1 s2.0 S0014292115001452 Main

European Economic Review 80 (2015) 165–193
Contents lists available at ScienceDirect
European Economic Review

journal homepage: www.elsevier.com/locate/eer
Ownership structure, governance, and innovation

Raoul Minetti a,n, Pierluigi Murro b, Monica Paiella c
a
Department of Economics, 486 W. Circle Drive, 110 Marshall-Adams Hall, Michigan State University, East Lansing, MI 48824-1038, United
States
b
Lumsa University, Italy
c
University of Naples Parthenope, Italy
a r t i c l e i n f o abstract
Article history: This paper tests the impact of firms' ownership structure on innovation in a context
Received 19 April 2014 featuring pronounced ownership concentration and conflicts between large and minority
Accepted 24 September 2015 shareholders. Using data for 20,000 Italian manufacturers, and accounting for the possible
Available online 9 October 2015
endogeneity of ownership levels, we find that ownership concentration negatively affects
JEL classification: innovation, especially by reducing R&D effort. Conflicts between large and minority
G32 shareholders appear to be a determinant of this effect. Moreover, risk aversion induced by
G34 lack of diversification exacerbates large shareholders' reluctance to innovate. Family
O3 owners support innovation more than financial institutions, but the benefits of financial
institutions increase with their equity stakes.
Keywords:
& 2015 Elsevier B.V. All rights reserved.
Ownership
Corporate governance
Agency problems
Technological change
1. Introduction
Technological innovation is a key determinant of firms' performance (OECD, 2010). Innovation allows firms to enhance
their productivity, break into domestic and foreign markets, and retain their leadership as market incumbents (Tellis et al.,
2009). There is a growing consensus among scholars and policymakers that, in turn, firms' ability to advance their tech-
nological frontier is influenced by their governance. Yet, there is little agreement on the way corporate governance exerts
this influence. On the one hand, it is sometimes argued that firms with dispersed ownership (relatively common in the
United States) have more incentives to engage in innovation because they diversify its risk across a large number of
investors (see, e.g., Aghion et al., 2013, for a discussion). Furthermore, in recent years some policymakers have voiced the
concern that firms with ownership concentrated in the hands of families (common in continental Europe and East Asia) may
be reluctant to reallocate resources from their traditional business to risky new technologies (Onida, 2004). On the other
hand, it is also claimed that firms with concentrated, stable ownership can better keep tight control of their activities,
monitor their management, and take long-term views, which is essential for investing in new technologies that need time to
yield results (The Economist, 2012).
These conflicting arguments are often based on anecdotes and case studies while the microeconomic evidence on the
link between ownership and innovation remains scant. As we elaborate below, a broad body of studies document that
ownership structure affects firms' growth and profitability but do not ascertain the role of innovation in this link. The
n
Corresponding author.
E-mail address: minetti@msu.edu (R. Minetti).
http://dx.doi.org/10.1016/j.euroecorev.2015.09.007
0014-2921/& 2015 Elsevier B.V. All rights reserved.
166 R. Minetti et al. / European Economic Review 80 (2015) 165–193
objective of this paper is to shed new light on the impact of ownership on innovation exploiting a rich, large scale survey of
20,000 Italian manufacturing firms conducted by the banking group Capitalia.
Italy constitutes an ideal testing ground for investigating the contrasting effects of ownership concentration, including
possible distortions in firms' decisions due to large shareholders' entrenchment and expropriation of minority shareholders.
The Italian business sector features a strong presence of medium-sized and small businesses, often characterized by
informational opaqueness in financial markets. Italian firms also exhibit pronounced ownership concentration (Bianco,
2003; Bianchi and Bianco, 2008). These features allegedly expose firms to conflicts of interest between main owners and
smaller shareholders, which could be detrimental to complex, long-term investments such as innovation. The structure of
the Italian business sector is similar to that of several European and Asian countries for which scholars and policy makers
voice concerns that excessively concentrated firm ownership may exacerbate conflicts between large and minority share-
holders (Claessens et al., 2002).1 This setting contrasts instead with that of the United States, where the business sector
consists predominantly of public companies with many dispersed shareholders and relatively low ownership concentration.
In U.S. public companies, more concentrated ownership, especially of institutional owners, is often (though not necessarily)
perceived to be a positive development for firms (Shleifer and Vishny, 1997).
Our sample comprises a large number of privately held firms as well as the universe of publicly listed Italian companies.
The data set provides information on firms' innovation which is based directly on firms' responses to survey questions. This
information includes details on the type (product or process) of innovations, as well as on the phases of the innovation
process (invention and adoption of new technologies). The data set also contains thorough information on firms' ownership
structure, such as the equity stakes and the types of the largest shareholders, the alignment between ownership and control,
and the involvement of shareholders in firms' management. Another advantage of our data is the availability of instruments
for firms' ownership structure. A challenge of any study on the relation between ownership and innovation is that unob-
servable factors can affect both. Moreover, reverse causality may be an issue as innovation can shape the ownership
structure. Our strategy for tackling these issues is to employ information on past regulation of Italian local financial markets.
Firms' access to external finance is a critical determinant of their needs and incentives to issue equity and open participation
to new shareholders. We thus employ information on the regulation of local financial markets introduced in Italy in the late
1930s to capture exogenous restrictions on the availability of external finance and construct instruments for firms' own-
ership structure.
As we explain in the paper, two elements make our instruments suitable. First, the regulation introduced in Italy in the
1930s profoundly affected the structure of local financial markets for several decades, and indeed in the early 1990s this
structure closely resembled that of the 1930s. Second, firms' ownership structure exhibits a high degree of persistence: such
a persistence is observed in many countries (Bebchuk and Roe, 1999) and is particularly pronounced in the Italian context
(Bianchi and Bianco, 2008; Istat, 2013). For these reasons, we expect that the regulation introduced in the 1930s affected
firms' ownership structure during the decades in which it was in place and that this effect persisted for several years after
the deregulation at the end of the eighties. On the other hand, the regulation is unlikely to have affected credit supply
conditions for long after its removal. For example, consider a firm seeking credit in 2000. We do not expect that its
probability of obtaining funds or its collateral requirement was significantly affected by a regulation removed more than ten
years earlier. Therefore, our instruments are unlikely to pick any direct effect on innovation of local financial market
conditions.
After accounting for its possible endogeneity, we find that ownership concentration negatively affects the probability
that firms introduce product innovations.2 The effects are sizable: increasing the equity share of the main shareholder by
one standard deviation reduces the likelihood of innovation by 15% (40% of the mean likelihood of innovation). Furthermore,
ownership concentration is neutral for total investment, which signals that its effect on innovation does not merely reflect a
broader effect on the total volume of investment. The estimates also reveal important non-linearities in the effect of
ownership concentration on innovation: its negative effect does not manifest itself at low levels of the equity stake of the
main shareholder, but kicks in especially at higher levels (e.g., above an equity share of 30%). We also obtain that ownership
concentration especially depresses innovation in medium-sized and large firms and in firms operating in traditional
industries. When we break down the innovation process into its phases, we uncover evidence that ownership concentration
depresses R&D more than technology adoption.
The analysis then turns to study the mechanisms through which ownership concentration affects innovation. While in
the United States managerial misbehavior allegedly constitutes the most prevalent form of corporate governance problems,
in Italy, like in other European and Asian countries, firms are often exposed to conflicts between large and minority
shareholders (Bianchi and Bianco, 2008). In fact, Italian firms feature a strong presence of individuals or families with sizable
equity stakes who may extract private benefits or take decisions in their own self-interest that are detrimental to minority
shareholders. Institutional owners are instead less common. Agency theory predicts that conflicts between large and
minority shareholders are severe when ownership is separated from control because controlling shareholders do not fully
internalize the effect of their decisions. The results do indeed reveal that firms in which the main shareholder does not
1
In many Asian and continental European countries, the structure of the business sector is similar to that of Italy, with a significant importance of
privately held small and medium-sized firms and widespread family ownership among privately held firms (see, e.g., OECD, 2014b,c, for a discussion).
2
Ownership concentration does not appear to have a significant effect on process innovations.
R. Minetti et al. / European Economic Review 80 (2015) 165–193 167
retain control are less likely to innovate. This suggests that conflicts among shareholders contribute to the negative effect of
ownership concentration on innovation.3 In addition to identifying a role of agency problems, the estimates suggest that risk
aversion induced by lack of financial or industrial diversification may be a further source of large shareholders' reluctance to
innovate.
The last part of the paper examines whether the type of the main shareholder matters for innovation. In most Italian
firms the main owner is a family or an individual. When the main shareholder is a financial institution, this generally
consists of a private equity fund rather than a bank. Firms in which an individual or a family owns the main equity stake
turn out to be more likely to innovate than firms in which a financial institution is the main shareholder. However, the
benefits of financial institutions for innovation increase with their equity stake. We argue that these results confirm the role
of agency problems in driving innovation.
There is an established empirical literature on ownership structure and corporate performance. Shleifer and Vishny
(1986), McConnell and Servaes (1990) and Zingales (1995) find a positive relation between ownership concentration and
corporate performance in the United States and other developed economies. Claessens (1997) identifies a positive relation
between ownership concentration and stock prices in transition economies (see also Claessens and Djankov, 1999). Other
studies find instead a negative effect of ownership concentration for non-U.S. firms. For instance, Claessens et al. (2002)
uncover a negative effect of large shareholders on the value of East Asian firms.
The evidence on the impact of ownership structure on innovation is scarce and focuses on the United States. Eng and
Shackell (2001) document a positive correlation between institutional ownership and R&D expenditures. Bushee (1998)
finds that managers are less likely to cut R&D to reverse a decline in earnings when institutional ownership is high. Aghion
et al. (2013) confirm this positive link between institutional ownership and the innovation effort of U.S. firms, especially
when CEOs are less protected from takeovers. As noted, since ownership structures exhibit relatively low concentration in
the United States, disentangling the negative effects on innovation of excessively concentrated ownership and large
shareholders' entrenchment can be hard using U.S. data. Our finding that ownership concentration is detrimental to
innovation contrasts with the predictions of the U.S.-based literature. This corroborates the view that the agency problems
that affect widely held public companies in the United States differ from those affecting firms in other countries.
The remainder of the paper is structured as follows. In Section 2, we discuss the theoretical predictions. Section 3
illustrates the data and the econometric methodology. Sections 4 and 5 present the results on the effects of ownership
concentration on innovation. Section 6 presents additional tests on the effects of shareholders' type. Section 7 concludes.
2. Theoretical predictions
To understand its relationship with ownership structure, it is crucial to keep in mind the properties of innovation. First,
new technologies are informationally opaque: they are hard to understand for third parties (e.g., courts) and few interim
signals, such as cash flows, are available on their final outcome (Rajan and Zingales, 2001). Second, innovations entail large
up-front effort and start-up costs (Hall, 2005). Third, innovations require a long gestation period (Goodacre and Tonks,
1995). Fourth, innovations are risky (Schivardi and Schneider, 2008). When a firm innovates, it can make mistakes (Car-
penter and Petersen, 2002). And innovations are risky also because they have low salvage value: the assets are often
intangible (e.g., intellectual property) or specific to the firm (Hall and Khan, 2003).
These properties (informational opaqueness, start-up costs, long-term horizon, riskiness) are relevant for the impact of
ownership structure on agency conflicts in innovative firms. Ownership concentration can mitigate conflicts between
shareholders and managers in widely held firms, such as the U.S. public companies (Shleifer and Vishny, 1997). Consider the
case in which managers are “lazy” or have career concerns. These problems are severe for innovations because they entail
large effort and start-up costs. Moreover, innovations are risky for a CEO: if things go wrong for purely stochastic reasons,
the board of directors may think he is a bad manager and fire him. This makes managers averse to innovation. In these
circumstances, a large shareholder can monitor managers, forcing them to choose innovation optimally (Jensen and
Meckling, 1976). Another problem for innovations is that shareholders tend to undervalue long-term innovative investments
(Stein, 1988) and this makes it easy for hostile acquirers to buy shares of the company at low prices. To protect shareholders,
managers will underinvest in innovations (Manso, 2011). Large shareholders' long-termism can reduce the pressure on
managers for myopic behavior, promoting innovations.
The U.S.-based literature focuses on the above benefits of large shareholders in mitigating conflicts between managers
and shareholders. However, a broad body of literature argues that large shareholders can become entrenched and represent
their own interests at the expense of minority shareholders (Shleifer and Vishny, 1997). Conflicts among shareholders are
allegedly frequent in companies with concentrated ownership in Europe and East Asia (La Porta et al., 1999). Expropriation
of minority shareholders can be easier for informationally opaque new technologies and, hence, minority shareholders can
be hostile to innovations. In addition, large, undiversified shareholders can distort investment in risky innovations because
of risk aversion (Bolton and Von-Thadden, 1998). Risk aversion can manifest itself as fear of losing resources in unsuccessful
3
Large Italian shareholders have an active role in management so that firms are less vulnerable to conflicts between managers and shareholders than
to conflicts among shareholders.
innovative investments. It can also manifest itself as fear of a large shareholder of losing control over the firm, if the
innovative investment does not succeed or even if it succeeds but expands the firm size beyond the level that the share-
holder can retain under his control.
3. Data and empirical strategy
3.1. Institutional background
Italy provides an ideal setting for isolating the link between ownership structure and innovation in a context where
concentrated ownership is predominant and conflicts between large and minority shareholders are frequent. The Italian
business sector features the prevalence of a governance model characterized by high ownership concentration. In 2000
(roughly the midpoint of our sample) the main shareholder owned about 65% of a non-listed manufacturing company on
average (Bianco, 2003; Bianchi and Bianco, 2008). The top shareholder was a family or an individual in 54% of cases, an
Italian company in 27%, a foreign firm in 13%, and a financial holding in 6% of cases. An analysis conducted by Istat (the
Italian National Statistics Office) in 2013 shows that this corporate governance model has changed very little in recent years
and concentrated ownership continues to be predominant in the industrial and services sector (Istat, 2013). The ownership
structure of Italian businesses is also very persistent over time. Bebchuk and Roe (1999) discuss in detail why firms'
ownership structure often exhibits high persistence. And Bianco and Bianchi (2008) demonstrate that indeed in Italy the
persistence of firms' ownership structure, including their ownership concentration, is particularly pronounced.
Turning to innovation, Italy exhibits a relatively low R&D intensity. Business formal R&D spending relative to the GDP
was 0.56% in 2001, compared to 1.62% in the OECD countries (OECD, 2003). Formal R&D spending is carried out more by
medium-sized and large firms in high- and medium-high tech industries. The relatively small size of Italian firms and their
specialization in low- and medium-tech industries may help explain the low R&D intensity of the Italian economy.
3.2. The empirical model
In the first part of the analysis, we study the effect of ownership concentration on innovation. Denote by y the difference
between the returns that a firm expects from a new technology and from an existing technology. The firm's decision to
innovate can be modeled as

