Wages, Employment, and Capital in Capitalist and Worker-Owned Firms

WAGES, EMPLOYMENT, AND CAPITAL
IN CAPITALIST AND WORKER-OWNED FIRMS
JOHN PENCAVEL,LUIGI PISTAFERRI,and FABIANO SCHIVARDI*
The authors how worker-owned and capitalist differ with

investigate enterprises
respect to wages, and in Italy, the market economy with the great
employment, capital
est incidence of worker-owned and worker-managed firms. Estimates calculated using
a matched data set for the 1982-94 corroborate
employer-worker panel years largely
the implications of orthodox behavioral models of the two types of enterprise. Co-ops
had 14% lower than on more volatile and
wages capitalist enterprises, average; wages;
less volatile Given the quality of the data set the authors argue,
employment. analyzed,
these results can be as broad
regarded having generality.
paper examines how two types of A second type of firm is that owned and
This firms differ in their responses to changes managed by its workers. We label such a
in their environment and assesses whether worker-owned and
worker-managed
firm a
these firms' reactions conform to textbook worker
cooperative.
In a worker co-op, capital
models of their behavior. One type of firm is borrowed from financial intermediaries or
is the familiar proprietorship or corporation provided by the workers who act as holders
in which those who supply the firm's capital of the equity. Ultimate managerial decisions
manage the company (or select the manag in a co-op rest with the worker-owners.
ers) and enjoy the residual returns. Such The general issue iswhether firms operate
capital-owned
and
capital-managed
firms are
differently when workers have a greater voice
the dominant form of economic organization over the activities of
enterprises.
Whereas
in the
private
sector of market economies and unionized workers in a capitalist firm must
their prevalence provides the rationale for bargain with the firm's owners (or the rep
calling such economies capitalist. Workers resentatives of the firm's owners) to achieve
are hired?often those who certain the worker-owned and worker
indirectly?by goals,
provide
the firm's
capital. managed firm does not require such a step,
and thus, in co-ops, observed outcomes do
not reflect the mixed objectives of workers
*John Pencavel is Levin Professor of Economics at

information on how to access see http://
Stanford University; Luigi Pistaferri is Associate Pro the former,
fessor of Economics, also at Stanford University; and www.centraledeibilanci.com/. The authors do not have
Fabiano Schivardi is a Senior Economist at the Research exclusive access to the confidential INPS data, as they
of the Bank of Italy. The authors thank are made accessible for research purposes to a number
Department
Robert Fairlie and seminar participants at the Hoover of institutions, such as the LABORatorio Revelli-Center
Institution for helpful comments on an earlier version for Employment Studies in Turin, the Ministry of Wel
of this paper. fare, and the Bank of Italy itself. Interested researchers
The data in this research are partly proprietary (the should contact one of these institutions to set up a joint
Centrale dei Bilanci data) and partly confidential (the research project. The authors thank the INPS for mak
Instituto Nazionale d?lia Previdenza Sociale data). For ing the workers' data available to them.
Industrial and Labor Relations Review, Vol. 60, No. 1 (October 2006). ? by Cornell University.
0019-7939/00/6001 $01.00
23
Sage Publications, Inc.
is collaborating with JSTOR to digitize, preserve, and extend access to
Industrial and Labor Relations Review
Downloaded from ilr.sagepub.com at NATIONAL UNIV SINGAPORE on June 29, 2015 ®
www.jstor.org
24 INDUSTRIALAND LABOR RELATIONSREVIEW
and management. Do firms inwhich workers in conjectures offered to explain the infre
have such ultimate voice operate differently quency of worker co-ops
as an
organizational
from firms in which this voice ismuted? form, we
examine differences in physical
These issues have been addressed in a capital inputs between capitalist and co-op
number of case studies, but in this paper we enterprises. There is a growing literature
move beyond case studies to an analysis of a using
micro-economic data to examine the
large number of capitalist firms and worker implications of conventional firms' profit
co-ops in Italy, the country with the most ex maximizing behavior (for instance, Bond
tensive worker co-ops of any modern market and Van Reenan 2006) and our research is
economy1 There are more than 38,000 worker in this vein.
co-ops (including those in agriculture) in Italy
(see Ammirato 1996) and their conspicuous I.
role in the economy has made them a subject Income-Maximizing
Models of the Capitalist Firm
of frequent study by economists (such as and Worker Co-op Compared
Bartlett et al. 1992, Jones and Svejnar 1985,
and Smith 1994), though never with as rich a This lays out and contrasts the main
section
data set as is compiled here. In this paper, we of economists' basic models of
implications
combine two one of and
surveys, workplaces the capitalist firm and the co-op. Many dif
another of workers, that permit us to follow ferent models have been proposed for both
the same firms and workers over time. Our we
capitalist firms and co-ops, but restrict
pooling of 13 years' data (1982-94) resulted ourselves to the canonical models from which
in observations on about worker co
2,000 other models take their reference. These
ops, which we compare with observations on models are and
elementary, perhaps naive,
over 150,000 capitalist workplaces. We also of the complexities of these
ignore many
have usable observations on an of
average firms and their environments. However, they
almost 13,000 individual workers per year, the workhorses for much economic
provide
some 2-4% of them working in co-ops. These
analysis and it is important to determine their
surveys allow us to match workers with their empirical relevance. Do the implications of
workplaces, and the panel feature of the data the orthodox models of the co-op and capi
means we control for invariant attributes of to the observations
talist conform
enterprise
firms and of individuals that are intrinsically we have collected? In particular, we focus
unobserved researchers and that we on to wages
by may labor market outcomes
relating
treat as fixed effects. In all, in these matched and employment and on the use of capital
we have the
by these firms.
worker-workplace panel data,
richest set of observations on worker co-ops
ever examined economists.
by
Our principal goal is to describe empirical The Capitalist Firm
regularities regarding wages and employment
in co-op and capitalist workplaces within a First consider the prototypical capitalist
firm as out in micro-economic text
framework suggested by textbook models of spelled
these organizations. In addition, in view of the books, with E denoting the firm's employ
on capital market issues ment of labor and K its services from physical
importance placed
capital. The price per unit of output is p,
the wage paid to each labor input is w, r is
the rental price of capital, and F denotes the
of well-known case studies include the
Subjects firm's fixed (that is, unavoidable
in the Basque of Spain contractual)
Mondragon group country
costs. In the classical case, these
(Whyte and Whyte 1991), plywood co-ops in the Pacific monetary
Northwest (Pencavel 2001), sanitation workers in the w, r, and F?are to the firm
values?p, given
San Francisco Bay Area (Russell 1985), and kibbutzim and define its economic environment. If the
in Israel (Warhurst 1999). Each of these organization
firm does not face a given value of p, then p
types differs from the others, of course, but all share
the feature that many of the workers "own" and "man
stands for the state of product market condi
the firm. tions, in
age" general.
Downloaded from ilr.sagepub.com at NATIONAL UNIV SINGAPORE on June 29, 2015

CAPITALISTAND WORKER-OWNED FIRMS 25
A price-taking firm faces a given wage that gets back what he put in (less the admission
is independent of p, r, and F. As we shall fee) and what he receives is independent
note below, this property is not shared with of the future stream of the
enterprise's
net
the standard model of the co-op. Treating returns. Hence, an
unregulated
market for
w as fixed in this firm's optimizing problem tradable co-op membership rights does not
does not necessarily require that the firm's exist in Italy.2
labor supply function be horizontal; it follows In the second model of the worker co-op
also if collective bargaining at the industry and the one investigated here, the co-op's
or regional or national level (as is typical in membership rules are specified by the legal
Italy) sets the wage that the capitalist firm system and, taking these rules as given, the
must is assumed to maximize income
pay. co-op per
The capitalist firm selects its inputs to worker-member so that members have
equal
maximize net revenues, and the resulting
shares. In this case, the levels of all inputs,
labor demand and capital services functions including labor, E, and capital, K, are selected
may be written as E = En(p, w, r) and K = to maximize
per-worker
net revenues, v. The
Kn(p, w, r). Normally, the labor demand resulting employment and capital services
function has the features that E rises with demand functions for the worker co-op do
an increase in p and E falls with an increase not contain v as a
predetermined
variable
in w. Analogously, according to the capital and may be written E = Ev(p, r, F) and K =
services demand function, K rises with an IC(p, %F). For the worker co-op, increases
increase in p and K falls with an increase in in p do not necessarily increase the demand
r. Changes in F do not induce alterations in for workers.3 Unlike the profit-maximizing
the optimal levels of E and K. capitalist firm, the worker co-op increases E
in response to an increase in F
The services demand func
The Worker co-op's capital
Co-op tion few un
possesses qualitative properties
Consider the twin worker-owned firm in less restrictions are placed on the form of
the same market environment with the same the production function. However, if labor
production function as the capitalist firm. An and capital are complementary inputs in
issue deferred to future research iswhether production so that the marginal product
the internal operations (the production func of one input rises when more of the other
tions) of co-ops are less or more efficient than
those of capitalist enterprises. Though there
are variations, two models of 2The absence of a "thick" market in membership
many principal
the co-op be entertained. rights may not be the consequence of the legal struc
might
ture; on the contrary, the legal structure may reflect
One model assumes that there is a market
the deeper problems in markets for co-op member
for co-op membership and that the prices of in a capitalist firm, in a co-op the right to
ship. Unlike
each shares fluctuate in response to a portion of net receipts entails not only ownership of
co-op's
in about an the share but also the individual's obligation to supply
changes expectations enterprise's
his labor to the organization, so the exchange of a co
fortunes (Sertel 1982; Fehr 1993). A draw one individual's
back of this model for our purposes is that op's share involves substituting labor
supply for another's. Hence factors that inhibit labor
there is little evidence of such markets for contribute to a "thin" market for co-op mem
mobility
membership in the Italian case. There are bership shares. In these circumstances, competition
admission fees to most Italian co-ops and for co-op membership shares may be less common than
bilateral so the legal system may be more
each member is required to contribute some monopoly,
inclined to regulate the market for membership rights.
of the enterprise's financial capital. How 3As is well known, if labor is the only so
input,
ever, these contributions do not constitute that r is not an argument of the employment demand
market membership then increases in p reduce
prices because Italian function, employment (and
so-called This inverse
law specifies both minimum and maximum output?the "perverse supply").
amounts that new members must
relation between p and E is no longer an unambiguous
advance
implication of the model of the income-maximizing
to their leav
co-ops. Furthermore, upon co-op once output depends on other
inputs in addi
ing the enterprise, normally the individual tion to labor.

