossi dell’approccio teorico alle problematiche ambientali in campo agricolo e naturalistico; il progetto di ricerca nazionale F.I.S.R. – M.I.C.E.N.A.

27 Dicembre 2006 Elvira LUSSANA Mediterraneo: una storia incompleta

28 Marzo 2007 Luca PIERONI Fabrizio POMPEI

Evaluating innovation and labour market relationships: the case of Italy

II

ISSN 1722-618X

I QUADERNI DEL DIPARTIMENTO DI ECONOMIA Università degli Studi di Perugia

1 Dicembre 2002

Luca PIERONI:

Further evidence of dynamic demand systems in three european countries

2 Dicembre 2002 Luca PIERONI Paolo POLINORI:

Il valore economico del paesaggio: un'indagine microeconomica

3 Dicembre 2002 Luca PIERONI Paolo POLINORI:

A note on internal rate of return

4 Marzo 2004 Sara BIAGINI: A new class of strategies and application to utility maximization for unbounded processes

5 Aprile 2004 Cristiano PERUGINI: La dipendenza dell'agricoltura italiana dal sostegno pubblico: un'analisi a livello regionale

6 Maggio 2004 Mirella DAMIANI: Nuova macroeconomia keynesiana e quasi razionalità

7 Maggio 2004 Mauro VISAGGIO: Dimensione e persistenza degli aggiustamenti fiscali in presenza di debito pubblico elevato

8 Maggio 2004 Mauro VISAGGIO: Does the growth stability pact provide an adequate and consistent fiscal rule?

9 Giugno 2004 Elisabetta CROCI ANGELINI Francesco FARINA:

Redistribution and labour market institutions in OECD countries

10 Giugno 2004 Marco BOCCACCIO: Tra regolamentazione settoriale e antitrust: il caso delle telecomunicazioni

11 Giugno 2004 Cristiano PERUGINI Marcello SIGNORELLI:

Labour market performance in central european countries

12 Luglio 2004 Cristiano PERUGINI Marcello SIGNORELLI:

Labour market structure in the italian provinces: a cluster analysis

13 Luglio 2004 Cristiano PERUGINI Marcello SIGNORELLI:

I flussi in entrata nei mercati del lavoro umbri: un’analisi di cluster

14 Ottobre 2004 Cristiano PERUGINI: Una valutazione a livello microeconomico del sostegno pubblico di breve periodo all’agricoltura. Il caso dell’Umbria attraverso i dati RICA-INEA

15 Novembre 2004 Gaetano MARTINO Cristiano PERUGINI

Economic inequality and rural systems: empirical evidence and interpretative attempts

16 Dicembre 2004 Federico PERALI Paolo POLINORI Cristina SALVIONI Nicola TOMMASI Marcella VERONESI

Bilancio ambientale delle imprese agricole italiane: stima dell’inquinamento effettivo

III