1 if y 40
y¼ ð1Þ
0 otherwise
y ¼ xa1 þz1 d11 þ u1 ; ð2Þ

where y is a dummy variable that takes the value of one if the firm innovates, x is a measure of the ownership concentration
of the firm, z1 is a vector of controls, and u1 is the residual.
In testing the effect of ownership concentration on innovation, we must account for the possibility that unobserved
factors are correlated with both. For example, the informational transparency of a firm may affect innovation (Cohen, 1995)
and may also affect ownership structure because it determines the firm's access to equity markets (Myers and Majluf, 1984).
Production efficiency implies a higher return of new technologies and may also attract new investors, affecting ownership
structure. Local market conditions may also be a common determinant of innovation and ownership structure. Moreover,
the endogeneity of ownership structure may stem from reverse causality. A final issue is the attenuation bias that can
originate from errors in the measurement of the equity stake of the shareholders, which is reported by the administrator or
the entrepreneur answering the survey questionnaire.
We address these endogeneity issues using an instrumental variable approach. Let z2 be a vector of instruments that are
correlated with ownership concentration but affect innovation only through ownership. Their effect on x is captured by d22
in the “ownership equation”
x ¼ z1 d21 þ z2 d22 þ u2 ; ð3Þ
where z1 refers to the controls in (2), z2 is the vector of instruments, and u2 is the residual.
We estimate the model in (1)–(3) using two methods, two-stage least squares (2SLS) and Newey (1987)'s Amemya's
generalized least squares (AGLS) for limited dependent variable models. The 2SLS estimation assumes that the prob-
ability of innovation is linear in x and z1. With dicotomous dependent variables, 2SLS work well for values of the
explanatory variables close to sample averages, but suffer from two limitations. First, the predicted values can fall
outside the unit interval of probabilities. Second, the partial effect of any explanatory variable (expressed in levels) is
restricted to be constant. Despite that, 2SLS provide a consistent estimate of the (partial) effect of ownership con-
centration, averaged across the distribution of the other controls.4 In addition to 2SLS, we use AGLS, a minimum chi-
4
We use heteroskedasticity-robust standard errors.
square estimator.5 This is less efficient than maximum likelihood, but is computationally robust and produces con-
sistent estimates and accurate standard errors when the dependent variable is dicotomous and the endogenous
explanatory variable is continuous (Newey, 1987). Furthermore, AGLS perform better than maximum likelihood when
instruments are not strong (Adkins, 2012). OLS and maximum likelihood estimates are also reported.
3.3. Data description
Our main data source is the sample of Italian manufacturing firms surveyed by the banking group Capitalia. The survey
has been used as a testing ground for other objectives by a few studies, including Parisi et al. (2006) and Benfratello et al.
(2008). We use four survey waves, which cover three-year periods ending in 1997, 2000, 2003, and 2006. The data set
includes a representative sample of all firms with 10–500 employees and the universe of firms with more than 500
employees (about 6% of the sample). Approximately 4500 firms were interviewed in each survey wave. Collected data
include information on product and process innovation, R&D and other innovation variables; equity shares and types of the
largest shareholders, as well as other details on the ownership structure; balance sheet data; company characteristics,
including demographics, management and workforce, participation in groups, relationship with customers, suppliers and
financiers. Three-, four- or five-digit industry (ATECO) codes are also reported. Some variables are available for each year
covered by the survey; some refer to the time of interview; others refer to the three-year period covered by the survey.
Table 1 displays summary statistics (see Table A1 for more details on the variables). The majority (68%) of the firms are in
the North of Italy. The average size is small to medium, with a mean of 105 employees and a median of 31. For comparison,
the businesses in the pooled 1998 and 1993 waves of the U.S. National Survey of Small Business Finances have a mean of 30
employees and a median of 6. Thus, the businesses in our sample are slightly larger than those in the NSSBF, although they
are still small- or medium-sized. Firms' legal structure is not reported in the early waves of the survey. When unavailable,
we obtained information on the legal structure from firms' web-sites and hand-matched this information with the surveys
using the VAT identification number. 94% of the firms have limited liability, 61% of which are private limited companies, 39%
public limited companies.
We also employ data of the Bank of Italy on the presence of banks in local markets and data of the Italian National
Statistics Office (Istat) on civil suits and population per judicial district, as well as on the value added and population of
provinces. Finally, we employ the index of financial development proposed by Guiso et al. (2004).
3.4. Measures of innovation and ownership structure
To measure innovation, we use firms' responses to this question: “In the last three years, did the firm introduce product
innovations, process innovations, organizational innovations related to product innovations, organizational innovations
related to process innovations?” . We define two dummies that take the value of one if the firm introduced product or
related innovations (innoprod)and process or related innovations (innoproc). We will later define variables capturing specific
aspects of innovation. In our sample, 39% of the firms report some product or related innovation and 51% report some
process or related innovation over the three years covered by the survey. Table 1 shows that innovators have higher
ownership concentration, are more likely to be in the North, are older and larger than non-innovators.
Employing self-reported data may generate concerns that the firms overstate or understate their innovations. However,
the Italian law (675/1996) prohibits using collected data for objectives different from that stated in the survey (the com-
pilation of statistical tables). Hence, firms should have no incentive to overstate their innovations to build a record as
creditworthy borrowers. Moreover, firms' responses went through multiple checks by highly qualified personnel. Third, a
pure measurement error in the dependent variable would bias the results only if systematically related to the explanatory
variables.
The first key explanatory variable is the ownership concentration of the firm, which we proxy by the equity share held by
the main shareholder (we will also perform some tests using the equity share of the two largest shareholders). In our
sample, the largest shareholder owns 57% of equity on average. There is substantial variation in ownership concentration
across businesses. The share of the main equity holder is 34% at the lowest quartile of firms, 50% at the median, and 85% at
the upper quartile. The survey also asks if the main shareholder has direct control over the firm (87% of cases), if in the years
of survey the firm engaged in acquisitions, divestitures and equity issues, and if the firm intends to go public.
3.5. Instruments
To implement the empirical model we need instruments for ownership concentration. Our strategy relies on identifying
exogenous restrictions on the local financial system that affect ownership concentration but not directly firms' innovation.
To this end, we exploit the 1936 banking law which subjected the Italian banking system to strict regulation of entry until
5
This consists of estimating a reduced form for the endogenous regressor and computing the residual. Then, generalized least squares are applied to a
reduced form for the probit. The explanatory variables of the probit include all the exogenous variables and instruments plus the residual from the first-
step estimation.
170
Table 1
Sample summary statistics.
Full sample Prod. innov. No prod. innov. Proc. innov. No proc. innov. R&D invest. No R&D. inv.
Mean Min Max Median Mean Mean Mean Mean Mean Mean
Dependent variables
Product innovationa 0.388 0 1 0 1.000 0.000 0.511 0.257 0.607 0.225
Process innovationa 0.512 0 1 1 0.679 0.411 1.000 0.000 0.660 0.406
R&D investmenta 0.426 0 1 0 0.668 0.274 0.547 0.299 1.000 0.000
Information technologya 0.764 0 1 1 0.817 0.725 0.804 0.713 0.835 0.703
Patentsa 0.016 0 1 0 0.026 0.009 0.022 0.009 0.027 0.007
Innovation expenditures not for R&D 8.749 0 17.310 9.210 8.802 8.682 8.931 8.516 8.434 9.507
(2.835) (3.006) (2.570) (2.943) (2.662) (3.159) (1.663)
Total investmenta 0.853 0 1 1 0.900 0.826 0.943 0.762 0.911 0.811
R. Minetti et al. / European Economic Review 80 (2015) 165–193

Total investment expenditure (100.000 €) 49,755.73 0.000 537,835.60 103,291.38 11,032.13 77,408.75 12,783.26 100,314.43 13,353.48 80,830.54
(462.245) (5.974) (607.083) (8.611) (710.588) (7.078) (630.739)
Skilled workers hiring 0.776 0 244 0.000 1.422 0.371 1.078 0.409 1.252 0.426
(8.236) (12.916) (2.461) (10.877) (2.494) (5.985) (9.584)
R&D workers (over total workers) 0.054 0 1.000 0.000 0.074 0.039 0.059 0.047 0.077 0.031
(0.128) (0.136) (0.120) (0.123) (0.133) (0.133) (0.119)
Endogenous Variables
Main shareholder quota 0.572 0.01 1.000 0.500 0.598 0.554 0.579 0.565 0.596 0.552
(0.281) (0.286) (0.276) (0.282) (0.279) (0.285) (0.275)
Two main shareholders quota 0.809 0.02 1.000 0.900 0.826 0.797 0.810 0.807 0.820 0.800
(0.232) (0.222) (0.238) (0.231) (0.233) (0.224) (0.238)
a
Main s.hold. is a family/individual 0.749 0 1 1 0.715 0.774 0.723 0.775 0.696 0.793
Main s.hold. is a financial institutiona 0.094 0 1 0 0.126 0.075 0.113 0.075 0.131 0.068
Main s.hold. is a firm or holdinga 0.118 0 1 0 0.119 0.116 0.131 0.104 0.134 0.105
Financial institution subscribera 0.011 0 1 0 0.016 0.008 0.012 0.010 0.016 0.007
Intention to go publica 0.017 0 1 0 0.025 0.012 0.022 0.011 0.026 0.010
Main s.holder has controla 0.870 0 1 1 0.863 0.878 0.874 0.867 0.871 0.873
External managers 0.626 0 1 1 0.636 0.593 0.658 0.592 0.661 0.558
(0.461) (0.454) (0.473) (0.447) (0.473) (0.447) (0.477)
Financial concentration 0.920 0.175 1.000 1.000 0.907 0.936 0.910 0.932 0.906 0.938
(0.178) (0.189) (0.163) (0.186) (0.168) (0.190) (0.160)
Corporationa 0.934 0 1 1 0.946 0.926 0.932 0.936 0.952 0.920
Control variables
Member of a groupa 0.240 0 1 0 0.288 0.203 0.277 0.201 0.310 0.181
‘Traditional’ sectora 0.486 0 1 0 0.434 0.522 0.455 0.519 0.423 0.536
High techa 0.047 0 1 0 0.059 0.039 0.054 0.039 0.067 0.032
Age 24.304 256 21 25.567 23.510 24.803 23.788 25.958 23.074
(17.670) (17.576) (17.663) (18.079) (17.226) (18.320) (17.018)
No. Employees 105.449 11 14,481 31 141.952 73.859 131.223 77.735 152.063 62.030
(354.115) (434.551) (250.186) (401.598) (292.107) (427.347) (242.254)
Total assets (100.000 €) 262.902 0.615 52,424.606 54.107 344.090 169.516 329.743 194.184 364.944 140.180
(132.684) (155.152) (85.368) (153.656) (106.902) (152.625) (81.422)
Sales (100.000 €) 242.500 0.394 56,483.807 55.145 308.773 174.109 283.463 198.377 342.057 141.034
(121.169) (128.338) (105.024) (121.634) (120.666) (144.653) (85.623)
Current assets/Total assets 7.129 5.092 11.927 7.320 7.127 7.136 7.028 7.231 7.138 7.129
(1.645) (1.608) (1.665) (1.606) (1.678) (1.594) (1.681)
Inventories (100.000 €) 57.100 0.000 35,052.486 9.416 78.847 30.903 71.714 42.199 77.717 28.037
(580.562) (727.766) (171.438) (658.057) (489.952) (565.829) (395.064)
No. banks 5.594 0 35 5 6.222 5.204 6.115 5.049 6.446 4.974
(5.024) (4.756) (5.147) (5.883) (3.855) (4.871) (5.056)
Length relation main bank 16.766 0 98 15 16.783 16.771 16.742 16.798 17.095 16.531
(12.223) (12.084) (12.314) (12.392) (12.039) (12.547) (11.986)
a
Credit rationing 0.134 0 1 0 0.131 0.135 0.130 0.137 0.125 0.140
Ateco 5 digitsa 0.309 0 1 0 0.303 0.314 0.300 0.318 0.308 0.310
Ateco 4 digitsa 0.421 0 1 0 0.407 0.430 0.429 0.413 0.397 0.438
Located in the Northa 0.683 0 1 1 0.720 0.660 0.694 0.673 0.724 0.653
Located in the Centera 0.208 0 1 0 0.195 0.216 0.204 0.212 0.204 0.212
Located in the Southa 0.108 0 1 0 0.085 0.124 0.102 0.115 0.072 0.136
No. branches 0.462 0.183 0.938 0.473 0.470 0.457 0.465 0.460 0.472 0.456
(0.112) (0.108) (0.114) (0.111) (0.112) (0.106) (0.116)
Provincial GDP growth 0.085 0.159 0.262 0.073 0.085 0.084 0.085 0.084 0.085 0.084
(0.047) (0.048) (0.047) (0.048) (0.047) (0.047) (0.048)
Provincial Herfindahl 0.066 0.031 0.257 0.060 0.065 0.067 0.066 0.066 0.064 0.067
(0.028) (0.026) (0.029) (0.027) (0.028) (0.025) (0.029)
Local financial development