input is used, then the co-op's demand for In fact, labor supply functions to the co-op
falls when r rises and its demand for may prevent attainment of the maximum
capital
are higher. in the
capital rises when fixed costs (See per-worker income described previous
Pencavel and Craig 1994.) subsection. Once labor supply considerations
If the co-op's optimal values of labor, capi become relevant, the
non-pecuniary employ
tal, and other inputs are substituted into its ment conditions in the two types of firms
objective function, the maximized value of need to be assessed. For instance, insofar
v?call this v*?may be written indirectly as as workers
place
a value on
participating
in
a function of the predetermined variables: the co-op's governance, they will be willing
v*= r, F). Thus, because each worker's to work at wages different from those prevail
?{p,
earnings
are i/*, when firms are
price-takers ing in the capitalist enterprise. Or, as noted
in all markets, changes in these prices induce above, the model of the income-maximizing
in the of each worker. co-op that the wages
changes earnings co-op implies co-op's respond
contrast, in the same market environment, to firm-specific prices and fixed costs and,
By
the wages of the worker in a capitalist firm therefore, they tend to fluctuate more than
are independent of firm-specific changes in wages in capitalist firms. Other things
market
prices.
For these reasons, the wages
equal, these greater wage fluctuations have
of co-op workers are
expected
to be more the effect of mitigating employment varia
variable than the wages of workers in capitalist tions. Consequently, employment loss may
firms. More specifically, the maximized value be viewed as less of a risk by co-op workers
of v is higher when p rises and v is lower than by workers in capitalist firms and co-op
when r or F falls. Thus, the co-op worker's workers may, accordingly, be willing to accept
earnings are predicted to vary with the firm's lower wages than workers in
capitalist
firms.
product market conditions (as embodied in These considerations frustrate clear

predic
p) and with the firm's input prices and its tions about the
sign
of the
capitalist-co-op
fixed costs. wage differential.
Figure 1 graphs the difference between Also possibly affecting these implications
employment in the capitalist and the co-op about capitalist-coop wage differentials is
firms. For the capitalist firm, employment is the presence of unionism. Most Italian
set at E{w), where the value of the marginal workers are covered by
collective bargain
that apply to all firms in a
product of labor (here given by p.AIPE) equals ing agreements
the wage,
w. For the income-maximiz
given industry or region. Typically these
given
is set at E(v), where minimum rates of pay,
ing co-op, employment agreements stipulate
the value of the marginal product of labor though plant-level bargaining may result in
income worker, v, which in turn at a
equals per wages particular workplace "drifting"
revenues worker, R/E. Because above this floor. In other instances, espe
equals per
the maximized value of v, v*, is no less than
cially in small establishments, the terms of
w, employment at the capitalist firm is not collective bargaining agreements may be
less than employment in the co-op. If the disregarded. Hence actual pay differentials
firm's wage, w, exceeded revenues across establishments in the same industry
capitalist
per worker, R/E, the capitalist firm would are greater than those specified in the multi
earn negative cease pro
profits and would employer agreements.
duction. In the period of our study, co-ops were not
legally required to adhere to the minimum
rates of pay set down in industry-wide union
Other Issues Bearing
contracts, although a number of them did.
on Wage Differentials
While in many cases workers in co-ops are af
This familiar implication that, other things filiated with the relevant industrial union, the
equal, the earnings of the co-op worker will co-op workers tend to display less militancy
exceed the earnings of the worker in the than workers in capitalist firms. This has
enterprise
assumes the absence of caused "frustration [among union leaders]
capitalist
labor constraints on the that workers must, of necessity,
any supply co-op. co-operative

Figure 1. Employment Differences between a Capitalist Firm and a Co-op.
E(v) E(w) E
out of an active role in trade union af and the use

opt concerning employment, wages,
fairs in favor of the struggle to make their of capital. These are our that
hypotheses
enterprises commercially viable" (Thornley our data
subsequently permit
us to address.
1983:332). The ability of co-ops to ignore First, with respect to wages,

union-negotiated pay scales and the less In the the wage is inde
militant posture of co-op workers lead to the capitalist enterprise,
pendent of the values of
that in lie below firm-specific product
conjecture wages co-ops may and fixedcosts.
prices, capital prices, By contrast,
those in capitalist workplaces. in the co-op, the wage on the
depends co-op's
The upshot of this discussion is that and fixed costs. In
product prices, capital prices,
theoretical reasoning cannot predict whether the wages of co-op workers are lower
particular,
wages are higher or lower in the capitalist when the prices of
capital
are
higher and when
firm than in its co-op twin. fixed costs are
higher.
The wage differential between the
capital
Summary Statement ist enterprise and the co-op may be
positive
or
negative, and research is to

empirical required
This exposition of the conventional models determine its value.
of the profit-maximizing capitalist firm and
Second, with respect to employment,
the worker
income-maximizing co-op sug
gests the following differences between them In the
capitalist firm, employment is lower when

28 INDUSTRIALAND LABORRELATIONSREVIEW
the wage rate is higher. In the co-op, employ logical approval, the cooperative movement
ment are determined variables
grew such that by 1921 some 20,000 co-ops
and wages jointly
and causation does not run from wages
simply were operating (Ammirato 1996:78). The
to The of the correlation
employment. sign
co-ops suffered during the years of Fascist
between wages and employment among co-ops but after the overthrow of Mus
is unspecified. government,
solini they recovered and the state reverted
In the firm, is indepen
capitalist employment to them various favors.4
costs. is granting
dent of fixed In the co-op, employment In the 1970s co-ops secured further tax
when fixed costs are
higher greater.
advantages, including exemption from all
will tend to be more and taxes on channeled to the reserves.
Wages responsive surpluses
less to market
employment responsive product Public works projects (transport, housing,
shocks in the in the capitalist
than firm.
co-op and schools) were steered toward co-ops and,
This is because in the capitalist firm the enterprise
to counter unemployment among young
takes wages as in which case demand
given product the state provided financial support
shocks (changes in output prices) induce the firm people,
to ventures. In
to make youths starting cooperative
quantity?including employment?reac
tions. Thus, in the capitalist firm, 1977-79, almost 1,300 co-ops were formed
employment
in this manner
(and other input quantities) will take the brunt (Ammirato 1996:95).
of product market shocks to the firm's economic The environment of the 1980s was less
environment, while in the co-op, such shocks will to the reduced
hospitable co-ops. Facing
have a first-order on each member's wages. financial as the state strove to cor
impact support
These effects on the member's will rect its large fiscal deficits,
co-op wages
co-ops modified
temper employment responses. their vision. They tried to shed their image
with to the use of physical of culture and aversion to
Third, respect working-class pri
vate enterprise and they adopted an attitude
capital,
favorable to economic efficiency and financial
In the
capitalist firm, the use of capital falls when
incentives. members were
the price of capital rises, but the use of capital is Co-op encouraged
of fixed costs. if
to view their organization as purely a financial
independent By contrast, capital
are investment and incentives were
and labor complementary inputs (defined monetary
above), the co-op's demand for capital falls when provided to bolster this view. Increasingly,
the price of capital rises and the use of capital with a business
supervisors practical mentality
increases when fixed costs increase. were hired into the co-ops from outside the
Before the relevance co-op movement. The Census data in Table 1

examining empirical
indicate that, over the period of our empirical
of these implications, consider the following
of the Italian movement. analysis from the early 1980s to the mid-1990s,
description co-op
worker co-ops grew in Nevertheless,
Italy.
by 1996, employment in co-ops constituted
II. Worker Co-ops in Italy
only 4% of total employment, with greater
in construction,
Brief History representation transport,
and some service industries.
The cooperative movement in Italy dates
back tomutual aid and insurance societies in Structure of
Co-ops
the nineteenth A confedera
century. co-op
According to Italian law, each co-op may
tion was established in 1886 that later became
La Lega (Nazionale delle Cooperative), the
umbrella of co-ops 4The Basevi Law of 1947 granted co-ops tax benefits,
largest organization today. on the
the amount of which depended importance of
The confederation's function was
principal labor costs in total costs. At maximum, a co-op was ex
to act as a on
pressure group government, empted from paying all taxes if labor costs represented
and, indeed, was secured 60% or more (reduced to 50% or more in 1992) of
legislation grant
from certain taxes and total costs (excluding raw material costs). The law also
ing co-ops exemption
other financial of co-ops,
them with preferential access to regulated aspects including
providing the amount each worker could invest in the organiza
public works projects. Favored in this way tion and the maximum interest he could earn on this
by the state and enjoying widespread ideo These limits were eased in 1971.
capital.