0.349 0.000 0.519 0.364 0.358 0.345 0.352 0.348 0.360 0.342
(0.113) (0.105) (0.117) (0.111) (0.115) (0.104) (0.119)
Pending trials 0.004 0.002 0.039 0.002 0.003 (0.004 0.004 0.004 0.003 0.004
(0.006) (0.006) (0.006) (0.006) (0.006) (0.005) (0.007)
Instrumental variables
Bank branches in 1936 20.938 3.668 61.777 19.128 21.253 20.750 21.132 20.743 21.388 20.626
(8.537) (8.477) (8.580) (8.671) (8.392) (8.468) (8.592)
Local/Total banks in 1936 0.809 0.261 1.000 0.831 0.821 0.802 0.812 0.806 0.820 0.802
(0.172) (0.167) (0.175) (0.171) (0.173) (0.166) (0.176)
Cooperative banks in 1936 0.698 0.000 2.679 0.586 0.696 0.700 0.688 0.709 0.693 0.702
(0.498) (0.489) (0.505) (0.493) (0.505) (0.492) (0.503)
Savings banks in 1936 0.251 0.000 1.612 0.076 0.257 0.247 0.251 0.251 0.262 0.243
(0.344) (0.353) (0.338) (0.349) (0.3407) (0.351) (0.338)
New branches (by entrants), 91–98 2.422 0.154 8.154 1.538 2.458 2.390 2.433 2.409 2.444 2.397
(2.269) (2.293) (2.244) (2.275) (2.260) (2.276) (2.254)
New branches (by incumbents), 91–98 23.830 0.154 88.615 16.308 24.417 23.341 24.052 23.582 24.386 23.299
(25.013) (25.365) (24.667) (25.086) (24.912) (25.309) (24.663)
Observations 18,603 7035 11,117 9350 8927 7740 10,411
Note: See the Appendix for exact definitions. For the non-dichotomous variables standard deviations are in parentheses.
a
A dummy variable.
171
the end of the eighties. The rationale is that, until its removal at the end of the eighties, this regulation plausibly affected
firms' ability and incentive to attract new shareholders and issue new equity. First, restrictions on the local supply of credit
can prevent investors from obtaining the liquidity necessary to purchase shares of firms (Caselli and Gennaioli, 2013). This
would affect the distribution of equity inside companies. Second, restrictions on the local supply of credit may force a firm to
issue equity. Myers (1984) argues that firms issue equity when they cannot issue the safest security (debt). Finally, the
existence or renewal of a loan can be a positive signal to potential shareholders (Shockley and Thakor, 1992). Thus,
restrictions on the local supply of loans may affect the ability to issue equity.6
We expect that the 1936 regulation affected firms' ownership structure during the decades in which it was in place and
that this effect persisted for several years after the deregulation at the end of the eighties. In fact, the ownership structure is
a highly persistent firm characteristic. Bebchuk and Roe (1999) discuss several reasons for which firms' ownership structure
exhibits strong persistence, including sunk costs, rent seeking and entrenchment of the stakeholders who benefit from the
existing ownership structure. For example, a large, entrenched shareholder may oppose a change in the equity shares that
would reduce his appropriation of resources. Moreover, because of free rider problems, it is hard to convince small
shareholders to sell their shares. As argued by Bebchuk and Roe (1999), these features imply that changes in firms' own-
ership structure could occur only when departures from efficiency are large. In Italy, the persistence of firms' ownership
structure is particularly pronounced (see, e.g., Bianco, 2003; Bianchi and Bianco, 2008). To verify this point in our data, we
exploited the panel structure of a subset of our sample and computed the average change in the equity share of the main
shareholder for the subset of firms that appear in two contiguous waves of the survey. The average change between 1997
and 2000 equaled 2.81%, while the average change between 2000 and 2003 was 2.46%. Thus, the ownership structure
appears to be very persistent.
For the above reasons, we expect that the effect of the 1936 regulation on firms' ownership structure persisted for several
years after the deregulation. On the other hand, the regulation is unlikely to have affected credit supply conditions for long
after its removal. For example, consider a firm seeking credit in 2000. We do not expect that its probability of obtaining
funds or its collateral requirement was significantly affected by a regulation removed more than ten years earlier. Therefore,
our instruments are unlikely to pick any direct effect on innovation of credit market conditions.
Let us now describe the instruments. The objective of the regulation was to enhance bank stability through restrictions
on bank competition. The 1936 banking law imposed strict limits on the ability of different types of credit institutions to
open new branches. Each credit institution was attributed a geographical area of competence based on its presence in 1936
and its ability to grow and lend was restricted to that area. A further directive issued in 1938 regulated differentially the
ability of credit institutions to grow. National banks could open branches only in the main cities; cooperative and local
commercial banks could open branches in the province where they operated in 1936; savings banks could expand within
the boundaries of the region (which comprises multiple provinces) where they operated in 1936.7 Guiso et al. (2003, 2004)
demonstrate that the regulation deeply affected local credit markets (creation and location of new branches) in the fol-
lowing decades. Between 1936 and 1985, in Italy the number of bank branches grew by only 87% versus 1228% in the United
States. By contrast, between the end of the 1980s and the late 1990s, that is, after the deregulation, it grew by about 80%,
almost twice as much as in the United States. To further verify this point, we have examined the degree of correlation
between the credit market structure in 1990 (the start of the deregulation period) and that in 1936. In Fig. 1, we plot the
number of bank branches per 100,000 inhabitants in the Italian provinces in 1990 and in 1936. The figure suggests strong
similarities between the distribution of branches in 1990 and the distribution in 1936. We have also computed the pairwise
coefficient of correlation between the number of bank branches per 100,000 inhabitants in the two years, obtaining a value
of almost 0.64 (significant at the 1% level). Similarly, the rank correlation coefficient equals 0.69. Altogether, these elements
suggest that the credit market structure in 1990 was strongly influenced by the 1936 regulation.
To identify the effects of ownership concentration on innovation, we exploit the differences in the constrictiveness of
regulation across local credit markets, as determined by the structure of the markets in 1936. The variation in the con-
strictiveness can be safely considered exogenous because in 1936 the local credit market structure was independent of local
market characteristics that could affect the ability to do banking and the ability of firms to grow. As discussed by Guiso et al.
(2003, 2004), in 1936 the distribution of types of banks across provinces, and hence the constrictiveness of regulation in a
province, did not reflect market forces but the interaction between previous waves of bank creation and the history of Italian
unification. In addition, the regulation was not designed with the needs of the provinces in mind. In fact, the differences in
the restrictions on the various types of banks were related to differences in banks' connections with the Fascist regime.
In practice, as instruments we use the four indicators that Guiso et al. (2003) employ to characterize the local structure of
the banking system in 1936, and thus the constrictiveness of regulation: (i) the number of bank branches in the province
(per 100,000 inhabitants); (ii) the share of bank branches owned by local banks over total banks in the province; (iii) the
number of savings banks in the province (per 100,000 inhabitants); and, (iv) the number of cooperative banks in the
province (per 100,000 inhabitants). Provinces with more bank branches in 1936, with a higher share of branches of local
banks, with relatively more savings banks and less cooperative banks should have suffered less from the regulatory freeze.
6
Very large firms might be less affected by local credit market conditions. Reestimating the regressions after excluding the largest firms leads to
similar results, as we will see.
7
Provinces are local entities with the size of U.S. counties. They are 103 and are grouped in 20 regions.
Fig. 1. Distribution of bank branches across Italian provinces (branches for 100,000 inhabitants).
Following Herrera and Minetti (2007), besides these indicators, other variables that are likely to reflect the constrictiveness
of the regulation are the (average annual) number of new branches created in a province by incumbent banks and by
entrant banks (net of branches closed) in the years immediately after the deregulation (per 100,000 inhabitants).
One could wonder whether cyclical variations in the economic activity of a province after the deregulation are correlated
both with our instruments and with current innovation. Thus, we control for the growth rate of the value added of the
province (averaged over 1991–98). To further reduce the risk that our historical instruments have an independent effect (i.e.,
other than through ownership) on current innovation, we include firm-level variables capturing current credit conditions
(credit rationing, length and number of relationships with banks) as well as proxies for current local lending conditions (the
number of bank branches in the province, the Herfindhal–Hirschman Index on bank loans and the financial development
index of Guiso et al., 2004). To conclude, a broader argument in defense of our instruments relates to the unclear importance
of province-level characteristics for innovation (see, e.g., Cohen, 1995). All in all, we have reasons to believe that the cor-
relation between our instruments and the residual in the innovation equation is negligible.
3.6. Control variables
We finally discuss the other explanatory variables. If an innovative firm fails, its financiers will recover the firm's assets.
Thus, the larger and the more pledgable the assets, the easier will be for the firm to obtain funds for the innovation. We
include the total assets of the firm and two proxies for asset liquidity, the ratios of current to total assets and of inventories
to total assets. The effects of these ratios are ambiguous ex ante: liquid assets are easy to pledge but easy to expropriate
(Myers and Rajan, 1998). Another determinant of innovation could be the verifiability of output. This depends on the
informational opaqueness of the firm and on the efficiency of courts. Young firms are more informationally opaque than
older ones because they lack an established track record. Hence, we control for the logarithm of age and its square. Small
firms are more informationally opaque than bigger ones because they are not scrutinized by the financial press or by rating
agencies (Petersen and Rajan, 1994). We measure size by total sales (results with the number of employees are similar). To
capture court efficiency, we control for the number of civil suits pending in the judicial district (per 1000 inhabitants). A
high number of pending suits can reflect inefficient courts (Bianco et al., 2005). The propensity to innovate also depends on
the return advantage of innovation. A first factor that may affect this advantage is size (Cohen and Klepper, 1996): bigger
firms can spread the fixed costs of innovation over a larger amount of sales, which we account for. Another factor is
diversification. Diversified firms can exploit economies of scope in innovation. Presumably, the lower the number of
industries in which the firm operates, the less diversified the firm is. We thus construct two dummy variables equal to one
when the firm is classified in a four- or five-digit ATECO sector. An additional factor is age. Older firms are less flexible when
innovating (Hall and Khan, 2003).
Current lending conditions may also affect innovation. We control for the number of relationships with banks, the
duration of the relationship with the main bank, and a dummy equal to one if the firm reports to be credit rationed. Further,
we control for structural characteristics of the banking sector at the time of the survey, such as the number of bank branches
in the province (per 100,000 inhabitants), the Herfindahl–Hirschman Index on bank loans in the province, and the Guiso
et al. (2004) financial development index. Finally, we include provincial GDP growth and two-digit sector dummies.
Industry dummies can capture intra-industry knowledge spillovers (Cohen, 1995). Last, some factors relevant for innovation,
including the quality of infrastructure, the entrepreneurial culture, and the degree of institutional development could differ
across the macro areas of Italy. We code dummies for whether a firm is in the Center or in the South.
4. The role of ownership structure
Before turning to the main results, in Table 2 we report tests of difference of means and medians of the key dependent
variables across segments of firms with different degrees of ownership concentration. For this purpose, we discretize the
equity share of the main shareholder using three thresholds, 25%, 50%, and 75%. We then compare means (Panel A) and
medians (Panel B) of the key dependent variables between firms with the main equity share below and above the defined
threshold. The t-tests suggest that on average firms with lower ownership concentration are less likely to introduce product
innovations, while the evidence for process innovations is less compelling. The table also reports tests of difference of means
Table 2
Tests of difference of means and medians.
Panel A: Tests of difference of means
Ownership concentration (main shareholder quota) Owner type
425% r 25% t-test Family Non-family t-test
Product innovation 0.395 0.344 4.860 0.369 0.444 8.810

Process innovation 0.517 0.495 2.049 0.494 0.563 8.174
R&D investment 0.433 0.381 4.900 0.395 0.522 14.866
450% r 50% t-test Fin. Instit. Non-fin. Inst t-test
Product innovation 0.425 0.356 9.206 0.516 0.374 11.241

Process innovation 0.533 0.498 4.558 0.611 0.501 8.896
R&D investment 0.465 0.392 9.599 0.591 0.409 14.559
475% r 75% t-test
Product innovation 0.444 0.365 9.494

Process innovation 0.531 0.507 2.885
R&D investment 0.483 0.402 9.609
Panel B: Tests of difference of medians
Ownership concentration (main shareholder quota) Owner type
425% (Pearson chi-sq.) p-value Family (Pearson chi-sq.) p-value
Product innovation 22.673 0.000 79.587 0.000

R&D investment 23.320 0.000 223.126 0.000
450% (Pearson chi-sq.) p-value Financial Institution (Pearson chi-sq.) p-value
Product innovation 84.886 0.000 132.229 0.000

R&D investment 92.049 0.000 209.641 0.000
475% (Pearson chi-sq.) p-value
Product innovation 92.170 0.000

R&D investment 93.358 0.000
Note: This table reports tests of differences of means (Panel A) and medians (Panel B) of the main dependent variables. Tests of difference of means are
conducted after discretizing the equity share of the main shareholder using three alternative thresholds, 25%, 50% and 75%.
and medians of the key dependent variables between family and non-family firms. These tests reveal that on average family
firms tend to have lower propensity to innovate than non-family firms.
4.1. Main results
Table 3 reports OLS and probit estimates of the likelihood of product or related innovation (columns 1–3) and process or
related innovation (column 4). Tables 4 and 5 report instrumental variable (IV) estimates. All our results are virtually
identical whether we lump product and process innovations together with related organizational innovations or not.
Henceforth, we focus on the results obtained by including organizational innovations. Let us first consider the OLS estimates
(Panel A). We find that the larger the equity share held by the main shareholder, the greater is the probability of product
innovation. Instead, the coefficient on the main equity share is insignificant in the regression for process innovation.
The OLS and the probit estimates are likely to be biased due to reverse causality and to the omission of variables cor-
related with both innovation and ownership concentration. The direction of this bias is ambiguous a priori. For example, our
controls for firms' informational transparency may be imperfect. Informational transparency can facilitate innovation by
attracting outside investors and can also reduce the need for monitoring by shareholders and, hence, for ownership con-
centration. This would lead to a downward bias in the estimate of the impact of ownership concentration on innovation. The
inability to control thoroughly for production efficiency may also be a source of downward bias. Higher efficiency can
increase the benefits of innovation and can also attract new equity holders, reducing ownership concentration. On the other
hand, reverse causality can generate an upward bias. The main shareholder of an innovative firm could have little incentive
and need to attract shareholders if she expects good profits from a new technology. Instrumental variable estimation allows
us to address these issues. We choose our instruments out of the set of variables reflecting the constrictiveness of the 1936
banking regulation, with the exact choice varying somewhat depending on the specific regression (details are in the notes to
the tables). The exact choice of instruments is based on their ability to predict ownership concentration in the first-stage
regressions, conditioning on the exogenous variables included. Tables 4 and 5 report the results of IV estimation. Table 4
displays the first-stage coefficients on the excluded instruments (the other coefficients are available upon request). The main
Table 3
Ownership concentration and innovation. OLS and Probit regressions.
Panel A: OLS
(1) (2) (3) (4)

I.PROD I.PROD-N. banks I.PROD-Cr. rat. I.PROC
Main s.holder 0.071nnn 0.073nnn 0.081nnn 0.021

Quota (0.016) (0.015) (0.015) (0.016)
Time dummies Y Y Y Y
Area dummies Y Y Y Y
þcontrols Y Y Y Y
Obs. 12113 13257 12893 12130
R2 0.08 0.08 0.08 0.06
Panel B: Probit
(1) (2) (3) (4)

I.PROD I.PROD-N. banks I.PROD-Cr. rat. I.PROC
Main s.holder 0.185nnn 0.198nnn 0.212nnn 0.041

Quota (0.043) (0.041) (0.042) (0.043)
þcontrols Y Y Y Y
Obs. 12,113 13,257 12,893 12,130
R2 0.06 0.07 0.06 0.05
Note: The table reports regression coefficients and associated standard errors (in parentheses). The dependent variables are reported at the top of each
column. Main shareholder quota refers to the capital share held by the main shareholder. Area dummies refer to the area of the country where the firm is
headquartered (Center or South). The controls included are (a) firm characteristics, namely total assets, current assets, inventories, sales, a second-order
polynomial in the age of the firm since founding, ATECO four- and five-digit code dummies, dummies for the sector of activity and duration of the credit
relationship with the main bank; (b) structural characteristics of the banking sector, namely the number of bank branches (per 100,000 inhabitants) and
the Herfindahl–Hirschman Index on bank loans, in the province; (c) variables controlling for the characteristics of the environment where the firms
operates, namely Guiso et al. (2003) financial development index, provincial GDP growth and a measure of the efficiency of the court system; (d) 24 sector
dummies. For exact definitions and details see the Data Appendix. In column (2) instead of the duration of the credit relationship with the main bank, we
use, as a control, the number of banking relationships. In column (3) instead of the duration of the credit relationship with the main bank, we use, as a
control, a dummy variable equal to one if the firm is credit rationed. The table reports the R2 (OLS) and pseudo R2 (Probit).
nnn
Significant at 1%.
Table 4
Ownership concentration and innovation. First stage of IV regressions.
(1) (2) (3) (4)

Product Prod – N.banks Prod – Cr. rat. Process
Main Shareholder quota
Branches in 1936 0.001nn 0.001nn 0.001nn 0.001nn

(0.0004) (0.0004) (0.0004) (0.0004)
Local banks/Tot. banks 0.056nn 0.053nn 0.054nnn 0.057nn
(0.022) (0.021) (0.021) (0.022)
Savings banks 0.008 0.005 0.004 0.007
(0.011) (0.010) (0.010) (0.011)
New branches 0.005nnn 0.005nnn 0.005nnn 0.005nnn
(by entrants) (0.002) (0.002) (0.002) (0.002)
þ controls Y Y Y Y
F statistics (Instr.) 5.78 5.98 5.58 5.83
Observations 12,113 13,257 12,893 12,130
Note: The table reports regression coefficients and associated standard errors (in parentheses). The dependent variables of the second stage are reported at
the top of each column. The set of instruments includes number of bank branches in the province in 1936 (per 100,000 inhab.), the ratio of local to total
bank branches, number of savings banks in the province in 1936 (per 100,000 inhab.), branches opened by new entrants in the province over the 1991–
1998 period (net of branches closed). Main shareholder quota refers to the capital share held by the main shareholder. Time dummies denote the year of
the survey. Area dummies refer to the area of the country where the firm is headquartered (Center or South). “þ controls” denotes the RHS variables of the
regressions in Table 2. In column (2) instead of the duration of the credit relationship with the main bank, we use, as a control, the number of banking
relationships. In column (3) instead of the duration of the credit relationship with the main bank, we use, as a control, a dummy variable equal to one if the
firm is credit rationed. The table reports the value of the F-statistics for a test of the weakness of the instruments.
nn
Significant at 5%.
nnn
Significant at 1%.
equity share is increasing in the number of bank branches in the province in 1936 and in the number of branches created by
new entrants over the 1991–1998 period, supporting the hypothesis that lower restrictions on the local supply of credit
reduce the need to sell equity. Instead, it is decreasing in the relative importance of local banks and in the diffusion of
savings banks (but the latter coefficient is not significant). The instruments are jointly highly significant (p-value ¼0.0000);
first-stage F-statistics are between 5 and 6.8 The p-values of the overidentification tests, reported in Table 5, show that,
except for the regressions for process innovation, we cannot reject the null hypothesis that the instruments are uncorrelated
with the regression residual. Finally, in Table 5 we also report the p-value for the Wald test of exogeneity of the main equity
share in the probit.9 We reject the null hypothesis that the equity share is exogenous with respect to the propensity to carry
out product innovation, but we generally do not reject this hypothesis for process innovation.
The second stage of the IV estimation is in Table 5. We comment the AGLS estimates; the 2SLS estimates are qualitatively
similar. Once we account for the endogeneity problem, the impact of ownership concentration on the likelihood of product
innovation is negative (z-statistic of 2.54) and large. The coefficient implies that increasing the main equity share by one
standard deviation would reduce the likelihood of product innovation by over 15% (almost 40% of the mean). The negative
effect of ownership concentration contrasts with the predictions of the literature for U.S. public companies, suggesting that
agency conflicts or risk aversion can hinder innovation in Italian firms with concentrated ownership. We will later study the
channels through which the effect of ownership concentration on innovation operates. The impact of ownership con-
centration on process innovation is negative too, but insignificant. One interpretation is that in Italian businesses process
innovations often consist of marginal advances for increasing the flexibility of production (Onida, 2004; Colombo and
Lanzavecchia, 1996). Such minor advances may be loosely affected by agency conflicts and, hence, be weakly related to the
ownership structure.
Interestingly, the estimated coefficients should be considered an upper bound to the true estimates of the effects: in fact,
because the instruments are not strong, the estimates are somewhat biased towards the OLS ones (Stock et al., 2002).
Finally, results are robust to using different sets of instruments, such as adding the squares of the variables used as
instruments and using other subsets of the variables reflecting the constrictiveness of the 1936 regulation.10 The reader may
8
An F-statistic between 5 and 6 signals that we could have a weak instruments problem and our estimates could be biased toward their OLS
counterparts. However, based on Stock and Yogo (2005)'s tabulation of the critical values for the weak instrument test, we reject the null of a relative bias
greater than 20 percent.
9
In the AGLS probit estimator, in the second stage we include the residual from the first-stage OLS as a regressor. The Wald test is a test of significance
on the coefficient of this residual.
10
In Italy in the period covered by the survey some reforms were enacted with possible consequences for corporate governance (see, e.g., Cumming
and Zambelli, 2013, 2010; see also Bianchi and Bianco, 2008, for a review). These reforms were implemented homogeneously in the country, so we cannot
Table 5
Ownership concentration and innovation. Second stage of IV regressions.
(1a) (1b) (2a) (2b) (3a) (3b) (4a) (4b) (5a) (5b)
2SLS AGLS 2SLS AGLS 2SLS AGLS 2SLS AGLS 2SLS AGLS
I.PROD I.PROD I.PROD I.PROD I.PROD I.PROD I.PROC I.PROC I.PROD I.PROD
Endogenous variable
Main shareholder 1.076nn 2.852nn 0.992nn 2.643nn 1.132nnn 3.029nnn 0.158 0.328
quota
(0.422) (1.148) (0.404) (1.102) (0.437) (1.186) (0.362) (0.946)
Two main shareholders 0.918n 2.315
quota
(0.541) (1.479)
Firms' characteristics
Center 0.029 0.075 0.033 0.084 0.032 0.084 0.009 0.027 0.017 0.040
(0.022) (0.060) (0.022) (0.059) (0.023) (0.062) (0.019) (0.049) (0.022) (0.061)
South 0.041 0.114 0.043 0.115 0.039 0.107 0.015 0.042 0.025 0.067
(0.035) (0.097) (0.034) (0.094) (0.035) (0.098) (0.030) (0.080) (0.033) (0.092)
Tot. assets 0.038n 0.073n 0.028n 0.060n 0.046nnn 0.100nnn 0.036nnn 0.146nnn 0.021n 0.025
(0.019) (0.040) (0.016) (0.034) (0.021) (0.038) (0.011) (0.046) (0.012) (0.032)
Current assets/Tot. 1.608nnn 4.635nnn 1.335nnn 3.810nnn 1.545nnn 4.455nnn 1.930nnn 4.983nnn 1.093nnn 3.249nnn
Assets
(0.441) (1.217) (0.417) (1.147) (0.458) (1.248) (0.386) (1.005) (0.343) (0.953)
Inventories 0.031 0.542nnn 0.034 0.321nn 0.044nn 0.487nnn 0.027 0.345nn 0.015 0.656nnn
(0.030) (0.157) (0.025) (0.144) (0.032) (0.152) (0.017) (0.161) (0.020) (0.158)
Sales 0.020 0.013 0.020nn 0.025 0.026 0.024 0.004 0.052nn 0.013 0.012
(0.015) (0.029) (0.013) (0.028) (0.018) (0.032) (0.010) (0.025) (0.010) (0.025)
Age 0.003nnn 0.009nnn 0.001 0.003 0.001 0.004 0.002nnn 0.004nn 0.003nnn 0.010nnn
(0.001) (0.002) (0.001) (0.003) (0.001) (0.003) (0.001) (0.002) (0.001) (0.003)
Age squared 0.000nnn 0.000nnn 0.000 0.000 0.000 0.000n 0.000n 0.000 0.000nnn 0.000nnn
(0.000) (0.000) (0.000) (0.000) (0.000) (0.000) (0.000) (0.000) (0.000) (0.000)
ATECO 5 digits 0.034nn 0.094n 0.025 0.070 0.038nn 0.103nn 0.019 0.047 0.017 0.047
(0.017) (0.048) (0.016) (0.045) (0.017) (0.048) (0.015) (0.040) (0.014) (0.039)
ATECO 4 digits 0.005 0.011 0.006 0.017 0.002 0.003 0.009 0.019 0.002 0.008
(0.014) (0.040) (0.013) (0.037) (0.014) (0.039) (0.013) (0.033) (0.013) (0.036)
External fin. variables
Length rel. main bank 0.063nnn 0.173nnn 0.018 0.044 0.051nnn 0.139nnn
(0.015) (0.041) (0.013) (0.033) (0.014) (0.038)
No. banks 0.015nnn 0.038nnn
(0.001) (0.004)
Credit rationing 0.064nnn 0.175nnn
(0.015) (0.043)
Local market conditions
No. branch/100,000 0.142n 0.378n 0.100 0.278 0.154nn 0.415nn 0.138nn 0.337nn 0.119n 0.314n
Inhabitants (0.072) (0.196) (0.067) (0.181) (0.072) (0.194) (0.062) (0.162) (0.067) (0.180)
Prov. GDP growth 0.001 0.006 0.059 0.183 0.078 0.210 0.143 0.378 0.001 0.001
(0.112) (0.307) (0.103) (0.287) (0.109) (0.301) (0.097) (0.254) (0.104) (0.289)
Prov. HHI 0.324 0.875 0.290 0.789 0.428 1.179 0.102 0.300 0.375 0.939
(0.276) (0.768) (0.265) (0.743) (0.279) (0.783) (0.241) (0.626) (0.341) (0.939)
Financial development 0.031 0.085 0.064 0.175 0.022 0.059 0.103 0.266 0.021 0.051
(0.085) (0.234) (0.082) (0.228) (0.086) (0.239) (0.073) (0.193) (0.083) (0.231)
Pending trials/100,000 1.541 4.924 1.663 5.158 1.429 4.426 0.356 0.958 1.118 3.664
Inhabitants (1.108) (3.350) (1.051) (3.172) (1.109) (3.336) (0.998) (2.701) (1.043) (3.133)
Time dummies Y Y Y Y Y Y Y Y Y Y
Overid. test (p-value) 0.9774 0.9730 0.9983 0.0023 0.4251

Wald test (p-value) 0.0017 0.0026 0.0009 0.6965 0.0632
Observations 12,113 12,113 13,257 13,257 12,893 12,893 12,130 12,130 12,064 12,064
Note: The table reports regression coefficients and associated standard errors (in parentheses). The dependent variables and the estimation method are
reported at the top of each column. The set of instruments includes number of bank branches in the province in 1936 (per 100,000 inhab.), the ratio of local
to total bank branches, number of savings banks in the province in 1936 (per 100,000 inhab.), branches opened by new entrants in the province over the
1991–1998 period (net of branches closed). Main shareholder quota refers to the capital share held by the main shareholder. Two main shareholders quota
refers to the capital share held by the two main shareholders. Time dummies denote the year of the survey. In columns (2a) and (2b), instead of the
duration of the credit relationship with the main bank, we use, as a control, the number of banking relationships. In columns (3a) and (3b), instead of the
duration of the credit relationship with the main bank, we use, as a control, a dummy variable equal to one if the firm is credit rationed. The table reports
the p-values of a Sargan test of overidentifying restrictions and of a Wald test of exogeneity of the variable that has been instrumented.
n
Significant at 10%.
nn
Significant at 5%.
nnn
Significant at 1%.
also wonder how the results change if we replace the equity share of the main shareholder with that of the two largest
shareholders. In some firms the second shareholder can form a block with the main one (Zwiebel, 1995), while in others he
can help monitor and discipline the main shareholder (Pagano and Röell, 1998). The estimated coefficient of the equity share
of the two largest shareholders retains a negative sign, but its statistical and economic significance is lower than that of the
main equity share (columns 5a and 5b of Table 5). This suggests that while in some firms the interests of the two largest
shareholders are aligned, in others the second shareholder dilutes the negative effect on innovation of the (equity share of
the) main shareholder.
The results for the firm-specific controls are consistent with those in the literature. The value of assets and the ratio of
inventories to total assets have a positive impact, whereas the ratio of current to total assets has a negative effect. The
coefficient of the sales of the firm is insignificant for product innovation, while it is negative and significant for process
innovation. The polynomial in age is significant and suggests a convex relationship. Regarding the characteristics of the local
environment, bank branch density has a positive effect while the coefficients on the Herfindhal on bank loans and on the
financial development index are generally not significant. Finally, the likelihood of innovation is decreasing in the duration
of the main credit relationship (see Minetti, 2011, for a theoretical explanation). As a robustness check, we use the number
of banking relationships as a control (columns 2a and 2b) and obtain that the likelihood of innovation is increasing in this
number, which is consistent with the view that multiple banks mitigate lenders' moral hazard (Rajan, 1992). Perhaps
surprisingly, the coefficient for credit rationing is positive (columns 3a and 3b).
4.2. Non-linear effects
Agency theory suggests that the effect of ownership concentration on firm value may be non-linear (Stultz, 1988). In
Panel A of Table 6, we allow for non-linearities in its effect on innovation. Fixed monitoring costs could imply that a large
shareholder monitors only if his equity share exceeds some threshold. However, a large shareholder could be more
entrenched and acquire the power needed to expropriate revenues especially when his equity stake is sufficiently large. We
first experiment by adding a quadratic term in the main equity share. The instruments are jointly highly significant and we
do not reject the overidentifying restrictions.11 In the 2SLS regressions, the likelihood of product innovation appears to be
convex in ownership concentration. Ownership concentration now becomes significant also in the regression for process
innovation.
The literature stresses that non-linearities in the effect of ownership concentration can take quite complex forms, which
would imply that simply inserting a quadratic term does not fully capture alignment and entrenchment effects. For example,
investigating the effect of managerial equity shares on firms' value, Morck et al. (1988) obtain that for relatively small equity
shares the effect of ownership concentration is positive, for an intermediate range the effect becomes negative, and finally
for even higher equity shares the effect turns slightly positive again. In our context, one could suspect that as long as
ownership concentration remains relatively low, a higher equity stake of the main shareholder does not carry extra costs in
terms of stronger entrenchment of the shareholder. In Table 6, Panel A (columns 6–11), we then experiment using
thresholds of the main equity share in line with those identified by Morck et al. (1988), accounting for the relatively higher
concentration of ownership in Italy than in the United States.12 We obtain that the negative effect of ownership con-
centration on innovation manifests itself at sufficiently high levels of the main equity share (above 30% or 45%) but not at
lower levels. This result carries through if we estimate the piecewise linear regression suggested by Morck et al. (1988) (see
column 5).13 As a robustness exercise, we also reran the regressions using the quartiles of the distribution of the main equity
share to construct thresholds. The inference we draw is virtually the same, with the negative effect of ownership con-
centration manifesting itself at sufficiently high levels of the main equity share (from the third quartile).
The reader could also wonder whether there are significant differences among firms of different size in the sample, e.g.,
between small firms with relatively few employees and large firms.14 In columns 1–4 of Table 6, Panels B and C, we report
the regressions for product innovation after partitioning the sample based on the number of employees. The impact of
ownership concentration is significant only for firms with at least 34 employees (the median number, see column 2), and
(footnote continued)
build on them to construct instruments. In untabulated tests, we experimented with using the 2003 corporate law reform, on the ground that this could
have had some differential effect on a subgroup of limited liability companies (SRL, Societa' a Responsabilita' Limitata). We constructed an instrument
consisting of the interaction between a dummy for SRL and a dummy taking the value of one after 2003, zero before. This instrument performed worse
than our preferred set of instruments in the first stage. In the second stage, the coefficient on ownership concentration retained a negative sign, although it
was estimated imprecisely.
11
Following Wooldridge (2002), to the instruments used so far, we add the square of the balance sheet variables included as controls in the regressions
for innovation (first-stage regressions available upon request).
12
In columns 6–8, we use 15% and 30% as thresholds. In columns 9–12, we consider 15%, 45%, and 85%. Morck et al. (1988) use 5%, 15% and 30% as
thresholds for managerial equity shares.
13
In the piecewise linear regression, the variable “main s.holder quota 0–15%” equals the main equity share when this is no greater than 15%, and
equals 15% when the share exceeds 15%. The variable “main s.holder quota 15–30%” equals zero when the share is no greater than 15%, equals the share
when this is between 15% and 30%, and equals 30% when the share exceeds 30%. Finally, the variable “main s.holder quota 4 30%” equals zero when the
share is no greater than 30% and equals the share when this exceeds 30%.
14
Ownership in larger firms is also more likely to be associated with a financial institution. In fact, in our data the share of firms above the median
number of employees that have a financial institution as the main shareholder equals 15.2% (versus 9.4% in the whole sample).
Table 6
Non-linear effects of ownership concentration on innovation.
Panel A: Ownership levels
(1) (2) (3) (4) (5) (6) (7) (8) (9) (10) (11) (12)
2SLS AGLS 2SLS AGLS 2SLS 2SLS 2SLS 2SLS 2SLS 2SLS 2SLS 2SLS
I.PROD I.PROD I.PROC I.PROC I.PROD I.PROD I.PROD I.PROD I.PROD I.PROD I.PROD I.PROD
Quota o 15% 15% o Quota o 30% Quota 430% Quota o 15% 15% o Quota o 45% 45% o Quota o 85% Quota 4 85%
Main s.holder quota 9.208n 24.426nn 13.110nn 36.358nn 7.738 5.488 0.797n 7.738 2.095 2.963n 2.238
(4.755) (11.679) (6.076) (15.342) (15.274) (6.926) (0.420) (15.274) (2.393) (1.823) (5.032)
(Main s.holder 7.110nn 11.829nn 10.100nn 28.185nn
quota)2