Table 1. Percentage of Co-Ops and of Co-Op Workers 1981-96.

by Industry,
Co-ops as a Percentage of All Establishments Co-op Workers as a Percentage of All Workers

Sector
1981 1991 1996 1981 1991 1996
1.14
Mining 2.63 1.16 2.15 3.19 2.56
Manufacturing 1.02 0.98
1.17 1.51 2.11 1.68
Utilities 6.08 6.35
6.28 1.04 0.25 0.27
Construction 0.74 1.53 2.43 4.93 4.61 4.26
Retail
0.27 0.55 0.27 1.44 2.57 2.01
Hotels
0.48 0.67 0.62 1.19 2.16 2.57
1.15
Transport 1.96 2.06 6.55 5.67 8.83
Financial Services 3.29 2.90 1.90 10.09 14.4012.90
IT & Real Estate 1.84 1.51 1.16 5.92 10.40 12.74
Other Services 0.49 0.81 1.52 2.41 3.02 3.63
All Sectors 0.61 1.02 1.06 2.50 3.66 4.06

Source: These data are taken from the Italian National Institute of Statistics Census of Industry and Services
conducted in 1981, 1991, and 1996.
determine its own rules of association and between worker-members and hired workers
by-laws, although statute law specifies certain who are not members is unlikely to be impor
key features. Co-op membership requires an tant. Non-members include
apprentices
and
individual to provide a portion of the co-op's probationary workers. Most co-ops distribute
capital. Sometimes an admission fee may profits
to members and non-members on the
also be required. A member who leaves the same terms, but there may be bonuses that
co-op is repaid the value of his contributed members enjoy. Only members may loan to
capital. A co-op must allocate at least 20% their co-ops and the interest paid on the loan
of each
year's
net revenues to its reserves. is both tax-free and higher than the interest
Profits into these funds are offered
paid exempt by conventional banks. Following
from income tax. Net revenues in an adverse shock, non-members are more
corporate
excess of that paid into reserves may be dis to lose their jobs. When
likely than members
tributed to working members in proportion members lose their jobs, the co-op will en
to their work, and dividends may be paid on courage another co-op to hire them. Co-ops
capital provided. Italian law specifies limits are not covered by employment protection
to distributions in the form of dividends. legislation,
so laid-off co-op workers are not
a member-worker receive on his entitled to severance
Usually, may pay.5
contributed capital no more than the rate By vote, the General Assembly selects a
paid on Treasury or Postal bonds. Council, the co-op's principal supervisory
Decisions about the distribution of net body. It appoints the managers and speci
revenues are made each General fies one
by co-op's general policies. Usually, manage
Assembly. Each member has one and only one rial position, the President, is elected, not
vote. Almost all co-op workers are members.
appointed, by the Council. Descriptions
Thus, in the survey of 49 co-ops in Tuscany of co-ops suggest extensive participation
and Emilia-Romagna undertaken by Bartlett by co-op members in decision-making and
et al. (1992) in the mid-1980s, on average
about 85% of workers were members: "In
most co-ops three of the work 5This description to the of our data
Lega quarters applies period
ers are members and the rest will be. When analysis, namely, 1982 to 1994. However,
from several
someone becomes eligible for membership, subsequent judicial and legislative decisions changed
a number of the rules. in 1998
a councillor or senior member invites him Legislative changes
that co-op workers are covered
to apply" (Holmstr?m specified by most of the
1989:91). As a con collective rules and employment
bargaining protection
sequence, for most purposes, the distinction legislation that apply to workers in capitalist firms.

considerable turnover of officers (Holm The Survey of Companies

str?m 1989).
Often basic pay scales in co-ops are those The company data are drawn from the
negotiated by the national union and the archives of the Italian Company Accounts
relevant confederation of employers at the Data Service, which collects balance sheet
regional or national level, but supplementary information and other items on over 30,000
payments,
as well as
promotions, working
Italian firms. Available since 1982, the data
conditions, and are determined are gathered by Centrale dei Bilanci, an
grievances,
inside the enterprise. Earnings may therefore organization established in the early 1980s
exceed the regionally negotiated or nationally jointly by the Bank of Italy, the Italian Bank
negotiated minima and there may be consid ing Association (ABI), and a pool of leading
erable differences in pay across firms in the banks to construct and share information
same or When tend on borrowers. Besides balance sheet items,
industry region. wages
to lie above those set down inmulti-employer there is information on the level of employ
collective bargaining agreements, it is dif ment, on the establishment's flow of funds,
ficult to gauge the extent to which the wage and on various attributes of the firm such as
structure diverges from that which would the year founded, its location, its legal form
obtain in the absence of unionism. of organization, its ownership status, and its
Within the co-op, the union may have a control structure. The entry on the legal
well-defined role in serving the interests of form of organization allows the identifica
workers who are not members, tion of co-ops.
although,
these constitute a small minority of Balancesheets for the banks' major clients
usually,
a work force and, in any event, most (defined according to the level of borrow
co-op's
non-members to and will become ing) are collected by the banks. The focus
aspire
members. The union may also have a role if on the level of borrowing skews the sample
the
co-op's representation
mechanisms are toward larger firms. The aim of increasing
not functioning well and a group of workers the
representation
of medium-sized firms,
feel their interests are being slighted. How however, resulted in an increase in the
sample
ever, a basic co-op principle is that workers size in 1993. Consequently, since 1993,
in the more from the middle of
participate enterprise's governance, companies range
and if the mechanisms for worker the size distribution are included. Given
participa
tion are functioning as designed, it is not that our method is not based on time series
clear what constructive role can be
played comparisons and we invariably include time
within the co-op by
a union. A union's role dummy variables in our statistical analysis,
as
watchdog
of management seems anoma this change in sampling in 1993 and 1994
lous when the workers select and have the should not affect our results.
authority to dismiss the managers. (See Because most of the leading banks are in
Holmstr?m 1989:89-108; Thornley 1983; the northern part of the country, the sample
Earle 1986.) has more firms headquartered in the North
than in the South. Also, banks deal mainly
with firms with a good history of borrowing,
III. Two Sources of Data
so the sample is tilted toward the more cred
types of data are exploited
Two in this itworthy firms. Even with these restrictions,
paper. One is an annual survey of companies these data provide a very comprehensive
and the other is an annual survey of workers. description of Italian firms. The firms in
The survey of workers is based on Social Se cluded in the sample (which excludes the
records. Because these records cover
sector) account for more than 50%
curity public
a number of years and include employer tax of total employment in Italy (Cingano and
codes, they allow for the construction of a Schivardi2004).
matched employer-worker longitudinal data Descriptive statistics on the firm sample
set. Consider first the information contained for three
representative years
are
provided
in the company survey. in Table 2, which restricts the data to the

Table 2. Statistics on Matched

Descriptive Companies: Company Sample.
1982 1988 1994
Variable Capitalist Co-op Capitalist Co-op Capitalist Co-op
No. of Companies 4,797 57 6,850 104
7,836 240
Average Employment 301 373 235 228

284 341
Median Employment 103 256 199 79153
72
Average Wages 0.0198 0.0245 0.0230 0.0293 0.0293 0.0520
Median Wages 0.0189 0.0197 0.0221 0.0233 0.0272 0.0260
Average Capital 9.52 4.23 8.92 6.80

16.32 6.60
Median Capital 1.48 2.37 1.18 2.89
1.52
1.29
Average Sales 38.43 29.42 36.03 50.03 45.55 40.24

Median Sales 8.77 17.65 9.00 18.45 10.95
9.92
Average User Cost of Capital 0.193 0.223 0.178 0.151

0.205 0.148
Median User Cost of Capital 0.180 0.197 0.170 0.193 0.146 0.146
Average Fixed Costs 0.0131 0.0091 0.0058 0.0033 0.0083 0.0043

Median Fixed Costs 0.0010 0.0023 0.0005 0.0013 0.0004 0.0005
Average Product Market Shock 0.072 -0.030 0.049 0.037 0.026 0.057
Fraction Manufacturing 0.772 0.157 0.740 0.307 0.671 0.246
Fraction Construction 0.066 0.368 0.066 0.240 0.060
0.133
Fraction Retail 0.113 0.158 0.132 0.2210.166 0.263
Fraction inNorth 0.737 0.982 0.721 0.807 0.732 0.708
Fraction in Center 0.157 0.000 0.164 0.135 0.159 0.208
Fraction in 0.106 0.083
South_ 0.018_0.115 0.057_0.109
Notes: The matched company includes the with one or more observations in the workers'
sample only companies
data set for the reference years. Wages are defined as annual labor costs plus dividends distributed to workers in co
ops, all divided by employment. All monetary variables are measured in billions of lira except for fixed costs, which
are measured in hundreds of billions of lira. All monetary variables are deflated
by the producer price index (base
1982). Companies are into four industry categories: construction, retail, and other.
organized manufacturing,
Source: Italian Company Accounts Data.
matched firms, that is, those companies fact, total annual labor costs per worker, and
for which observations on workers are also
although these labor costs consist principally
available. The share of co-ops in our
sample
of labor earnings, they also include taxes on
is comparable with that of the population labor. For the co-ops, we include in wages
up to 1992: according to the Census, the the dividends distributed toworkers. Usually,
share of all was these were in co-ops. However,
co-ops among enterprises wages higher
0.61% in 1981 and 1.02% in 1991; in our these
comparisons
hold
nothing
constant.
sample,
the share of co-ops
was 1.17% in The industrial distribution of capitalist and
1982 and 1.50% in 1988. The share of co co-op firms was (and remains) quite different
ops in our sample rose to 2.97% in 1994, the and this alone will give rise to wage differ
in the sample ences: the capitalist firms in our
consequence of the change sample were
design in 1993 described above. Average concentrated in
manufacturing,
whereas
and median employment in the co-ops relatively
more
co-ops
were in construction
tended to be larger than in the capitalist and the retail industry.
firms. The empirical frequency distribu Capital (measured in billions of 1982
tion of companies by employment reveals lira) consists of the value of land, buildings,
that there were a few capitalist firms with office equipment, machinery, vehicles, and
extremely high levels of employment, but patents. According to Table 2, usually the
these were offset by a large number of very average stock of capital was greater in the
small capitalist workplaces. capitalist enterprises than in the co-ops,
What is called "wages" in Table 2 is, in while the median stock of capital was higher

Figure 2. Frequency Distribution of Capital-Labor Ratios for Capitalist and Co-op Enterprises.
35 4.
?.- ? .. ^?* * --* i-4.