(3.435) (8.459) (4.392) (11.114)
Main s.holder quota 5.913
0–15% (23.451)
Main s.holder quota 4.218
15–30% (7.309)
Main s.holder quota 1.095nn
over 30% (0.539)
Time and Area Y Y Y Y Y Y Y Y Y Y Y Y

dummies
þ controls Y Y Y Y Y Y Y Y Y Y Y Y
Overid. test (p- 0.0046 0.0756 0.7439 0.9353 0.3096 0.0000 0.5979 0.3004 0.3640 0.8107
value)
Wald test exog. (p- 0.0004 0.0000
value)
Observations 12,113 12,113 12,130 12,130 12,113 393 1530 10,192 393 3973 4721 3028
Panel B: 2SLS
Size Industry
(1) (2) (3) (4) (5) (6) (7) (8) (9) (10)
I.PROD I.PROD I.PROD I.PROD I.PROD I.PROD I.PROD I.PROD I.PROD I.PROD
Empl: o 34 Empl:Z 34 Empl: o 15 Empl: Z 15 Drop bottom 5% Drop bot- Traditional High Tech Scale No scale
& top 5% tom 5%
Main s.holder 0.566 1.152n 0.072 0.888nn 0.731nn 0.650nn 2.065nn 0.119 1.054 1.446nn
quota
(0.533) (0.622) (0.686) (0.378) (0.339) (0.321) (0.987) (0.614) (0.694) (0.601)
Time and area Y Y Y Y Y Y Y Y Y Y

dummies
þ controls Y Y Y Y Y Y Y Y Y Y
Overid. test (p- 0.9400 0.6462 0.609 0.937 0.755 0.943 0.9273 0.8589 0.3910 0.5647
value)
Observations 5943 6170 1343 10,770 10,947 11,605 5998 520 2468 9645
179
180
Table 6 (continued )
Panel B: 2SLS
Size Industry
(1) (2) (3) (4) (5) (6) (7) (8) (9) (10)
Empl: o 34 Empl: Z 34 Empl: o 15 Empl: Z 15 Drop bottom 5% Drop bot- Traditional High Tech Scale No scale
& top 5% tom 5%
Panel C: AGLS
Size Industry
(1) (2) (3) (4) (5) (6) (7) (8) (9) (10)

Empl: o 34 Empl: Z 34 Empl: o 15 Empl: Z 15 Drop bottom 5% Drop bot- Traditional High Tech Scale No scale
& top 5% tom 5%
Main s.holder 1.636 2.887n 0.129 2.362nn 1.873nn 1.644n 6.021nn 0.893 2.709 3.874nn
quota
(1.594) (1.646) (2.099) (1.034) (0.922) (0.862) (2.834) (1.898) (1.838) (1.644)
Time and Area Y Y Y Y Y Y Y Y Y Y

dummies
þcontrols Y Y Y Y Y Y Y Y Y Y
Wald test exog. (p- 0.2308 0.0303 0.983 0.005 0.0164 0.0201 0.0004 0.5898 0.0855 0.0011
value)
Observations 5943 6169 1334 10,770 10,947 11,605 5995 520 2466 9643
Note: The table reports regression coefficients and associated standard errors (in parentheses). The dependent variables are reported at the top of each column. In Panel A we show the regression for non-linear
effects of ownership (in this panel the estimation method is reported at the top of each column). Panels B and C display regressions that allow the effect of ownership concentration to differ across firms of different
size and across firms operating in different industries. The regressions in Panel B are estimated by two-stage least squares (2SLS) and the regressions in Panel C are estimated by two-stage conditional maximum
likelihood (AGLS, Newey's minimum chi-squared estimator). We instrument both the main shareholder quota and its square. The set of excluded instruments includes the ratio of local bank branches to total bank
branches, savings banks (per 100,000 inhab.) in the province in 1936, the branches opened by new entrants in the province over the 1991–1998 period (net of branches closed), and the number of bank branches
(per 100,000 inhab.). For the regressions in columns (1)–(4) of Panel A, we have also used total assets, the ratio of current assets to total assets, inventories and sales all squared. For the regressions in columns (7)–
(10) of Panels B and C we have also used the number of cooperative banks (per 100,000 inhabitants) in the province in 1936. Time dummies Denote the year of the survey. Area dummies refer to the area of the
country where the firm is headquartered (Center or South). “þ controls” denotes the RHS variables of the regressions in Table 3. The table also reports the p-values of a Sargan test, as a test of overidentifying
restrictions, and of a Wald test, as a test of exogeneity of the variables that have been instrumented.
n
nn
Significant at 5%.
this result remains unchanged if we split the sample at the median value of sales. An interpretation is that in very small
firms minority shareholders are probably close to the main one (perhaps part of the extended family or network) and
conflicts among shareholders that can inhibit innovation are less important. To dig deeper into the effect of firm size on the
relation between ownership concentration and innovation, in the regressions we experiment with other size thresholds. It is
sometimes argued that the decisions of Italian firms are affected by labor market regulations, which offer stronger
employment protection to workers with permanent contracts in firms with at least 15 employees. We then investigate
whether the results change when we drop firms with less than 15 employees from the sample (see column 4). The results
remain virtually unchanged. Finally, to make sure that the results are not overly influenced by the smallest and the largest
firms, we also experiment with dropping the smallest 5% and the largest 5% firms for number of employees (columns 5 and
6). Again, the results carry through.15
When we distinguish across industries (columns 7 and 8), ownership concentration has a negative impact on product
innovation for firms in traditional sectors (such as textiles, food, and tobacco), while its impact is insignificant in high-tech
sectors. This could be due to the fact that for high-tech firms innovating is really a condition for survival, so the ownership
structure has a limited impact. These results carry through if we adopt the classification of high-tech sectors put forth by
Benfratello et al. (2008). Ownership concentration has also a negative impact in sectors where economies of scale are
modest, which include many traditional firms, whereas its coefficient is insignificant in sectors where economies of scale are
important (columns 9–12).
Finally, in untabulated tests, we also allowed the effect of ownership concentration on innovation to vary depending on
whether the firm is located in the North, Center or South. The results carried through and the estimated coefficients on the
interaction terms of the main equity share with the geographical dummies were not significant.
4.3. Changes of ownership structure
Not only the current ownership structure can affect a firm's innovation, but also its changes can have a role in innovation
decisions. Some survey questions help capture these “dynamic” effects. The survey asks whether financial institutions
subscribed new shares over the three years ending in the year of the survey. In Table 7, columns 1–4, the key explanatory
variable is a dummy equal to one if a financial institution subscribed shares. We instrument this dummy using the same
variables used for ownership concentration. In the linear probability model, we use two-stage least squares and instrument
the dummy by its fitted probability.16 In the probit, we employ a two-step method based on a least square approximation
proposed by Arendt and Holm (2006). This consists of estimating a linear probability model for the dummy, computing the
residual and then estimating a probit for innovation adding the fitted residual as a covariate (in the table, we label this
procedure as “augmented model” ).17 The dummy has a significant and positive effect on product and process innovation.18
The second survey question regards the intention to go public in the following year. We code a dummy equal to one if the
firm has this intention and either instrument it with its fitted probability in the linear probability model or estimate a
“residual augmented” probit, as detailed above. The coefficient of the dummy is positive and significant for both product and
process innovation (columns 5–8 of Table 7). The third survey question regards the intention of the firm to sell minority
equity stakes to financial institutions in the following three years. Again, the dummy capturing this intention turns out to
have a positive and significant coefficient (columns 9–10).
Overall, the results in Table 7 that issuing equity and going public tend to promote innovation are consistent with our
baseline results. In fact, issuing equity and going public dilute the cash flow rights of the main shareholder and reduce
ownership concentration (which we found to depress innovation).19
4.4. Phases of the innovation process
The innovation process comprises two phases: research and adoption of new technologies. In Table 8, we study at what
stage ownership structure is more relevant. To ascertain the effect on research, we construct a dummy that takes the value
of one if the firm carried out R&D expenditures in the three years covered by the survey. The effect of the main equity share
on the probability of R&D is negative and significant (column 1).
To examine the effect on technology adoption, we code a dummy equal to one if the firm answered affirmatively to this
question: “In the last three years did the firm carry out investment for the introduction of hardware, software, tele-
communication networks?” . Investment in information technology can proxy for the adoption of innovations because most
15
If we segregate the data by quartiles of employees, consistent with the previous results, we obtain that it is especially the medium-large firms that
appear to drive the negative effect of ownership concentration on product innovation. The results are available from the authors.
16
Angrist and Pischke (2009) suggest to instrument an endogenous dichotomous regressor with the fitted probability from a first stage probit reduced
form model for the endogenous dummy. However, this procedure presents difficulties when there is more than one endogenous variable and the
endogenous variables share the same instruments, which is our case when we add to the regression the main equity share.
17
This procedure is computationally less demanding than maximum likelihood and always converges. Arendt and Holm (2006) conduct Monte Carlo
exercises and find that it works well and overall performs as well as full maximum likelihood in small samples when the endogeneity is not too severe.
18
When we control for ownership concentration in the non-linear probability model, we estimate the coefficients of interest by running AGLS on the
“residual augmented” model.
19
In untabulated regressions we found that acquisitions and divestitures have no effect on innovation.
182
Table 7
Changes to ownership structure and innovation.
Panel A: Product innovation
(1) (2) (3) (4) (5) (6) (7) (8) (9) (10) (11) (12)
2SLS “Augmented” 2SLS “Augmented” 2SLS “Augmented” 2SLS “Augmented” 2SLS “Augmented” 2SLS “Augmented”
model model model model model model
Main s.holder 0.115 0.001 0.307 0.499 0.408 0.829

Quota (0.330) (0.394) (0.243) (0.614) (0.278) (0.608)
nnn nnn
Fin. institution 3.302 9.585 3.053n 11.232nnn
Subscriber (1.161) (3.230) (1.740) (2.290)
Intention to 2.593nnn 7.535n 1.558 6.014n
Go public (0.662) (4.098) (0.985) (3.325)
Intention to sell 5.594nnn 17.592nnn 4.583nn 14.624nnn

To private equity (1.881) (4.691) (2.008) (0.568)
Time dummies Y Y Y Y Y Y Y Y Y Y Y Y
Area dummies Y Y Y Y Y Y Y Y Y Y Y Y
þcontrols Y Y Y Y Y Y Y Y Y Y Y Y
Overid. test (p- 0.1237 0.0891 0.1370 0.0148 0.5209 0.8715
value)
Observations 12,662 12,662 12,045 12,045 12,314 12,314 11,723 11,723 6585 6579 6320 6314
Panel B: Process innovation
(1) (2) (3) (4) (5) (6) (7) (8) (9) (10) (11) (12)
2SLS “Augmented” 2SLS “Augmented” 2SLS “Augmented” 2SLS “Augmented” 2SLS “Augmented” 2SLS “Augmented”
model model model model model model
Main s.holder 0. 512n 1.189 0.562nn 1.960nnn 0.879nnn 2.173nnn

Quota (0.265) (1.008) (0.252) (0.614) (0.219) (0.332)
Fin. institution 3.425nnn 13.703nnn 2.039 10.970nnn
Subscriber (1.329) (3.759) (1.288) (4.154)
Intention to 3.570nnn 10.856n 1.977n 5.557
Go public (0.779) (6.554) (1.044) (6.612)
Intention to sell 4.588n 0.657 0.241 1.231
To private equity (2.583) (5.800) (2.844) (0.779)
Time dummies Y Y Y Y Y Y Y Y Y Y Y Y
Area dummies Y Y Y Y Y Y Y Y Y Y Y Y
þcontrols Y Y Y Y Y Y Y Y Y Y Y Y
Overid. test (p- 0.0593 0.0134 0.3136 0.0503 0.5198 0.8818
value)
Observations 12,586 12,679 12,062 12,062 12,330 12,330 11,739 11,739 6593 6594 6329 6314
Note: The table reports regression coefficients and associated standard errors (in parentheses). The dependent variable of the regressions in Panel A is the dummy for Product Innovation and the dependent variable
of the regressions in Panel B is the dummy for Process innovation. The estimation method is reported at the top of each column. We instrument both the main shareholder quota and the variables proxying for
changes in the ownership structure. The set of instruments includes number of bank branches in the province in 1936 (per 100,000 inhab.), the ratio of local bank branches to total bank branches, savings banks
(per 100,000 inhab.) in the province in 1936, branches opened by new incumbents in the province over the 1991–1998 period (net of branches closed). Time dummies denote the year of the survey. Area dummies
refer to the area of the country where the firm is headquartered (Center or South). “þ controls” denotes the RHS variables of the regressions of Table 2. The table also reports the p-values of a Sargan test, as a test of
overidentifying restrictions, and of a Wald test, as a test of exogeneity of the variables that have been instrumented.
n
nn
Significant at 5%.
nnn
Significant at 1%.
Table 8
The effects of ownership on R&D, technology adoption and total investment.
Panel A: 2SLS
(1) (2) (3) (4) (5) (6) (7) (8) (9) (10)
R&D INFO TECH PATENTS INNO Skilled R&D workers I. PROD I. PROC Total Tot. invest.
invest. not workers investment expenditure
for R&D hiring
Main s.holder 1.022nn 0.801 0.222nn 1.214 9.462n 0.155n 1.076nn 0.158 0.189 0.519
Quota (0.417) (0.514) (0.103) (2.383) (5.852) (0.086) (0.422) (1.186) (0.250) (0.503)
Time dummies Y Y Y Y Y Y Y Y Y Y
Area dummies Y Y Y Y Y Y Y Y Y Y
Over. test (p- 0.8425 0.6175 0.9306 0.6017 0.7271 0.4783 0.9774 0.0023 0.0453 0.5977
value)
Observations 12,112 10,584 12,094 2895 8029 9856 12,113 12,130 12,130 9090
Panel B: AGLS
(1) (2) (3) (5) (6) (7) (8) (9)

R&D INFO TECH PATENTS Skilled R&D workers I. PROD I. PROC Total
workers Investment
hiring
Main s.holder 2.784nn 2.671 5.791nn 65.907n 0.347nn 2.852nn 0.328 0.957
Quota (1.130) (1.688) (2.711) (40.903) (0.174) (1.148) (0.946) (1.249)
Time dummies Y Y Y Y Y Y Y Y
Area dummies Y Y Y Y Y Y Y Y
þcontrols Y Y Y Y Y Y Y Y
Wald test (p- 0.0015 0.0670 0.0207 0.0223 0.0195 0.0017 0.6965 0.4676
value)
Observations 12,109 10,582 12,055 8029 9856 12,113 12,130 12,130
Note: The table reports regression coefficients and associated standard errors (in parentheses) for the effects of ownership concentration on R&D, tech-
nology adoption, and total investment. The dependent variables are reported at the top of each column. The regressions in Panel A are estimated by two-
stage least squares (2SLS); the regressions in Panel B, columns (1)–(3) and (7)–(9), are estimated by two-stage conditional maximum likelihood (AGLS,
Newey's minimum chi-squared estimator); columns (5) and (6) are estimated with AGLS tobit estimator. We instrument the variable Main s.holder quota.
The set of instruments includes number of bank branches in the province in 1936 (per 100,000 inhab.), the ratio of local to total bank branches, number of
savings banks in the province in 1936 (per 100,000 inhab.), branches opened by new entrants in the province over the 1991–1998 period (net of branches
closed). In column (9) we also use the number of branches opened by incumbents in the province over the 1991–1998 period (net of branches closed). In
column (5) we also use the number of cooperative banks (per 100,000 inhabitants) in the province in 1936. For convenience in columns (7) and (8) we
report the baseline regression results of Table 4 (from columns 1a, 1b, 4a and 4b). Time dummies denote the year of the survey. Area dummies refer to the
area of the country where the firm is headquartered (Center or South). “þ controls” denotes the RHS variables of the regressions in Table 2. The table also
reports the p-values of a Sargan test, as a test of overidentifying restrictions, and of a Wald test, as a test of exogeneity.
n
nn
Significant at 5%.
manufacturing firms do not invent new hardware, software, and communication equipment but acquire them from ICT
firms, research centers, and universities (Confindustria, 2003). The coefficient of the main equity share is insignificant
(column 2). We perform two more tests on the adoption of innovations. First, the survey reports whether a firm acquired
patents abroad. The effect of ownership concentration on patent acquisition is positive (column 3). Second, we consider the
(log) expenditures for innovation that are not allocated to R&D (acquisitions of plants, know how, training, marketing of new
products).20 The effect of ownership concentration is negative but insignificant (column 4). In sum, the results suggest that
ownership concentration discourages research more than the adoption of innovations.
Finally, we investigate whether ownership concentration affects investment in human capital, an important input for
inventing and adopting new technologies. In columns 5 and 6, the dependent variables are respectively the number of
workers with a college degree hired in the year of the survey and the percentage of workers in R&D activities. The effects of
ownership concentration on these variables are negative and significant at the 10% level.
4.5. Innovation and total investment
The reader may wonder whether the effect of ownership concentration on innovation simply reflects a broader effect on
total (traditional or innovative) investment. In Table 8, we test whether ownership concentration helps explain firms' total
investment. We define a binary variable that takes the value of one if in the three years covered by the survey the firm
20
This information is available only in the last two waves of the survey.
invested for purchasing plants or equipment, zero otherwise; and a variable equal to the average investment expenditure
over the three years. For comparison purposes, in columns 7 and 8, we carry over the baseline estimates from Table 4. In
column 9, we report the estimates for the propensity to invest; in column 10, those for the average investment expenditure.
The estimates imply that ownership concentration does not matter for total investment, thus suggesting that the effect of
ownership concentration is specific to innovation.
5. Disentangling the ownership–innovation links
As noted, firms' ownership structure may affect innovation decisions by influencing agency conflicts between majority
and minority shareholders and between shareholders and managers. It can also affect innovation by influencing large
shareholders' diversification and risk aversion. In Tables 9 and 10, we seek to disentangle the contribution of these
mechanisms.
5.1. Agency problems
Agency conflicts between large and minority shareholders are common in Italian businesses (Onida, 2004). In Table 9, we
investigate further whether the effect of ownership concentration on innovation captures problems of entrenchment and
expropriation of minority shareholders by large shareholders. Claessens et al. (2002) argue that entrenchment and value
extraction by a large shareholder are more severe when control rights are not aligned with cash-flow rights, because the
incentive to extract value is less restrained by the controlling shareholder's cash-flow stake. In columns 1–2, we include a
dummy for whether the main shareholder has control over the firm. The alignment between control rights and cash flow
rights promotes innovation, although the coefficient on the dummy tends to lose significance when we control for own-
ership concentration (columns 3–4).21
In columns 5 and 6, we display the results after sub-sampling firms according to whether they are part of a group or not.
Firms that are part of a group might be more exposed to conflicts among shareholders. In fact, due to the chain shareholding
inside the group, controlling shareholders could effectively have limited cash flow rights inside the firms. We obtain that
attributing control to the main shareholder positively affects product innovation for firms that are not part of a group, while
it has no effect for group affiliates. Main shareholders effectively hold less cash flow rights inside group-affiliated firms, so
that attributing control to them could produce less benefits in aligning cash flow rights with control rights.
In a context like the Italian one where firms feature concentrated ownership, conflicts among shareholders are more
relevant than conflicts between shareholders and managers. Yet, our data also allow to investigate the presence of conflicts
involving managers. The last two survey waves ask each firm the share of external managers on its board. In columns 7–10,
we restrict the analysis to these two waves and control for this share. As instruments, we use the variables reflecting the
constrictiveness of the 1936 regulation, again on the ground that governance practices are persistent over time.22 The
likelihood of product innovation decreases with the share of external managers. This suggests that separating ownership
from management exacerbates conflicts between shareholders and managers, stifling innovation.
5.2. Risk and diversification
In Table 10, we explore the role of risk and diversification in the innovation process. If the financial portfolio of a firm
is not diversified, large investors could be reluctant to undertake risky innovations (Bolton and Von-Thadden, 1998). The
survey asks firms about the allocation of their financial investments among equity participation in Italian companies,
equity participation in foreign companies, short-term Italian bonds, medium- and long-term Italian bonds, foreign
bonds, other financial instruments. The rate of response is about 35%. We measure the concentration of firms' financial
portfolio with the Herfindahl–Hirschman index of the various asset shares.23 In columns 1–2, the coefficient of the
interaction between this index and the measure of ownership concentration is negative and significant both in the
regression for product innovation and in that for process innovation. Hence, the negative effect of ownership con-
centration appears to depend on the firm's profile of financial diversification: the less diversified the firm is, the lower
its propensity to innovate if its ownership is more concentrated.24 In columns 3 and 4, we use a measure of industrial
diversification. Plausibly, the lower the number of industries in which a firm is active, the less diversified its production.
21
We address the endogeneity issues as we did with the dummies proxying for shocks to the ownership structure (see Section 4.3 for the procedure).
The instruments are drawn from the same set of variables used for ownership concentration. It is plausible that the conditions in the local financial market
affect both equity holders' shares in a firm and their involvement in the firm (that is, whether they retain control).
22
Restrictions on the local supply of credit can hinder the ability of firms to pay the (possibly high) salaries of external managers and induce firms to
rely on competencies inside the family. They can also affect individuals' ability to invest in human capital and, hence, the supply of managerial capital.
23
We instrument the Herfindahl index using the same variables used for ownership concentration. Local credit market conditions can affect firms'
financial portfolio diversification through the financial information that local banks are able to provide to the firms and through firms' incentives to hold
deposits in banks.
24
The distinction between the portfolio of a firm and that of the main shareholder is blurred especially for small- or medium-sized firms. The findings
are confirmed when we focus on firms with less than 50 employees.
Table 9
Disentangling the link ownership-innovation. Agency problems.
Panel A: Product Innovation
Main Shareholder has control External Managers
(1) (2) (3) (4) (5) (6) (7) (8) (9) (10)
2SLS “Augmented” 2SLS “Augmented” 2SLS 2SLS 2SLS AGLS 2SLS AGLS
model model
Main s.holder 1.080nn 0.715 1.485nn 3.833nn

quota
(0.428) (0.776) (0.709) (1.815)
Main s.holder has 1.217n 3.565nn 0.493 4.401nnn 0.482 1.211n
Control (0.693) (1.420) (0.894) (1.456) (0.826) (0.627)
External managers 0.841n 1.634nnn 0.181 0.284
(0.449) (0.395) (0.690) (1.780)

dum.
Sample All All All All Group No Group All All All All
Overid. test (p- 0.0000 0.9928 0.1618 0.8037 0.4301 0.8692
value)
Wald test (p-value) 0.0155 0.0032
Observations 12,742 12,742 12,113 12113 3097 9618 2611 2603 2483 2483
Panel B: Process innovation
Main Shareholder has control External Managers
(1) (2) (3) (4) (5) (6) (7) (8) (9) (10)
2SLS “Augmented” 2SLS “Augmented” 2SLS 2SLS 2SLS AGLS 2SLS AGLS
model model
Main s.holder 0.222 2.977nnn 0.307 0.738

quota
(0.407) (1.144) (0.562) (1.504)
Main s.holder has 0.585 1.360 0.862 1.321 1.016 0.265
Control (0.585) (1.447) (0.854) (2.065) (0.957) (0.511)
External managers 0.343 1.035 0.325 1.073
(0.377) (0.737) (0.561) (1.465)

dum.
Sample All All All All Group No Group All All All All
Overid. test (p- 0.0184 0.0125 0.4557 0.0074 0.2582 0.0609
value)
Wald test (p-value) 0.1873 0.2486
Observations 12,759 12,759 12,130 12,130 3110 9632 2617 2617 2489 2489
Note: The table reports regression coefficients and associated standard errors (in parentheses) for the role of agency problems in the link ownership-
innovation. The estimation method is reported at the top of each column. The dependent variable in the regressions in Panel A is Product Innovation. The
dependent variable in the regressions in Panel B is Process Innovation. We instrument all the variables for which the coefficients are reported in the table.
The set of instruments for the regressions in columns (1)–(6) includes number of bank branches in the province in 1936 (per 100,000 inhab), the ratio of
local to total bank branches, number of cooperative banks (per 100,000 inhab.), number of savings banks in the province in 1936 (per 100,000 inhab.),
branches opened by new entrants and by incumbents in the province over the 1991–1998 period (net of branches closed). For the regressions in columns
(7)–(10) we include as instruments: branches opened by new entrants in the province over the 1991–1998 period (net of branches closed) and the Guiso
and Jappelli index of financial awareness of households. Finally, for the regressions in columns (9)–(10) we also add as instruments the number of bank
branches in the province in 1936 (per 100,000 inhab) and the ratio of local to total bank branches. Time dummies denote the year of the survey. Area
dummies refer to the area of the country where the firm is headquartered. “þ controls” denotes the RHS variables of the regressions in Table 2. The table
also reports the p-values of a Sargan test, as a test of overidentifying restrictions, and of a Wald test, as a test of exogeneity.
n
nn
Significant at 5%.
nnn
Significant at 1%.
We interact the main equity share with the dummy variable for whether the firm is classified in a five-digit ATECO
sector, which indicates low industrial diversification. For product innovation, the coefficient on the interaction variable
is negative and significant. Thus, diversification, whether financial or industrial, mitigates the negative effect of own-
ership concentration on innovation.
Finally, in columns 5–6 we allow for differences in the legal structure by inserting a dummy that takes the value of
one if the firm is a corporation. Shareholders of corporations are protected by limited liability, so they might be less
Table 10
Disentangling the link ownership-innovation. Risk and diversification.
Panel A: 2SLS
(1) (2) (3) (4) (5) (6) (7) (8)

I.PROD I.PROC I.PROD I.PROC I.PROD I.PROC I.PROD I.PROC
Main s.holder 0.442 0.169 1.126nnn 0.253 1.031nn 0.014

quota
(0.466) (0.429) (0.396) (0.338) (0.431) (0.447)
Main s.holder 0.438nnn 0.220nnn
quotan
Financ. (0.083) (0.074)
Concentration
Main s.holder 0.053nn 0.027
quotan
Ateco 5 digit (0.026) (0.022)
Corporation 1.087n 0.979 0.493 1.530n
(0.628) (0.644) (0.879) (0.927)
Time and Area Y Y Y Y Y Y Y Y

dum.
þ controls Y Y Y Y Y Y Y Y
Over. test (p-value) 0.1072 0.4663 0.9737 0.0109 0.1971 0.0322 0.9913 0.0353
Observations 4332 4337 12,113 12,130 12,742 12,759 12,113 12,130
Panel B: AGLS or Augmented Model
(1) (2) (3) (4) (5) (6) (7) (8)

AGLS AGLS AGLS AGLS “Augmented” “Augmented” “Augmented” “Augmented”
model model model model
I.PROD I.PROC I.PROD I.PROC I.PROD I.PROC I.PROD I.PROC
nnn
Main s.holder 1.119 0.346 3.007 0.611 0.114 2.591nn
quota
(1.231) (1.126) (1.085) (0.881) (0.686) (1.017)
Main s.holder 1.152nnn 0.554nnn
quotan
Financ. (0.227) (0.210)
Concentration
Main s.holder 0.148nn 0.071
quotan
Ateco 5 digit (0.071) (0.058)
Corporation 3.205nn 2.668n 4.691nnn 5.781nnn
(1.392) (1.635) (1.179) (1.634)
Time and Area Y Y Y Y Y Y Y Y