30 4. mi
5
mm i.
?*** *;%*
254. mm
* mm*^mfc*^-+^-*'*.m?.
204
15 J.* ? ,*
' '
<4U2 filil?'-.:;1""" ?:<i??
Notes: The horizontal axis corresponds to deviations in capital-labor ratios from their industry mean values. The
"< -0.08" contains observations -0.09 and -0.08). The lowest category
categories represent intervals of 0.01 (so between
is < 0.12 and the highest category is > 0.11.
in the co-ops. However, these capital stock According to Figure 2, capital-labor ratios
comparisons reflect in part the different in co-ops display a wider dispersion than
technologies across industries, those in capitalist enterprises: there were
production
with capitalist enterprises more heavily rep relatively more co-ops with very low capital
resented in manufacturing and co-ops in labor ratios (at the left tail of the distribu
retail industry and in construction. To assess tion), but also relatively more co-ops with
whether co-ops were relatively undercapital very high capital-labor ratios. According to
ized and whether input combinations were these data, input ratios in co-ops differed
systematically different across the two types from those in capitalist firms, but the dif
of firms, we calculate the ratio of capital to ferences are not well described by simply
employment for each firm and average it characterizing co-ops as using less capital
over all years observed. Then we regress relative to labor.
this average capital-labor ratio on industry In our fitted equations below, we make use
dummy variables to remove the effects of of the rental price of capital and of fixed costs.
industry-specific technologies. Finally, in For the former, we define a comprehensive
Figur? 2 we plot the frequency distribution measure of the user cost of capital for enter
of the residuals from this regression sepa prise j in year t that takes account of taxes,
rately for co-ops and for capitalist firms. investment credits, depreciation allowances,

financial debt, depreciation, and each firm's regress the logarithm of real sales on a firm
reported borrowing rate.6 This unusually fixed effect and a linear time trend and allow
precise
measure of the user cost of
capital,
this trend to be different across industries and
denoted has a mean value of 0.197. across of the Then we mea
rjt, regions country.
To measure fixed costs, F-t,
we use each sure the product market shock for each firm
company's long-term debt payments in each in each year by the deviation of the logarithm
year (measured in real terms). These long of real sales from the value predicted from
term debts represent contractual payments
this regression (including the firm's fixed
the firm must make on its commitments effect) .8 In our estimating equations below,
from
previous years and
they correspond
this variable is denoted by Sp an indicator
to the unavoidable nature of fixed costs in of firm /s transitory shocks in the product
economic models of the firm. These ex market in year t; in Table 2, this variable is
penditures, which must be made regardless labeled "product market shock."
of the decisions and actions,
enterprise's
illustrate the principle that "sunk costs are The of Workers
Survey
sunk." The mean of F:t is 0.005 and its fre
TheItalian National Institute for Social
quency distribution is skewed to the right,
Security (Istituto Nazionale della Previdenza
with 25% of observations having a Fjt value
of zero. This definition of fixed costs is the Sociale) requires firms to file a yearly report
for each worker on their payroll. The data
best available in these data.7 To ensure that
are used to estimate the amount of withhold
our results with respect to other relationships
are not simply an artifact of this particular ing tax the employer has to pay on behalf of
the employees and to the INPS as contribu
way of computing fixed costs, all the equa
tions for pensions and health insurance.
tions reported below were estimated also by
measure The forms are roughly comparable to those
omitting this of
Fjt. collected the Internal Revenue Service
The argument in Section I also alluded by
in the United States. However, while U.S.
to the effects on wages and employment of
administrative data are usually provided on a
transitory product market shocks. With the
available data, such shocks for each firm in a grouped basis, the INPS has truly individual
records. Moreover, in the United States,
given year are measured by the deviation of
records are censored at the of
its sales in that year from the sales predicted earnings top
the tax bracket, while the Italian data are not
by a linear regression. More precisely, we
subject to top-coding.
The data cover all employees in the pri
6To be precise, the user cost of capital for establish
vate sector. The INPS supplies observations
ment j in year t is given by on a of the universe,
subsample namely,
= (! - - - + all those workers born on two
rjt "jtX1 V"1 igjtjP- xji> particular
days of the year. We restrict the data to
where 5 reflects tax rates, investment tax cred the period from 1982 to 1994 so they can
corporate
its, depreciation allowances, and subsidies (which vary by be paired with the establishment data. The
x is the corporate tax rate, gis the ratio of finan form information on each worker's
region), reports
cial debt to total liabilities, ?is the average borrowing rate
annual earnings and on the number of days
paid by the firm, eis the nominal return on equity (the
and weeks worked.9
on Italian Treasury bonds), n is the sector-specific
yield
expected increase in the prices of capital goods (from
the Bank of Italy survey of investment in manufactur
ing), and ? is the sector-specific rate of depreciation. 8Other of the equation
specifications imput
is defined as where (ppi) is the shocks were
(Djf.Jj)/ (ppi) ?, ing product market applied with small
^Fji to
producer price index, Dis the stock of long-term debt consequences for the estimates reported below.
all financial intermediaries, and/is the long-term nomi Of course, if were pure measurement error, it
S?
nal interest rate on borrowing. The latter is regarded should be uncorrelated with wages and employment.
as confidential information and is not available for any information on hours worked is not available,
establishment, so we use the economy-wide medium there is less variation in hours worked in Italy
though
and long-term rate on bank borrowing. Fixed costs are than in many other countries. consist of two
Wages
measured in hundreds of billions of 1982 lira. components: normal and occasional. Occasional wages

Workers are identified by broad job cate IV. Empirical Specifications

gory: apprentices, production workers, cleri and Results for Wages
cal workers, and Information on
managers. A priori reasoning in Section I could not
education is unavailable, but days and weeks a
yield prediction regarding the sign of the
of work and the date employment ended
co-op-capitalist wage differential, although
(if this occurs) are From each
provided. we surmised that, whereas wages in
capitalist
worker's social number, information
security firms are independent of variations in the
can be retrieved on the worker's
gender, year prices of non-labor of
of birth (and thus age), and place of birth. inputs (in particular,
the user cost of capital) and of fixed costs,
tax code allows us to Tp
Finally, the employer's the wages of co-op workers are lower when
match information on the worker with that Fp
r-tand Fjt are each higher. Also, the models
on the firm.
of the capitalist firm and the worker co-op
Table 3 presents information on the char
in Section I hypothesized that wages would
acteristics of workers matched with their tend to be more volatile in the co-op. That
workplaces. Workers aged less than 18 years the capitalist
is, when firm operates with
and more than 65 are as are
years omitted, a horizontal labor supply curve or when
all those observations lacking data on wages wages are fixed by multi-employer collective
and the extent of work in the year. Whereas in
bargaining agreements, wages the capital
the unit of observation in Table 2 is the
ist firm are independent of shocks to the
company, the unit of observation in Table firm's labor demand function. Such shocks
3 is the worker. According to the observa
will have first-order effects on employment,
tions on workers, and median
average wages hours, and other input quantities such as
were a little lower for co-op workers than for
contrast, in the co-op, shocks to
capital. By
workers in capitalist firms. When a simple
labor demand will be directly transmitted to
descriptive regression equation isfitted to the the co-op workers' wages.
company data and to the worker data, holding these data to address these issues,
constant and
Using
the wage dif
region industry, our observations on describe worker
wages
ference between the and the
co-ops capitalist i in company j in calendar year t. These
measured in the company data
enterprises are unbalanced data because a worker
may
and in the worker data is similar.10
not be observed in every year. If a worker
was employed by two firms in a year, there
include sums
drawn from the wage supplementation will be two observations on this worker that
fund laid-off or short-time
for workers, seniority and An that a method
year. equation provides
loyalty premia, one-time bonuses, moving expenses and
to investigate these differences in wages
business travel refunds, the monetary value of goods
in kind, and allowances for lost tips and commissions. among individuals in capitalist firms and
On average, occasional wages are less than 10% of the worker co-ops is
total. Our measure of gross wages is the sum of the two
= + + + +
components. For co-op workers, wages include dividends (1 ) lnwijt xijta zj? rjtyx Fjty2
distributed to them by their firm, where the information
+
about dividends comes from the company data set. SJtys+CijtQ0+rJtCijtQ1+FJtCtJtQ2
10More precisely, for the data on individual workers,
we the logarithm of real annual wages on indus
regressed
try dummy variables, variables, calendar
where \nw stands for the logarithm of real
regional dummy
year dummy variables, and a dummy variable taking the x are observed characteristics of the
wages,
value of unity for a co-op observation. Similarly, for the individual worker, z are observed features of
data on establishments, we the logarithm of
regressed the enterprise where the individual works,
real annual labor costs per worker on the same set of
variables. In the former the estimated coef and Cis an indicator taking the value of unity
regression
ficient on the co-op dummy was-0.083 (with an estimated when individual iworks in a co-op. r-t is the
standard error of 0.015), and in the latter regression it user cost of is fixed as
capital, F-t costs, and,
was -0.016 (with an estimated standard error of 0.007).
described in Section III, our measure of
Given the different concepts, it is natural that the two S7iis
estimates should not be identical. However, their com
observed transitory shocks to firm/s product
mon market in year t. Factors that affect wages but
sign and similar magnitude suggest that the "raw"
differences in Tables 2 and 3 conceal basic similarities. are unobserved the researcher are
by repre