dum.
þ controls Y Y Y Y Y Y Y Y
Wald test (p-value) 0.0057 0.4447 0.0013 0.4336
Observations 4332 4337 12,113 12,130 12,742 12,759 12,113 12,130
Note: The table reports regression coefficients and associated standard errors (in parentheses) for the role of risk aversion in the link ownership-innovation.
The dependent variables are reported at the top of each column. The regressions in Panel A are estimated by two-stage least squares (2SLS) and the
regressions in Panel B are estimated by two-stage conditional maximum likelihood (AGLS, Newey's minimum chi-squared estimator) or “Augmented”
Model. We instrument all the variables for which the coefficients are reported in the table. The basic set of instruments includes number of bank branches
in the province in 1936 (per 100,000 inhab.), the ratio of local to total bank branches, number of cooperative banks (per 100,000 inhab.) in the province in
1936, branches opened by new entrants in the province over the 1991–1998 period (net of branches closed). For the regressions in columns (1)–(4) we also
use the interaction of the basic instruments with the index that we use for diversification. For the regressions in columns (5)–(8) we use also the number of
savings banks in the province in 1936 (per 100,000 inhab.) and the number of branches opened by incumbent banks in the province over the 1991–1998
period (net of branches closed). Time dummies denote the year of the survey. Area dummies refer to the area of the country where the firm is head-
quartered (Center of South). “þ controls” denotes the RHS variables of the regressions in Table 2. The table also reports the p-values of a Sargan test, as a
test of overidentifying restrictions, and of a Wald test, as a test of exogeneity.
n
nn
Significant at 5%.
nnn
Significant at 1%.
averse to risky innovations. We instrument the “corporation” dummy using the same variables used for ownership
concentration. As expected, the coefficient on the dummy is positive and significant. The reader could however suspect
that this is picking up the effect of dispersed ownership (corporations tend to have a more dispersed ownership
structure). When we add the main equity share to the regression (columns 7–8), the coefficient on the dummy retains
its significance.
6. Families and financial institutions
As Table 1 shows, in the majority of the firms in our sample the main shareholder is an individual or a family. A reason for
this importance of family firms is that for several decades legal prescriptions introduced in the 1930s prevented banks from
holding shares in firms. Despite a change in the legislation in the 1990s (d.lgs. 481/92 and 385/93), financial institutions
continue to have a limited role as firm owners. Among financial institutions, investment and equity funds are the most
frequent shareholders, while banks, insurance companies and pension funds hold limited equity stakes (Bianco and Bianchi,
2008; OECD, 2014a). Barucci and Falini (2005) report that in 2002 banks' equity share in listed firms equalled 3.4% on
average. This is also confirmed by the last wave of the Capitalia survey, which is the only survey wave asking firms about the
type of financial shareholders: banks are the main shareholder in only a handful of cases.
In Table 11, we add, in separate regressions, a “family business” dummy, when the main shareholder is an individual or
family, and a “financial institution” dummy, when the main shareholder is a financial institution. The owner type dummy is
treated as endogenous and instrumented using the same variables used for ownership concentration.25 With the “family
business” dummy, we have also experimented by adding to the set of instruments an index of financial awareness similar to
that in Guiso and Jappelli (2005). This is based on the Bank of Italy Surveys of Household Income and Wealth (SHIW) and is
computed at the regional level as the fraction of households aware of basic financial instruments (e.g., stocks and mutual
funds). Financial awareness can affect the incentive of families to retain a large stake in their companies. By contrast, it
should not affect innovation directly, also in light of the fact that the index is an average at the regional level and is based on
household survey data dating back to 1995. Results (not tabulated) do not change when we add this instrument.26
Based on the estimates (columns 1–4 for product innovation, 11–14 for process innovation), family firms are somewhat
more likely to carry out product innovations than firms in which the main shareholder is a financial institution, another firm
or a holding company.27 This might appear in contrast with the positive effect of diversification (Table 10) because family
owners are typically less diversified than institutional owners. We also experiment with controlling for lack of diversifi-
cation by interacting the family owner dummy with our proxies for diversification. The results are confirmed as family firms
are somewhat more likely to innovate, but the effect is smaller if diversification is low (columns 4 and 14). An interpretation
for the effect of family ownership is that families can be long-term value maximizers, since the company will be passed to
future generations of family members (Caselli and Gennaioli, 2013) and family owners tend to have a strong attachment to
their firms (Sraer and Thesmar, 2007). This long-termism benefits new technologies, which entail long gestation periods. By
contrast, institutional owners may have a short horizon because they often have a limited investment holding period and, in
addition, they may be assessed by the financial press and by courts on the ground of their short-term performance. Italian
institutional investors allegedly exhibit both these problems (OECD, 2014a,b,c). Because of their limited size (which prevents
them from exploiting economies to scale), many institutional investors face difficulties in engaging in illiquid long-term
investments, such as holding long-term equity positions in firms. Moreover, quarterly assessments of asset liquidity
exacerbate institutional investors' short-termism. Finally, the fact that many Italian firms are small and informationally
opaque makes it difficult for institutional investors to assess the risk associated with holding long-term equity positions
in firms.
When we insert a dummy capturing whether a financial institution is the main owner, the estimated coefficient on this
dummy is negative but insignificant for product innovation (column 5). This result remains unaltered if we restrict attention
to the last survey wave (which reports the type of financial institution) and redefine the financial institution dummy as
taking the value of one if the main owner is a non-bank financial institution (see column 7 of Table 11).28 And the result also
carries through if we insert two distinct dummies for ownership by financial institutions, one for non-banks and one for
banks (column 8).29
When we add to the regressions ownership concentration and its interaction with the financial institution dummy, the
dummy gains significance (column 10). However, the coefficient on the interaction implies that the effect of institutional
ownership becomes positive when the equity share exceeds 70%. This is in line with Aghion et al. (2013) who find a positive
association between innovation and institutional ownership concentration in the United States. An interpretation is that
larger equity stakes motivate financial institutions to monitor more, alleviating the agency problems that hinder innovation.
25
In fact, the owner type is likely to reflect the same needs and incentives to open participation to new shareholders as ownership concentration (see,
e.g., Cosh et al., 2009, for an analysis of the endogeneity of owner types). We use 2SLS when considering the linear probability model and estimate a
“residual augmented” model when considering the probit specification, as detailed in Section 4.3.
26
We do not expect this to be a good instrument for ownership concentration. If we insert it in the baseline regressions, it is insignificant and the
estimation results are virtually unchanged.
27
In untabulated tests, we investigate the impact of family ownership on the phases of innovation (R&D and investment in information technology).
The coefficients are estimated imprecisely. We also obtain that, like ownership concentration, family ownership does not influence total investment.
28
In column 6, we report the estimated coefficient for the financial institution (bank or non-bank) dummy when we restrict attention to the last
survey wave.
29
As noted, for the other survey waves we lack information on the type of financial institution. However, in Table 7 we obtain that a firm's intention to
sell minority equity shares to a financial institution positively affects innovation. Furthermore, in untabulated tests we obtain that this effect turns out to be
especially strong for young firms. Since private equity funds generally play a role in promoting long-term, innovative projects in young firms, these results
may confirm the role of private equity funds in firms' innovation.
188
Table 11
The owner type and innovation.
Panel A: 2SLS
(1) (2) (3) (4) (5) (6) (7) (8) (9) (10) (11) (12) (13) (14) (15) (16) (17)
I.PROD I.PROD I.PROD I.PROD I.PROD I.PROD I.PROD I.PROD I.PROD I.PROD I.PROC I.PROC I.PROC I.PROC I.PROC I.PROC I.PROC
Main s.holder 0.708nn 1.700 0.688n 0.700nn 1.880nn 0.074 3.383nn 0.031 0.226 1.306
quota
(0.319) (1.454) (0.416) (0.334) (0.860) (0.272) (1.728) (0.358) (0.335) (0.942)
Family 0.708nnn 0.439n 0.463 0.660 0.082 0.018 3.169nn 0.169
(0.245) (0.253) (1.322) (0.533) (0.216) (0.220) (1.571) (0.419)
Familyn Main 1.252 4.367nn
s.holder quota (1.794) (2.120)
Familyn Financial 0.373nnn 0.170nnn

Concentration (0.079) (0.064)
Financial 0.532 0.143 0.211 6.043n 0.945n 0.767 5.499
institution
(0.495) (0.415) (0.547) (3.408) (0.541) (0.548) (3.730)
Fin. institutionn 8.261n 8.932n
Main s.holder (4.678) (5.136)
quota
Non-bank fin inst. 0.153 0.425
(0.424) (0.617)
Bank 12.744
(24.065)
Time and Area Y Y Y Y Y Y Y Y Y Y Y Y Y Y Y Y Y

dum.
þ controls Y Y Y Y Y Y Y Y Y Y Y Y Y Y Y Y Y
Over. test (p-value) 0.5278 0.8396 0.8260 0.1674 0.0159 0.1642 0.1862 0.3989 0.1190 0.9683 0.0008 0.0002 0.2267 0.5139 0.0052 0.0004 0.4020
Observations 12,505 11,876 11,876 4194 12,742 3256 3256 3256 12,113 12,113 12,522 11,893 11,893 4199 12,759 12,130 12,130
Panel B: “Augmented” Model
(1) (2) (3) (4) (5) (6) (7) (8) (9) (10) (11) (12) (13) (14) (15) (16) (17)
I.PROD I.PROD I.PROD I.PROD I.PROD I.PROD I.PROD I.PROD I.PROD I.PROD I.PROC I.PROC I.PROC I.PROC I.PROC I.PROC I.PROC
Main s.holder 1.320n 6.032n 1.648n 1.813nn 4.756nnn 0.403 9.532n 0.062 0.082 3.067nn
quota
(0.898) (3.689) (1.296) (0.853) (1.535) (0.853) (5.493) (1.093) (0.853) (1.355)
Family 2.064nnn 0.662 4.828n 0.643 0.144 1.662nnn 10.480nn 0.412nn
(0.501) (0.411) (3.003) (0.903) (0.456) (0.526) (4.767) (0.177)
FamilynMain 5.915 12.483n
s.holder quota (4.158) (6.627)
Familyn Financial 0.891nnn 1.897nnn
Concentration (0.193) (0.653)
Financial 1.664 0.328 2.635nnn 11.995nn 2.276nn 3.827nnn 11.761nn
institution
(1.230) (1.036) (0.662) (5.847) (1.071) (0.777) (5.880)
Fin. institutionn 21.170nnn 22.578nnn
Main s.holder (7.286) (7.551)
quota
Non-bank fin inst. 0.351 0.317
(1.029) (1.111)
Time and Area Y Y Y Y Y Y Y Y Y Y Y Y Y Y Y Y Y

dum.
þ controls Y Y Y Y Y Y Y Y Y Y Y Y Y Y Y Y Y
Observations 12,505 11,876 11,876 4194 12,742 3256 3256 3256 12,113 12,113 12,522 11,893 11,893 4199 12,759 12,130 12,130
Note: The table reports regression coefficients and associated standard errors (in parentheses). The dependent variables are reported at the top of each column. The regressions in Panel A are estimated by two-
stage least squares (2SLS), the regressions in Panel B are estimated by the “Augmented” model. We instrument all the variables for which the coefficients are reported in the table. The set of instruments includes
the number of bank branches in the province in 1936 (per 100,000 inhab.), the ratio of local bank branches to total bank branches, cooperative banks (per 100,000 inhab.), savings banks (per 100,000 inhab.), in the
province in 1936, branches opened by new entrants and branches opened by incumbent banks in the province over the 1991–1998 period (net of branches closed). In columns (4) and (11), we also use the
interaction of the basic instruments with the index that we use for diversification. Time dummies denote the year of the survey. Area dummies refer to the area of the country where the firm is headquartered
(Center or South). Regression 8, in Panel B, includes a dummy equal to one if the bank is the main owner. “þ controls” denotes the RHS variables of the regressions in Table 2. The table also reports the p-values of a
Sargan test, as a test of overidentifying restrictions.
n
nn