Table 3. Statistics on Workers: The Matched

Descriptive Worker-Company Sample.
1982 1988 1994
Variable Capitalist Co-op Capitalist Co-op Capitalist Co-op
No. of Observations 10,154 163 13,192 304 544
14,749
Average Wages 12.106 10.480 13.877 11.883 14.708 11.892

Median Wages 11.813 10.611 13.170 12.444 13.730 11.238
Average DaysWorked 252 254 264 251 264 251
Median DaysWorked 305 302 312 311 311312
Average Weeks Worked 45
45 45 43 45
43
Average Age 38.237.7 37.7 36.5 38.0 37.1
Fraction Male 0.741 0.773 0.742 0.681 0.725 0.662
Fraction Operatives 0.706 0.644 0.654 0.507 0.610 0.642
Fraction 0.355 0.329 0.484 0.372 0.339
Clerical_0.281
are measured
Wages in millions
of lira and deflated by the consumer price index (base 1982). For co-op workers,
wages include dividends distributed by the co-op. The matched sample includes only workers for whom company
information is available for the reference years.
Source: Istituto Nazionale d?lia Previdenza Sociale.
sen ted by v and 8. Unobserved factors that be cor related with C11 Our panel data permit
vary across people but are fixed for a given us to observe workers when
they
move across
individual over time are denoted by v. This firms, so the correlation between!) and Ccan
includes schooling as well as personality traits be addressed by averaging variables over time
and habits that endure over time and that may and writing ( 1 ) as
be relevant to the labor market. 8
represents - -
across (2) +
unobserved factors that vary people \nwijt ln?y= (*.-, ^)o
in different firms and over time. - + - +
(zjt-^)? +Jrjt rpYi {Fjt Fj)y2
The arguments in Section I suggested that +
the wages of workers in capitalist firms will {Sjt-fyys+(C#-Ci/)%
tend to be insensitive to differences +
in rand Wv-rjCJOi* (F^-FjQpd,
F, whereas the wages of workers in co-ops will
(5^-^)03+ (8^-8?),
tend to be a negative function of both r and
where a bar over a variable indicates an av
F. If this is the case, Yi = Y2= 0, while 0j and
over time. The movement of workers
are erage
each Section I
02 negative. Analogously, between and firms allows us
co-ops capitalist
reasoned that transitory shocks in product
to estimate the 0's. (That is, without such
markets would tend to affect wages more in
and more in
worker mobility,
Cijt
would equal its mean
co-op enterprises employment value over the period.) This within-group
capitalist enterprises. This would suggest
specification is a familiar technique with
that 03 is positive. The wage differential be
tween workers in capitalist firms and those panel data. Inequation (2), capitalist-co-op
in co-ops depends on the estimates of 60, 0^
and 03.
02, on
Because v is unobserved, es
nThe sign of the bias depends the relationship
least-squares between wages and risk aversion. those who
Suppose
timates of 0O in equation (1) are biased if v is are less risk-averse receive higher wages (though with
correlated with C. For example, ifwages are greater variance) and suppose such individuals are
more volatile in co-ops, other more inclined to work in co-ops. Then the bias of the
things equal,
least-squares estimate of 0O that ignores unobserved
more risk-averse seek
people may employ will tend to be negative.
heterogeneity That is, if the
ment in a capitalist enterprise. If, in turn, true value of 0O is positive, the estimate of 0O that ignores
attitudes toward risk are absorbed in v, then this form of unobserved heterogeneity will be lower
the unobserved will than the true value.
equation's components

wage differentials are measured controlling In the bottom panel of Table 4, d\nw/dr
for unchanging characteristics of workers and d\nw/dF report the impact on wages of
(such as schooling and gender), although the user cost of capital and fixed costs. For
the magnitude of the partial correlation workers in capitalist firms, differences in r-t
between wages and each of the fixed char bore no
meaningful
association with wages.
acteristics of workers cannot be
computed.12 According to the least-squares estimates in
In essence, equation (2) asks, what change in columns (1) and (2), in capitalist enterprises,
wages is experienced by a worker who moves higher values of fixed costs appear to have
between a and a been associated with lower However,
co-op enterprise capitalist wages.
enterprise? this negative association appears fragile: it
The top panel of Table 4 presents the pa turns positive in column (3) and is insigni
rameter estimates of equations (1) and (2) ficantly different from zero in column (4).
and the bottom panel shows the implications The within-group estimates in column (4)
of those results. In the estimates of equa for daily wages are consistent with the argu
tion ( 1 ), in addition to the co-op dummy, Q ments in Section I.
are related to the worker's For workers, the effect of differ
wages age, gender, co-op
and location, as well ences in on is
occupation, industry, Fjt wages consistently negative.
as to the company's size (the logarithm of However, the magnitude of the effects is very
the workplace's employment), its user cost small. Consider the-0.367 value of dlnw/dFin
of capital, fixed costs, and deviations in the the first column as an example. This implies
logarithm of its real sales. The estimates of that, evaluated at the co-op median value of
equation (1), shown in the two "Ordinary 7^

in 1994 (namely, 0.0005), the elasticity of
Least-Squares" columns, suggest that, other wages with respect to fixed costs for co-ops
things equal,
a worker in a co-op earned was -0.00018, or a co-op with 10% higher
15-16% less than a worker in a capitalist long-term debts paid 0.18% lower wages to
its co-op workers. The effect of variations in
enterprise.
However, as above, this least on the wages of co-op workers is positive in
argued /y,
squares estimate may be biased if the factors the
ordinary least-squares
results in columns
affecting the choice of employment between (1) and (2) but negative (as implied by the
co-ops and capitalist firms are correlated with theoretical arguments in Section I) in the
the determinants of earnings in the two types wi thin-group estimates in columns (3) and
of firm. Equation (2) is designed to address (4). The value of dlnw/dr of -0.120 for co
this
problem
and its estimates are contained op workers in the third column of Table 4
a 10% higher user
in the columns beneath "Within-Group" in implies that a co-op with
Table 4. The estimated cost of than another would have
capitalist-co-op wage capital co-op
differential falls only slightly to 14%. These tended to payjust over 1% lower weekly wages.
estimates of the wage differential would be Thus, relatively large differences in the price
judged statistically significantly less than zero of capital were associated with relatively small
by
conventional criteria.13 differences in wages. The within-group esti
mates confirm a association between,
negative
on the one hand, differences in wages across
12What factors affecting wages are neglected in equa and, on the other hand, differences
co-ops
tions (1) and (2)? Suppose worker i ismore productive in Tuand F-p as conjectured in Section I. The
when working in enterprise j than when working in an economic of these associations,
magnitudes
other enterprise and suppose this greater productivity is
however, are not
unobserved to the researcher. Such unobserved match important.
components are not identified in these equations.

^These estimates of the capitalist-co-op wage dif
ferential (and, indeed, of the other effects in this wage is because, as mentioned earlier, the distribution of F
are sensitive to the point of evaluation. The is skewed to the right and F assumes some very high
equation)
point of evaluation for the estimates in this table is at values for some large firms. Workers in these large
sample mean values of r,F, and 5 in the workers sample. firms are observed more
frequently in the panel data
(Note that the sample mean value of Fat 0.298 ismuch and this pushes up the sample mean of F in the work
higher than the mean values reported in Table 2. This ers' data set.)

CAPITALIST AND WORKER-OWNED FIRMS 37
Table 4. Estimates of theWage Equations (1) and (2).