Significant at 5%.
nnn
Significant at 1%.
189
7. Conclusion
This paper has tested the impact of firms' ownership structure on innovation decisions in a context where ownership
concentration is pronounced and conflicts often arise between main shareholders and minority ones. We have found that,
after accounting for the possible endogeneity of ownership levels, ownership concentration has a large, negative effect on
product innovation. This effect is stronger for firms' R&D effort, for medium-sized and large firms, and for firms in tradi-
tional sectors. We have investigated the channels through which ownership structure affects innovation. The analysis
reveals that conflicts between large and minority shareholders may indeed hinder innovation when ownership is con-
centrated. Moreover, risk aversion induced by lack of financial and industrial diversification exacerbates large shareholders'
reluctance to innovate. In the last part of the paper, we have examined whether the type of the main shareholder plays a
role in innovation decisions. Firms in which a family is the main shareholder are more likely to innovate than firms in which
a financial institution is the main shareholder, but the benefits of financial institutions for innovation increase with their
equity stake.
Technological change is a major mechanism through which firms grow and expand domestically and abroad. This
analysis suggests that, by influencing innovation, corporate governance can be a driving force of these processes. An
interesting research topic could be to relate the dynamics of industrial sectors to the link between corporate governance and
innovation. For example, one could investigate how shocks to the profitability or riskiness of industries affect innovation
depending on the ownership structure predominant in the industries. We leave this and other important issues for future
research.
Acknowledgments
We wish to thank the editor, an anonymous referee, Magda Bianco, Zsuzsanna Fluck, Ana Maria Herrera, Maurizio
Iacopetta, Andrea Ichino, Franco Peracchi, Jeff Wooldridge, and seminar and conference participants at Bank of Italy, Einaudi
Institute for Economics and Finance, LUISS-Guido Carli University, Michigan State University, Universitat Pompeu Fabra,
University of Bologna, University of Nice-Sciences Po, Italian Ministry of Economy and Finance (Treasury Department),
Organisation for Economic Cooperation and Development (Paris), Financial Intermediation Research Society (Sydney) for
helpful comments and conversations. All remaining errors are ours.
Appendix A
See Table A1.
Table A1
Variables and sources. This table describes the definitions of the variables used in the empirical analysis. Four main data sources are used in the analysis:
four waves of the Capitalia Survey of Italian Manufacturing Firms (SIMF), which cover three-year periods ending in 1997, 2000, 2003 and 2006; the
province-level database of the Italian National Statistics Office (ISTAT); the Statistical Bulletin of the Bank of Italy (SBBI); and the book “Struttura funzionale
e territoriale del sistema bancario italiano 1936–1974” (SFT) by the Bank of Italy.
Variable Definition and source (in parentheses)
Dependent variables
Product and Process innovation The survey asks each firm: “In the last three years, did the firm realize (1) product innovations,
(2) process innovations, (3) organizational innovations related to product innovations, (4) organiza-
tional innovations related to process innovations?”. The dummy for product (process) innovation is
equal to one if the firm reports to have realized product (process) innovations or organizational
innovations related to product (process) innovations over the three years covered by the survey, zero
otherwise (SIMF)
Investment in R&D The survey asks each firm: “In the last three years, did the firm carry out R&D expenditures?”. The
dummy for R&D investment is equal to one if the firm answers “yes”, zero otherwise (SIMF)
Information technology The survey asks each firm: “In the last three years, did the firm carry out investment for the intro-
duction of hardware, software, telecommunication networks?”. The dummy for information tech-
nology is equal to one if the firm answers “yes”, zero otherwise (SIMF)
Patents The survey asks each firm: “In the last three years, did the firm acquire or sell patents abroad?”. The
dummy for patents is equal to one if the firm acquired patents abroad in the years of the survey, zero
otherwise (SIMF)
Innovation expenditures not for R&D The natural logarithm of expenditures for technological innovation that are not directly allocated to
R&D (SIMF)
Total investment The survey asks each firm: “In the last three years, did the firm carry out investment for purchasing
plants or equipment, and, if so, for what amount in each year?”. The dummy for total investments is
equal to one if the firm answers “yes”, zero otherwise (SIMF)
Total investment expenditure
Table A1 (continued )
Variable Definition and source (in parentheses)
The survey asks each firm: “In the last three years, did the firm carry out investment for purchasing
plants or equipment, and, if so, for what amount in each year?”. Total investment expenditure is a
variable equal to the average investment expenditures over the three years (SIMF)
Hiring of high skilled workers Number of workers with a bachelor's degree hired in the last year of the survey (SIMF)
R&D workers The percentage of firm's employees employed in R&D activities (SIMF)
Endogenous variables
Main shareholder quota Capital share held by the main shareholder (SIMF)
Two main shareholder quota Capital share held together by the two main shareholders (SIMF)
Owner type The survey asks each firm to report the characteristics of the main shareholders owning the firm. We
distinguish whether the main shareholder is a family/single person, a financial institution, a firm or
holding (SIMF)
Business type The survey asks each firm whether it is publicly listed. In the survey, the information on whether the
firm is a private limited company (LTD) or a public limited company (PLCs) is available only for the
2003 and 2006 surveys (SIMF)
Financial institution subscriber A dummy equals to one if a financial institution underwrote new shares of the firm in the three years
of the survey; zero otherwise (SIMF)
Intention to go public A dummy equals to one if firm plans to go public in the following year; zero otherwise (SIMF)
Main shareholder has control A dummy equals to one if the main shareholder reports to have control over the firm; zero otherwise
(SIMF)
External managers The percentage of external managers in the board of the firm (SIMF)
Financial concentration The concentration of firms' financial portfolio, measured as the Herfindahl–Hirschman index of the
various asset shares. The survey asks each firm to report the allocation of its financial investments
among equity participation in Italian companies, equity participation in foreign companies, short-term
Italian bonds, medium- and long-term Italian bonds, foreign bonds, other financial instruments (SIMF)
Control variables
Group, Consortium The survey asks each firm to report whether it belongs to a group of firms and whether it belongs to a
consortium. The dummies for participation in a group and consortium take the value of one if the firm
answers “yes” to the questions, zero otherwise (SIMF)
Sector of activity The survey reports the sector of activity of firms (ATECO code). Based on this information, firms are
classified as traditional, high tech and scale intensive using Pavitt's taxonomy (SIMF)
Age of the firm Number of years since inception (SIMF)
Number of employees Total number of employees in the last year of the survey (SIMF)
Total assets, sales, current assets and These variables are balance sheet data. They are available for each year covered by the survey. We use
inventories the three-year average (SIMF)
Number of banks, Duration relationship with The survey asks each firm to report the number of banks with which it maintains a stable credit
main bank relationship and the duration (in years) of the relationship with the main lender, at the time of
interview (SIMF)
Credit rationing The survey asks each firm: “In the last year of the survey, would the firm have liked to obtain more
credit at the market interest rate?”. The dummy for credit rationing is equal to one if the firm answers
“yes”, zero otherwise (SIMF)
Ateco n-digit Dummy that takes the value of one if the firm reports its ATECO classification as an n-digit number;
0 otherwise (SIMF)
Center, South Dummy variables that take the value of one if the firm is located in a central or southern province;
zero otherwise (SIMF)
Number of branches For each province and year we calculated the number of branches per 1000 inhabitants; then we
computed the average over the years 1991–1998 (SBBI)
Provincial GDP growth Average growth rate of the value added of the province where the firm is located over the years 1985–
1994 (SBBI)
Provincial Herfindahl Average Herfindahl–Hirschman Index (HHI) on bank loans in the province during the 1985–1995
period (SBBI)
Local financial development We use the measure proposed by Guiso et al. (2004). This is based on the estimates of the fixed effects
for geographical region from a probit for the probability that, ceteris paribus, a household is shut off
from the credit market in Italy
Efficiency of the court system We considered the number of civil suits pending in each of the 27 district courts of Italy, scaled by the
population of the district. We imputed this variable to the firms according to the districts where they
are headquartered (ISTAT)
Instrumental variables
Bank branches in 1936 Number of bank branches in the year 1936 in the province, per 100,000 inhabitants (SFT)
Local/Total banks in 1936 Ratio of local to total bank branches in the year 1936 in the province (SFT)
Cooperatives banks in 1936 Number of cooperative banks in the year 1936 in the province, per 100,000 inhabitants (SFT)
Saving banks in 1936 Number of savings banks in the year 1936 in the province, per 100,000 inhabitants (SFT)
New branches entrant For each province and year we calculate the number of branches created by entrant banks per 100,000
inhabitants. Then, we compute the average over the years 1991–1998 (SBBI)
New branches incumbent For each province and year we calculate the number of branches created minus those closed by
incumbent banks per 100,000 inhabitants. Then we compute the average over the years 1991–1998
(SBBI)
Appendix B. Supplementary data
Supplementary data associated with this paper can be found in the online version at http://dx.doi.org/10.1016/j.euro
ecorev.2015.09.007.
References
Adkins, L., 2012. Testing parameter significance in instrumental variables probit estimators: some simulation results. J. Stat. Comput. Simul. 82 (10),
1415–1436.
Aghion, P., Van Reenen, J., Zingales, L., 2013. Innovation and institutional ownership. Am. Econ. Rev. 13 (1), 277–304.
Angrist, J., Pischke, J., 2009. Mostly Harmless Econometrics: An Empiricist's Companion. Princeton University Press.
Arendt J., Holm, A., 2006. Probit Models with Binary Endogenous Regressors. Discussion Papers on Business and Economics, No. 4/2006. Department of
Business and Economics Faculty of Social Sciences, University of Southern Denmark.
Barucci, E., Falini, J., 2005. Institutional Investors and Corporate Governance in Italy: An Analysis of their Stockholdings. Working Paper, Polytechnic of
Milan.
Bebchuk, L.A., Roe, M.J., 1999. A theory of path dependence in corporate ownership and governance. Stanf. Law Rev. 52, 127–170.
Benfratello, L., Schiantarelli, F., Sembenelli, A., 2008. Banks and innovation: microeconometric evidence on Italian firms. J. Financ. Econ. 90, 197–217.
Bianchi, M., Bianco, M., 2008. L'evoluzione della corporate governance in Italia: Meno piramidi più coalizioni?. Il Mulino, Bologna.
Bianco, M., 2003. L' industria italiana. Il Mulino, Bologna.
Bianco, M., Jappelli, T., Pagano, M., 2005. Courts and banks: effect of judicial costs on credit market performance. J. Money Credit Bank. 37 (2), 223–244.
Bolton, P., Von-Thadden, E.L., 1998. Blocks, liquidity, and corporate control. J. Finance 53 (1), 1–25.
Bushee, B., 1998. The influence of institutional investors on myopic R&D investment behavior. Account. Rev. 73 (3), 305–353.
Carpenter, R., Petersen, B., 2002. Capital market imperfections, high-tech investment, and new equity financing. Econ. J. 112 (F54–F72), 54–72.
Caselli, F., Gennaioli, N., 2013. Dynastic management. Econ. Inq. 51 (1), 971–996.
Claessens, S., 1997. Corporate governance and equity prices: evidence from the Czech and Slovak republics. J. Finance 52 (4), 1641–1658.
Claessens, S., Djankov, S., 1999. Ownership concentration and corporate performance in the Czech Republic. J. Comp. Econ. 27 (3), 498–513.
Claessens, S., Djankov, S., Fan, J., Lang, L., 2002. Disentangling the incentive and entrenchment effects of large shareholdings. J. Finance 57 (6), 274–2771.
Cohen, W., 1995. Empirical studies of innovative activity. In: Stoneman, P. (Ed.), Handbook of the Economics of Innovation and Technological Change,
Blackwell Publishers, Oxford.
Cohen, W., Klepper, S., 1996. Firm size and the nature of innovation within industries: the case of process and product R&D. Rev. Econ. Stat. 78 (2), 232–243.
Colombo, U., Lanzavecchia, G., 1996. L'industria senza ricerca. Nuova Civiltà Macch. 3–4, 96–121.
Confindustria (Italian Confederation of Industry), 2003. Mappa delle competenze scientifico-tecnologiche, Rome.
Cosh, A., Cumming, D., Hughes, A., 2009. Outside enterpreneurial capital. Econ. J. 119 (540), 1494–1533.
Cumming, D., Zambelli, S., 2013. Private equity performance under extreme regulation. J. Bank. Finance 37 (5), 1508–1523.
Cumming, D., Zambelli, S., 2010. Illegal buyouts. J. Bank. Finance 34 (2), 441–456.
Eng, L., Shackell, M., 2001. The impact of long-run performance plans and institutional ownership for firms research and development expenditures. J.
Account. Audit. Finance 16 (2), 117–139.
Goodacre, A., Tonks, I., 1995. Finance and technological change. In: Stoneman, P. (Ed.), Handbook of the Economics of Innovation and Technological Change,
Blackwell, Oxford, pp. 298–341.
Guiso, L., Sapienza, P., Zingales, L., 2003. The Cost of Banking Regulation. Unpublished Working Paper. Chicago Graduate School of Business.
Guiso, L., Sapienza, P., Zingales, L., 2004. Does local financial development matter? Q. J. Econ. 119 (3), 929–969.
Guiso, L., Jappelli, T., 2005. Awareness and stock market participation. Rev. Finance 9, 537–567.
Hall, B., 2005. The financing of innovation. In: Shane, S. (Ed.), Handbook of Technology Management, Blackwell Publishers, Oxford.
Hall, B., Khan, B., 2003. Adoption of new technology. In: Jones, D.C. (Ed.), New Economy Handbook, Elsevier Science, Amsterdam, pp. 230–251.
Herrera, A.M., Minetti, R., 2007. Informed finance and technological change: evidence from credit relationships. J. Financ. Econ. 83 (1), 223–269.
Istat, 2013. Primo Studio Sulla Competitivita' dei Settori Produttivi. Istituto Nazionale di Statistica Italiano, Rome.
Jensen, M.C., Meckling, W., 1976. Theory of the firm: managerial behavior, agency costs, and capital structure. J. Financ. Econ. 3 (4), 305–360.
La Porta, R., Lopez-de-Silanes, F., Shleifer, A., 1999. Corporate ownership around the world. J. Finance 54 (2), 471–517.
Manso, G., 2011. Motivating innovation. J. Finance 66 (5), 1823–1860.
McConnell, J., Servaes, H., 1990. Additional evidence on equity ownership and corporate value. J. Financ. Econ. 27 (2), 595–612.
Minetti, R., 2011. Informed finance and technological conservatism. Rev. Finance 15 (3), 633–692.
Morck, R., Shleifer, A., Vishny, R., 1988. Management ownership and market valuation: an empirical analysis. J. Financ. Econ. 20 (1–2), 293–315.
Myers, S., 1984. The Capital structure puzzle. J. Finance 39 (3), 575–592.
Myers, S., Majluf, N.S., 1984. Corporate financing and investment decisions when firms have information investors do not have. J. Financ. Econ. 13, 187–221.
Myers, S., Rajan, R., 1998. The paradox of liquidity. Q. J. Econ. 113 (3), 733–771.
Newey, W., 1987. Efficient estimation of limited dependent variable models with endogenous explanatory variables. J. Econom. 36 (3), 231–250.
OECD, 2003. OECD Science, Technology and Industry Scoreboard 2003, Paris.
OECD, 2010. Measuring Innovation: A New Perspective, Paris.
OECD, 2014a. Report on Effective Approaches to Support Implementation of the G20/OECD High-Level Principles on Long-Term Investment Financing by
Institutional Investors (Annex). OECD, Paris.
OECD, 2014b. Corporate Governance Factbook, Paris.
OECD, 2014c. Entrepreneurship at a Glance, Paris.
Onida, F., 2004. Se il piccolo non cresce. Il Mulino, Bologna.
Pagano, M., Röell, A., 1998. The choice of stock ownership structure: agency costs, monitoring, and the decision to go public. Q. J. Econ. 113 (1), 187–225.
Parisi, M.L., Schiantarelli, F., Sembenelli, A., 2006. Productivity, innovation and R&D: micro evidence for Italy. Eur. Econ. Rev. 50 (8), 2037–2061.
Petersen, M., Rajan, R., 1994. The benefits of firm-creditor relationships: evidence from small business data. J. Finance 49 (1), 3–37.
Rajan, R., 1992. Insiders and outsiders: the choice between informed and arm's length debt. J. Finance 47 (4), 1367–1400.
Rajan, R., Zingales, L., 2001. Financial systems, industrial structure, and growth. Oxf. Rev. Econ. Policy 17 (4), 467–482.
Schivardi, F., Schneider, M., 2008. Strategic experimentation and disruptive technological change. Rev. Econ. Dyn. 11, 386–412.
Shleifer, A., Vishny, R., 1986. Large shareholders and corporate control. J. Polit. Econ. 94 (3), 461–488.
Shleifer, A., Vishny, R., 1997. A survey of corporate governance. J. Finance 52 (2), 737–783.
Shockley, R., Thakor, A., 1992. Information Content of Commitments to Lend in the Future: Theory and Evidence on the Gains from Relationship Banking.
Indiana University Discussion Paper, No. 523.
Sraer, D., Thesmar, D., 2007. Performance and behaviour of family firms: evidence from the French Stock Market. J. Eur. Econ. Assoc. 5, 709–751.
Stein, J., 1988. Takeover threats and managerial myopia. J. Polit. Econ. 96 (1), 61–80.
Stock, J.H., Yogo, M., 2005. Testing for weak instruments in linear IV regressions. In: Andrews, D.W.K., Stock, J.H. (Eds.), Identification and Inference for
Econometric Models: Essays in Honor of Thomas J. Rothenberg, Cambridge University Press, Cambridge, pp. 80–108.
Stock, J.H., Wright, J.H., Yogo, M., 2002. A survey of weak instruments and weak identification in generalized method of moments. J. Bus. Econ. Stat. 20 (4),
518–529.
Stultz, R., 1988. Managerial control of voting rights: financing policies and the market for corporate control. J. Financ. Econ. 20 (1–2), 25–54.
Tellis, G., Prabhu, J., Chandy, R.K., 2009. Radical innovation in firms across nations: the preeminence of corporate culture. J. Mark. 73 (1), 3–23.
The Economist, 2012. The Big Engine That Couldn't. May issue.
Wooldridge, J.M., 2002. Econometric Analysis of Cross Section and Panel Data. MIT Press, Cambridge.
Zingales, L., 1995. What determines the value of corporate votes? Q. J. Econ. 110 (4), 1047–1073.
Zwiebel, J., 1995. Block investment and partial benefits of corporate control. Rev. Econ. Stud. 62 (2), 161–185.

1 s2.0 S0014292115001452 Main

Uploaded by

Document Information

Original Description:

Original Title

Copyright

Available Formats

Share this document

Share or Embed Document

Sharing Options

Did you find this document useful?

Is this content inappropriate?

Copyright:

Available Formats

1 s2.0 S0014292115001452 Main

Uploaded by

Copyright:

Available Formats

European Economic Review 80 (2015) 165–193

Contents lists available at ScienceDirect

European Economic Review

Ownership structure, governance, and innovation

3. Data and empirical strategy

3.1. Institutional background

3.2. The empirical model

y ¼ xa1 þz1 d11 þ u1 ; ð2Þ

3.3. Data description

3.4. Measures of innovation and ownership structure

R. Minetti et al. / European Economic Review 80 (2015) 165–193

R. Minetti et al. / European Economic Review 80 (2015) 165–193

3.6. Control variables

4. The role of ownership structure

Panel A: Tests of difference of means

Ownership concentration (main shareholder quota) Owner type

425% r 25% t-test Family Non-family t-test

Product innovation 0.395 0.344 4.860 0.369 0.444 8.810

450% r 50% t-test Fin. Instit. Non-ﬁn. Inst t-test

Product innovation 0.425 0.356 9.206 0.516 0.374 11.241

475% r 75% t-test

Product innovation 0.444 0.365 9.494

Panel B: Tests of difference of medians

Ownership concentration (main shareholder quota) Owner type

425% (Pearson chi-sq.) p-value Family (Pearson chi-sq.) p-value

Product innovation 22.673 0.000 79.587 0.000

450% (Pearson chi-sq.) p-value Financial Institution (Pearson chi-sq.) p-value

Product innovation 84.886 0.000 132.229 0.000

475% (Pearson chi-sq.) p-value

Product innovation 92.170 0.000

4.1. Main results

(1) (2) (3) (4)

Main s.holder 0.071nnn 0.073nnn 0.081nnn 0.021

(1) (2) (3) (4)

Main s.holder 0.185nnn 0.198nnn 0.212nnn 0.041

(1) (2) (3) (4)

Main Shareholder quota

Branches in 1936 0.001nn 0.001nn 0.001nn 0.001nn

F statistics (Instr.) 5.78 5.98 5.58 5.83

Observations 12,113 13,257 12,893 12,130

Overid. test (p-value) 0.9774 0.9730 0.9983 0.0023 0.4251

4.2. Non-linear effects

Panel A: Ownership levels

R. Minetti et al. / European Economic Review 80 (2015) 165–193

Time and Area Y Y Y Y Y Y Y Y Y Y Y Y

Time and area Y Y Y Y Y Y Y Y Y Y

R. Minetti et al. / European Economic Review 80 (2015) 165–193

Time and Area Y Y Y Y Y Y Y Y Y Y

4.3. Changes of ownership structure

4.4. Phases of the innovation process

Panel A: Product innovation

Main s.holder 0.115 0.001 0.307 0.499 0.408 0.829

R. Minetti et al. / European Economic Review 80 (2015) 165–193

Panel B: Process innovation

Main s.holder 0. 512n 1.189 0.562nn 1.960nnn 0.879nnn 2.173nnn

(1) (2) (3) (5) (6) (7) (8) (9)

4.5. Innovation and total investment

5. Disentangling the ownership–innovation links

5.1. Agency problems

5.2. Risk and diversiﬁcation

Panel A: Product Innovation

Main Shareholder has control External Managers