Ordinary Least-Squares Within-Group
(1) (2) (3) (4)

Description ln(Weekly Wages) ln(Daily Wages) ln(Weekly Wages) ln(Daily Wages)
Parameter Estimates
(Q -0.078 -0.095 0.028 0.015

Co-op
(0.019) (0.018) (0.026) (0.026)
User Cost of Capital (r) -0.011 -0.00 I -0.011 0.009
(0.012) (0.011) (0.010) (0.010)
(User Cost of Capital) *Co-op 0.107 0.110 -0.110 -0.113
(0.084) (0.079) (0.084) (0.086)
Fixed Costs (F) -0.004 -0.004 0.005 -0.002
(0.001) (0.001) (0.002) (0.002)
(Fixed Costs) *Co-op -0.363 -0.247 -0.502 -0.463
(0.099) (0.093) (0.074) (0.075)
Product Market Shock (5) 0.037 0.017 0.052 0.029
(0.003) (0.003) (0.002) (0.002)
(Product Market Shock) *Co-op 0.032 0.045 0.052 0.078
(0.024) (0.022) (0.019) (0.019)
Implied Effects
Log Wage Differential -0.164 -0.146 -0.140 -0.141

Capitalist-Coop
(0.028) (0.026) (0.028) (0.028)
d\nw/dr. Firms -0.011 -0.001 -0.011 0.009
Capitalist
(0.012) (0.011) (0.010) (0.010)
din w/dr. 0.096 0.110 -0.120 -0.104
Co-ops
(0.083) (0.078) (0.084) (0.085)
dlnw/dF: Firms -0.004 -0.004 0.005 -0.002
Capitalist
(0.001) (0.001) (0.002) (0.002)
d\nw/dF: -0.367 -0.252 -0.498 -0.465
Co-ops
(0.099) (0.093) (0.074) (0.075)
d\nw/dS: Firms 0.037 0.017 0.052 0.029
Capitalist
(0.003) (0.003) (0.002) (0.002)
d\nw/dS: Co-ops 0.068 0.062 0.103 0.107
(0.024) (0.022) (0.019) (0.019)
Notes: Estimated standard errors are in parentheses beneath their estimated coefficients. In deriving the esti
mates of equation (1) reported in columns (1) and (2) above, in addition to the variables listed above, the equations
are estimated the following variables: age, age squared, gender, two regional variables (distinguishing
including
the North, Center, and South), three industry variables (distinguishing manufacturing, construction, retail trade,
and other industries), two occupation variables (distinguishing operatives, clerical workers, and managers), the
of employment (to allow for any company size effects on wages), and dummy variables for each calendar
logarithm
is evaluated at the overall mean values for r, F, and S, namely, r =
year. The capitalist-co-op log wage differential
0.188, F= 0.298, and S= 0.044. These use 137,682 observations.
equations
In describing models of the capitalist of product market shocks to the firm's labor
firm and the worker co-op in Section I, we demand function. Such shocks will have
hypothesized that wages would tend to be first-order effects on
employment, hours, and
more volatile in the co-op. In
particular,
when other input quantities. This is not the case
the capitalist firm operates with a horizontal in the co-op, where shocks to labor demand
labor supply curve or when wages are fixed will affect the
co-op's wages. The presence
collective of (1) and (2) is designed to
by multi-employer bargaining, Spin equations
wages in the capitalist firm are independent measure these effects.

In fact, the estimated values of dlnw/d A sp?cification enabling us to examine

S indicate that wages were responsive to these differences between capitalist and co
product market shocks in both capitalist op enterprises is
and co-ops. However, the
enterprises posi =
tive fillip to wages was in the co-ops (3) + +
larger \nEjt lnwjtr\0 (ln^C^
and, in the estimates, the in + + +
within-group yi2 (rjtCpr\s +Fjtr\4 (FjtCj)r}5
teraction between S and the co-op dummy
is significantly greater than zero (as judged
by conventional statistical criteria). In the where lnii-, stands for the logarithm of em
estimates in columns (3) and ployment and lna;, represents the logarithm
within-group
(4), the effect of such shocks is roughly two of real average annual wages in firm j in year
to three times higher in co-ops than it is esti t. The elasticity of employment with respect
mated to have been in capitalist enterprises: to wages in capitalist firms is given by r]0.
in co-ops, a positive 10% product market C takes the value of unity for a co-op and
shock was associated with 1% higher wages, zero for a capitalist firm. The logarithmic
whereas in capitalist enterprises the same employment-wage
association in co-ops is
shock was associated with between 0.3% and given by Tio+ Th.
0.5% higher wages.14 These inferences are The user cost of capital paid by the firm
insensitive to the presence of fixed costs, is given by r, the company's fixed costs by F,
Fjt,
in the and transitory product market shocks by S.
equation.
Interactions between rand Cand between
F and C allow the effects on employment
V. Empirical Specifications and
of input prices and fixed costs in co-ops to
Results for Employment and Capital
differ from those in capitalist firms. Interac
tions between S and C allow the two kinds
Employment of firms to differ, as well, in the employment
The models outlined in Section I suggest z stands
impact of product market shocks,
that, for capitalist firms, a causal relationship for other observed attributes of firms, in
from wages to employment exists such that, their location and their
cluding regional
other is a
things equal, employment negative general industry. Factors affecting the
function of For are
wages. co-ops, wages
logarithm of employment but unobserved
jointly determined with employment and a the researcher are co and v.
by given by
well-defined causal relationship from wages to Those factors to each
specific enterprise
employment does not obtain. In addition, the and constant over time are
represented by
models in Section I imply that employment stands for factors that across
(Oy. vljt vary
is independent of fixed costs in the capitalist firms and over time.
enterprise but is a positive function of fixed A simple way of estimating equation (3)
costs in the co-op. Furthermore, employment is to first-difference it from one year to the
is to be more to transi next and thereby eliminate
expected responsive the enterprise
tory product market shocks experienced by fixed effects, tOj. For the co-op enterprises,
the capitalist enterprise than to those expe because are determined with
wages jointly
rienced by the co-op. we need to the
employment, recognize po
tential correlation between the equation's
stochastic term and We cannot
wage changes.
use a for as
14This result is compatible with and within-group strategy employment
Guiso, Pistaferri,
of firms'
we have done for wages because, in addition
Schivardi's (2005) analysis of the transmission
shocks to wages. that shocks to to the problem of fixed unobserved hetero
They found permanent
firms' value added are transmitted
geneity, we have the problem that the wage
to workers in the form
of changes in their compensation, a that may a we
finding is address
endogenous, problem using
be attributable to risk of bankruptcy and to incentive
lagged values of the wage as instrumental
effects. The magnitude of the effect is small, however,
and its estimation does not distinguish between variables. The endogeneity of wages means
co-ops
and capitalist firms. the wage is correlated
contemporaneously

Table 5. Implied Elasticities of Employment and of Capital with Respect

to the User Cost of Capital, Fixed Costs, and the Product Market Shock.
Wages,
Employment Capital
(A) (B) (C) (D)

Elasticity of Employment Evaluated Capitalist Co-Op Capitalist Co-op
or to: at the...
Capital with Respect Enterprises Enterprises Enterprises Enterprises
Wages Mean -0.803 0.017 -0.200 0.679

(0.032) (0.133) (0.067) (0.262)
Median -0.803 0.017 -0.200 0.679
(0.032) (0.133) (0.067) (0.262)
User Cost of Capital Mean -0.019 -0.002 -0.076 -0.070
(0.002) (0.022) (0.004) (0.047)
Median -0.018 -0.002 -0.072 -0.065
(0.002) (0.020) (0.004) (0.044)
Fixed Costs Mean 0.0007 0.0017 0.0024 0.0043
(0.0002) (0.0018) (0.0004) (0.0038)
Median 0.0006 0.0002 0.0002 0.0004
(0.0001) (0.0002) (0.00003) (0.0004)
Product Market Shock Mean 0.383 0.265 0.201 0.189
(0.003) (0.036) (0.008) (0.073)
Median 0.383 0.265 0.201 0.189
(0.003) (0.036) (0.008) (0.073)
The entries next to "mean" present elasticities evaluated at the sample mean values of the variables distinguishing
the capitalist mean from the co-op enterprises mean. A similar definition to "median" elastici
enterprises applies
ties. An alternative way of presenting elasticities is to compute the elasticities for each firm and then calculate the
mean and median values of these implied elasticities. In fact, these average and median elasticities were close to
those reported above.
with the error term of the employment equa The consequences of this instrumental vari
tion. If we used a within-group strategy to able regression are reported in column (A) of
remove the firm fixed effect, our instruments Appendix Table Al. To help draw inferences
would be invalid (in short panels) because from these estimates, columns (A) and (B) of
the within-group error term is the deviation Table 5 list the point estimates and estimated
of the error term from its time-series standard errors of the elasticities of
original employ
mean and this contains its past values that are ment with to w, r, F, and 5 evaluated
respect
correlated with the past values of the wage. A at mean and median values of
capitalist
and
first difference-instrumental variable strategy co-op variables.
enterprise
solves both the unobserved heterogeneity For capitalist firms, there is a well-defined
problem and the problem caused by the effect of wages on
employment,
with an em
endogeneity of wages. ployment-wage elasticity of-0.80.16 For the

The wages of co-ops and capitalist firms are co-ops, the
employment-wage relationship
treated
symmetrically so, using two-year
and is insignificantly different from zero with a
three-year lagged values as instruments, both
Aln ( C and Aln (W:t) are instrumented.15
Wjt).
16If we restrict the estimation to the observations on

capitalist firms only, wages may not need to be treated as
15These are valid instruments if all in which case we may
0)j captures endogenous, apply least-squares.
persistent factors and vi 7 is serially uncorrelated. The Ifwe do this, the estimated elasticity of demand for labor
first-differenced equation also includes regional, yearly, with respect to wages for capitalist firms is -0.705 with
and industry dummy variables. a standard error of 0.003.

40 INDUSTRIALAND LABORRELATIONSREVIEW
very small point estimate.17 Employment in differences, treating A\n(wjt)

and
A\n(wjt).
the capitalist firms is relatively insensitive to C as
endogenous
and
using two-year and
changes in the user cost of capital, r: a 10% three-year lagged values of these variables
increase in the user cost of capital reduces as instrumental variables. The resulting
employment by 0.2%. This elasticity is even estimates of equation (4) are contained in
smaller for the co-ops and its point estimate column (B) of Appendix Table Al and their
is
virtually
zero. The estimates attached to
implications are presented in columns (C)
fixed costs imply that, for both capitalist enter and (D) of Table 5.
prises
and co-ops, employment
was
virtually
For the capitalist firms, the demand for
independent of differences in fixed costs. capital with respect to the user cost of capital
According to the estimates of the parame is negative; the coefficient on wages is also
ters attached to the transitory product market negative,
so that and labor are
capital comple
shock variable, employment in both capitalist ments (as implied also by the estimates in
and co-op enterprises responded positively column (A) for employment); and capital
to
transitory product
market shocks. How is relatively unresponsive to differences in
ever, the
employment response
was smaller fixed costs. The point estimates suggest
for co-ops than for
capitalist enterprises
and that differences in fixed costs had larger ef
the interaction between the product market fects on capital for the co-ops than for the
shock and the co-op dummy is significantly capitalist enterprises, although
these co-op
less than zero. A positive 10% product market estimates are not calculated with
precision.
shock is associated with almost 4% greater For co-ops, the association between
capital
employment in the capitalist enterprises and and wages was positive. This finding is fully
about 2.5% greater employment in the co compatible with an income-maximizing in
ops. More generally, the estimates suggest stitution because wage income enters the
that, for the co-ops, employment is relatively co-op's objective function. The co-ops and
insensitive to changes in all the right-hand had similar elasticities
side variables, a finding that replicates previ of demand for capital with respect to the user
ous research on
employment
in
co-operative
cost of capital. Transitory product market
enterprises (see Pencavel and Craig 1994). shocks were associated with about the same
changes in capital in the capitalist firms as in
the co-ops: a positive 10% product market
Capital
shock was associated with 2% greater use of
A specification analogous to equation (3) in and
capital capitalist co-op enterprises.
maybe applied to describe the use of physical
across and
capital enterprises years:
Assessment
= + +
(4) \nKjt \nwjt[i0 (lnwjtCp u^ These results are broadly compatible with
the implications of the simple models of the
+ + + +
SjtV>6 (Sjfij)^7 *02 W27 *V capitalist firm and the co-op we sketched in
where for the logarithm of capital. Section I. Thus, a well-defined impact of
Inkstands on and on is im
same
Applying the logic we did for employ wages employment capital
ment, we estimate this in first plied by the textbook model of the capitalist
equation
firm and this is borne out by the estimates
in Table 5: a negative employment-wage
elasticity and a capital-wage elasticity are
17The absence of a meaningful association between
in co-ops speaks to an alternative measured for Because
wages and employment capitalist enterprises.
model that has been ascribed to Italian co-ops, constitute the maximand for a
namely wages co-op,
one inwhich
employment ismaximized subject to paying the model of the income-maximizing co-op
a wage to that firms. If
comparable paid by capitalist
developed in Section I does not permit the
this were the case, increases in wages induced by union
in the non-co-op sector would question to be asked, "What is the effect of a
management bargaining
on and
induce contractions of employment by co-ops, yet no co-op's wages employment capital?"
such relationship is estimated for co-ops.
Similarly, it is a fundamental implication of

profit-maximizing behavior of the capitalist Table 6. Least-Squares Estimates
firm that, among firms remaining profitable, of the Co-ops' Input Demand Functions.
increases in fixed costs have no impact on

Right-Hand
employment. This is consistent with the very Side Variable In(Employment) In(Capital)
small estimated elasticities of employment User Cost of Capital, r 0.061 -0.478
and capital with respect to fixed costs. By (0.186) (0.212)
contrast, increases in fixed costs are
predicted Fixed Costs, F 0.316 0.434
to lead to greater employment and greater (0.373) (0.423)
in co-ops. It takes a Product Market S 0.503 0.409
capital generous reading Shock,
of our results to discern support for that pre (0.035) (0.040)
diction. This may suggest that our measure R20.205 0.140
of fixed costs, long-term debt payments, is
because this measure contains the nobs_975_969
wanting: Estimated standard errors are in parentheses beneath
price of financial capital, its variations could their estimated coefficients.
embody a price effect and not simply the ef
fect of in the contractual
changes enterprise's
More for the co-ops,
payments. generally,
the use of employment and capital is pretty i^and S; and, assuming labor and capital are
much independent of changes in fixed costs. in the production
complementary function,
The estimates of the coefficients on r, and
w, the demand for capital is a negative function
5 for both capitalist enterprises and co-ops of r and a positive function of F and S.
are almost totally unaffected if the variable the statistical of differ
Though significance
measuring fixed costs, F, is omitted from the ences inFis questionable, the measured effect
estimating equations. of variations in the user cost of capital on the
Finally, employment was somewhat more demand for capital is economically meaning
responsive to product market shocks in ful: a 10-percentage-point increase in the user
capitalist enterprises than in co-ops. This is cost of capital reduced the demand for capital
consonant with the reasoning that transitory
by almost 5%. These co-ops clearly altered
shocks affect wages more and employment their use of labor and capital in response to
less in co-ops than in capitalist firms. Product market shocks.
transitory product
market shocks, we found, were associated
with about the same changes in capital in VI. Conclusions
as in co-ops.
The estimates of equations (3) and (4) do We have used data on individual workers
not, of course, constitute the reduced and individual companies to compare the
co-ops'
the wage and employment outcomes of capitalist
form input demand functions, because
worker's on the with those of worker co-ops. We
co-op wage appears right enterprises
hand side of these have also investigated differences in the use
equations. Least-squares
estimates of the coefficients on r,F, and S of of capital. Previous research has relied on
the relevant input demand functions for the case studies to investigate these issues and
co-ops are reported in Table 6.18 The signs has used data sets on workers and firms much
of each of the estimated coefficients in Table smaller and less extensive than those we ex
6 are consistent with the arguments in Sec ploit in this paper. In this sense, our results
tion I: employment is a positive function of can claim to constitute general findings about
and co-op In
capitalist enterprises. organiz
our about these outcomes, we
ing thoughts
have drawn upon conventional models of the
18These estimates are derived from using only the
profit-maximizing capitalist enterprise and
to regress the year-to-year
sample of co-op observations the
income-maximizing
worker co-op.
in employment and the year-to-year in
change change With respect to employment and capi
on the first differences in r, F, and 5 in addition
capital
to year dummy variables, and tal, these models' implications are largely
region dummy variables,
variables. congruent with our empirical findings.
industry dummy

That is, well-defined employment-wage and be more sensitive to market shocks

product
are estimated for in co-ops than in capitalist firms. Our
capital-wage relationships simple
capitalist firms that may be interpreted as the model of the
co-op
income-maximizing
elasticities of demand for labor and for capital led us to expect that co-op wages would be
with respect to wages; the employment-wage associated with input prices and
negatively
elasticity is calculated to have been -0.80 and fixed costs and these hypotheses, at least for
the capital-wage elasticity -0.20. The labor the estimates, are consistent
within-group
demand and the capital demand equations with our results. However, the of
magnitude
estimated for capitalist enterprises suggest these links is small: we find that co-ops fac
that labor and capital were complements ing a 10% higher price of capital tended to
as increases in the cost of capital reduced pay about 1% lower wages; co-ops with 10%
employment and increases inwages reduced higher fixed costs paid 0.18% lower wages to
the use of their workers.
capital. co-op
For co-ops, the
partial
correlation be The empirical regularities we have docu
tween
wages and
employment
was
essentially
mented have been organized around simple
zero, while capital was higher in those co-ops income-maximizing models of the capitalist
paying relatively high wages. The income and co-op enterprises, and the implications
maximizing model of the co-op suggests that of these models have generally fared well
these links between employment and wages when pitted against the data. Consequently,
and
capital
and wages
are not causal. Con the need for other models might seem less
sistent with the profit maximization model compelling even no doubt,
though, imagi
of the capitalist enterprise, employment and native minds could advance other models.
capital in such firms were unaffected by The hard work consists not in envisaging
changes in fixed costs. Finally, employment other models but, first, in identifying tests
was somewhat less
responsive
to
product
that discriminate between the simple mod
market shocks in co-ops than in capitalist els advanced here and more
complicated
firms, which is consistent with the notion models, and, second, in
organizing
a
large
that where workers command and set of data to to
enterprises representative speak
a greater voice will protect workers from these tests.
employment
reductions.
Certainly, the models we have investigated
With respect to wages, we such as
hypothesized ignore many confounding features,
that would not be corre considerations, the costs of
capitalist wages intertemporal
lated with input prices and fixed costs. The making input adjustments,
and issues associ
results in Table 4 suggest that, in the capi ated with bargaining. Moreover, because of
talist
enterprises
we studied, the association lack of data, we have ignored the prices of
between wages and the user cost of other such as raw materials and
capital many inputs,
was
certainly weak and the link between energy, and variations in these other input
wages and fixed costs was fragile, in some prices are not likely to be uncorrelated with
instances positive and in other instances variations in the price of capital. These issues
As we surmised, tended to warrant further
negative. wages investigation.

Appendix Table Al
Instrumental Variable Estimates of the Employment and Capital Equations
Dependent Variable
(A) (B)
Side Variable In (Employment) ln(Capital)
Right-Hand
of Wages, \nw -0.803 -0.200

Log
(0.032) (0.067)
[0.0080] [0.0081]
of Wages.Co-op Dummy, (lnw).C 0.819 0.879
Log
(0.136) (0.269)
[0.0178] [0.0212]
User Cost of Capital, r -0.094 -0.386
(0.009) (0.019)
User Cost of Capital.Co-op r.C 0.086 0.036
Dummy,
(0.108) (0.237)
Fixed Costs, F 0.136 0.456
(0.032) (0.067)
Fixed Dummy, EC 0.054 0.011
Costs.Co-op
(0.201) (0.424)
Product Market Shock, S 0.383 0.201
(0.003) (0.008)
Product Market Dummy, S.C -0.118 -0.012
Shock.Co-op
(0.036) (0.074)
Test 5.636 3.341
Sargan
(2) (2)
[0.060] [0.188]
These are fitted in first difference form. Entries in parentheses are estimated standard errors. Num
equations
bers in square brackets represent partial R^ statistics from the first stage regression equation (which addresses the
predictive power of the excluded instruments). In the case of the Sargan (over-identification) test, we report the
value of the statistic, the degrees of freedom (in parentheses), and the p-value of the test [in square brackets].
and calendar year dummy variables are each
statistically significant at the one percent level.
Regional, industry,
The employment is fitted to 95,240 observations and the capital equation to 95,129 observations. (Some
equation
firms report no capital, and we do not include these observations when fitting the capital equation.)

REFERENCES
Piero. 1996. La Lega: The Making a
Ammirato, of Suc Italian Producer
Cooperatives." Econ?mica, Vol. 52,
cessful Cooperative Network. Aldershot: Dartmouth No. 208 (November), pp. 449-65.
Publishing. Pencavel,John. 2001. Worker Participation: Lessons from
Bartlett, Will,John Cable, Saul Estrin, Derek C.Jones, the Worker Co-ops of the Pacific Northwest. New York:
and Stephen C. Smith. 1992. "Labor-Managed Russell
Coop Sage Foundation.
eratives and Private Firms in North Central Italy: An Pencavel, and Ben Craig. 1994. "The Empiri
John,
Empirical Comparison." Industrial and Labor Relations cal Performance of Orthodox Models of the Firm:
Review, Vol. 46, No. 1 (October), pp. 103-18. Conventional Firms and Worker Cooperatives."
and John Van Reenan. 2006. "Micro
Bond, Stephen, Journal of Political Economy, Vol. 102, No. 4 (August),
economic Models of Investment and Employment." pp. 718-44.
In James J. Heckman and Edward E. Learner, eds., Russell, 1985. in the
Raymond. Sharing Ownership
Handbook of Econometrics. Amsterdam: Elsevier, State of New York
Workplace. Albany: University
forthcoming. Press.
Cingano, Federico, and Fabiano Schivardi. 2004. Sertel, Mur?t R. 1982. "A Rehabilitation of the Labor
"
"Identifying the Sources of Local Productivity Growth. Firm." In Mur?t R. Sertel, ed., Workers and
Managed
Journal of theEuropean Economic Association, Vol. 2, pp. Incentives, pp. 11-26. Amsterdam: Contributions to
720-42. Economic Analysis, Vol. 140, North-Holland.
Earle, John. 1986. The Italian Cooperative Movement. Smith, Stephen C. 1994. "Innovation and Market Strat
London: Allen & Unwin. egy in Italian Industrial Econometric
Cooperatives: "
Fehr, Ernst. 1993. "The Simple Analytics of a Mem Evidence on Organizational
Comparative Advantage.
bership Market in a Labor-Managed In Economic Behavior and Vol. 23,
Economy." fournal of Organization,
Samuel Bowles, Herbert Gintis, and Bo Gustafsson, No. 3 (May), pp. 303-20.
eds., Markets and Democracy: Participation, Accountability, Thornley, Jenny. 1983. "Workers' Co-operatives and
and Efficiency, pp. 260-76. Trade Unions: The Italian Experience." Economic
Cambridge: Cambridge
University Press. and Industrial Democracy, Vol. 4, No. 3
(August), pp.
Guiso, Luigi, Luigi Pistaferri, and Fabiano Schivardi. 321-44.
2005. "Insurance within the Firm." Journal of Political Warhurst, 1999. Between Market, State, and
Christopher.
Economy, Vol. 113, No. 5 (October), pp. 1054-87. Kibbutz: The Management and Transformation of Socialist
Holmstr?m, Mark. 1989. Industrial Democracy in Italy: London: Mansell.
Industry.
Workers Co-ops and the Self-Management Debate. Aider Whyte, William F., and Kathleen K. Whyte. 1991. Mak
shot: Gower. The Growth and Dynamics
ing Mondragon: of theWorker
Jones, Derek C, and Jan Svejnar. 1985. 2nd ed. Ithaca, N.Y.: ILR Press,
"Participation, Cooperative Complex,
Profit-Sharing, Worker Ownership, and Efficiency in Cornell University.

Wages, Employment, and Capital in Capitalist and Worker-Owned Firms

Uploaded by

Copyright:

Available Formats

You might also like

Wages, Employment, and Capital in Capitalist and Worker-Owned Firms

Uploaded by

Document Information

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

Wages, Employment, and Capital in Capitalist and Worker-Owned Firms

Uploaded by

Copyright:

Available Formats

WAGES, EMPLOYMENT, AND CAPITAL

IN CAPITALIST AND WORKER-OWNED FIRMS

JOHN PENCAVEL,LUIGI PISTAFERRI,and FABIANO SCHIVARDI*

The authors how worker-owned and capitalist differ with

*John Pencavel is Levin Professor of Economics at

Downloaded from ilr.sagepub.com at NATIONAL UNIV SINGAPORE on June 29, 2015

ing the enterprise, normally the individual tion to labor.

Downloaded from ilr.sagepub.com at NATIONAL UNIV SINGAPORE on June 29, 2015

product market conditions (as embodied in These considerations frustrate clear

Downloaded from ilr.sagepub.com at NATIONAL UNIV SINGAPORE on June 29, 2015

Figure 1. Employment Differences between a Capitalist Firm and a Co-op.

out of an active role in trade union af and the use

1983:332). The ability of co-ops to ignore First, with respect to wages,

negative, and research is to

Downloaded from ilr.sagepub.com at NATIONAL UNIV SINGAPORE on June 29, 2015

Before the relevance co-op movement. The Census data in Table 1

Downloaded from ilr.sagepub.com at NATIONAL UNIV SINGAPORE on June 29, 2015

Table 1. Percentage of Co-Ops and of Co-Op Workers 1981-96.

Co-ops as a Percentage of All Establishments Co-op Workers as a Percentage of All Workers

All Sectors 0.61 1.02 1.06 2.50 3.66 4.06

Downloaded from ilr.sagepub.com at NATIONAL UNIV SINGAPORE on June 29, 2015

considerable turnover of officers (Holm The Survey of Companies

Downloaded from ilr.sagepub.com at NATIONAL UNIV SINGAPORE on June 29, 2015

Table 2. Statistics on Matched

Average Employment 301 373 235 228

Average Capital 9.52 4.23 8.92 6.80

Average Sales 38.43 29.42 36.03 50.03 45.55 40.24

Average User Cost of Capital 0.193 0.223 0.178 0.151

Average Fixed Costs 0.0131 0.0091 0.0058 0.0033 0.0083 0.0043

Downloaded from ilr.sagepub.com at NATIONAL UNIV SINGAPORE on June 29, 2015

?.- ? .. ^?* * --* i-4.

Downloaded from ilr.sagepub.com at NATIONAL UNIV SINGAPORE on June 29, 2015

Downloaded from ilr.sagepub.com at NATIONAL UNIV SINGAPORE on June 29, 2015

Workers are identified by broad job cate IV. Empirical Specifications

Downloaded from ilr.sagepub.com at NATIONAL UNIV SINGAPORE on June 29, 2015

Table 3. Statistics on Workers: The Matched

Average Wages 12.106 10.480 13.877 11.883 14.708 11.892

Downloaded from ilr.sagepub.com at NATIONAL UNIV SINGAPORE on June 29, 2015

equation (1), shown in the two "Ordinary 7^

components are not identified in these equations.

Downloaded from ilr.sagepub.com at NATIONAL UNIV SINGAPORE on June 29, 2015

Table 4. Estimates of theWage Equations (1) and (2).

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

(Q -0.078 -0.095 0.028 0.015

Log Wage Differential -0.164 -0.146 -0.140 -0.141

Downloaded from ilr.sagepub.com at NATIONAL UNIV SINGAPORE on June 29, 2015

In fact, the estimated values of dlnw/d A sp?cification enabling us to examine

Downloaded from ilr.sagepub.com at NATIONAL UNIV SINGAPORE on June 29, 2015

Table 5. Implied Elasticities of Employment and of Capital with Respect

(A) (B) (C) (D)

Wages Mean -0.803 0.017 -0.200 0.679

endogeneity of wages. ployment-wage elasticity of-0.80.16 For the

16If we restrict the estimation to the observations on

Downloaded from ilr.sagepub.com at NATIONAL UNIV SINGAPORE on June 29, 2015

very small point estimate.17 Employment in differences, treating A\n(wjt)

Downloaded from ilr.sagepub.com at NATIONAL UNIV SINGAPORE on June 29, 2015

profit-maximizing behavior of the capitalist Table 6. Least-Squares Estimates

increases in fixed costs have no impact on