Job Creation and Job Destruction in the Theory of Unemployment

Dale T. Mortensen
Christopher A. Pissarides
Job Matching, Wage Dispersion,
and Unemployment
edited by
Konstantinos Tatsiramos
Klaus F. Zimmermann
Oxford
University Press
Oxford, 2011, 1-16.
Mortensen and Pissarides:
Job Creation and Job Destruction in
the Theory of Unemployment
Konstantinos Tatsiramos and Klaus F. Zimmermann
The 2005 IZA Prize in Labor Economics was awarded to Dale T.
Mortensen and Christopher A. Pissarides for their path-breaking
contributions to the analysis of markets with search and matching
frictions. The annual IZA Prize is awarded for outstanding academic
achievement in the field of labor economics. Both their individual
contributions and their joint development of a dynamic equilibrium
model of labor markets account for much of the success of job search
theory and the flows approach in becoming a leading tool for microeconomic and macroeconomic analysis of labor markets. The 2005
IZA Prize Laureates have been selected, jointly with Peter Diamond, as
the recipients of the 2010 Nobel Memorial Prize in Economic Sciences.
Their models, which are now widely used in labor economics and macroeconomics, have highly enriched research on unemployment as an
equilibrium phenomenon, on labor market dynamics, and cyclical adjustment. Their research on labor market search and job matching has
also significantly directed and shaped the empirical literature.
The award ceremony was preceded by the IZA Prize Conference
in honor of Professors Mortensen and Pissarides, where a number of
distinguished scholars, including Nobel Laureate James Heckman,
presented papers on topics such as the effects of cognitive and noncognitive abilities on labor market outcomes, the future of unions,
and the role of civic attitudes on the design of labor market institutions. Previous winners of the IZA Prize in Labor Economics include
Jacob Mincer, Orley Ashenfelter, and Edward Lazear. These past winners have contributed to the field of labor economics by undertaking
pioneering research on topics such as human capital investments,
on-the-job learning, female labor supply decisions, trade union
1
Introduction by the Editors
membership, program evaluation methodologies, compensation
schemes and worker productivity, and personnel economics. Professors Mortensen and Pissarides join this esteemed group of economists
for their exceptional work in labor economics.
This book contains five papers which helped to shape the equilibrium search model. The first chapter by Mortensen (1982) and the second chapter by Pissarides (1985) are among the contributions which
have initiated the research on what is known today as the search and
matching model of the labor market. The third chapter by Pissarides
(1986) provides the empirical foundations for the flows approach
to the labor market analysis. The fourth chapter is the joint paper by
Mortensen and Pissarides (1994), which is a complete statement of the
equilibrium search and matching model with endogenous job creation
and job destruction. The fifth chapter by Mortensen (1990) examines
models of wage inequality.
An introduction written by the authors discusses the flow view of
the labor market that replaced the disequilibrium theory and formed
the basis for the development of a coherent theory of worker and employer behavior. The authors provide a summary of each chapter with
a discussion of their important contributions and their relation to the
existing literature. Finally, the book concludes with a final chapter
written by the authors with an extensive summary of the developments in the literature since the Mortensen-Pissarides model was put
in place.
Dale Mortensen has pioneered the study of individual workers’
job search decisions. The insight that search frictions can be modeled as random arrival of trading partners – or stochastically arriving
matching opportunities – has provided crucial thrust to the idea that
searching for wage offers and jobs is costly when workers and firms
lack full information about prices in the labor market. This representation of frictions has also revolutionized research on labor turnover
and reallocation, as well as research on personal relationships. Dale
Mortensen has also advanced the field in his joint work with Kenneth
Burdett (Burdett and Mortensen 1980), explaining search and layoff
unemployment as equilibria phenomena and showing how a search
model with wage posting can generate wage dispersion (Burdett and
Mortensen 1998).
Christopher Pissarides has broken new ground in studying macroeconomic implications of the flows approach to labor market analysis
by using the matching functions as a tool to study equilibrium unem2
Introduction by the Editors
ployment. The matching function relates job creation to the number
of unemployed, the number of job vacancies, and the intensities with
which workers search and firms recruit. It successfully captures the
key implications of frictions that prevent an instantaneous encounter of trading partners and has proved a particularly powerful tool
for modeling two-sided search frictions that stem from information
imperfections about potential trading partners, not least because it
can be incorporated in models without adding excessive complexity.
Pissarides (1979, 1985, 1994) further developed the matching model,
which is at present the leading tool for studying imperfect labor markets in macroeconomics, in subsequent studies of equilibrium unemployment dynamics.
The essence of the main ideas of the search and matching framework and the insights of the immense literature that was stimulated by
their fundamental contributions are presented in two state-of-the-art
joint articles that are the prime references for researchers interested in
the field: “Job Reallocation, Unemployment Fluctuations and Unemployment Differences” (Mortensen and Pissarides 1999a) focuses on
the macroeconomic implications of the flows approach, while “New
Developments in Models of Search in the Labor Market” (Mortensen
and Pissarides 1999b) centers on the implications of search decisions at
the individual worker level. Pissarides has elaborated the matching approach in “Equilibrium Unemployment Theory” (1990, 2000), which
has become the leading monograph in the field.
The search and matching framework advanced by Dale Mortensen
and Christopher Pissarides is a theoretical tool for the analysis of labor markets suitable for policy analysis. In a joint paper, Mortensen
and Pissarides (2003) compared the impact of policy for the European Union and the United States, showing that the impact of policy is
quantitatively different at different skill levels because the smaller the
surplus created by a match, the bigger the impact of a given policy on
wages and job creation. Christopher Pissarides has also carried out a
quantitative evaluation of the tax recommendations in the 1994 European Commission White Paper.
The matching model has been applied to the study of labor market policy, especially in Europe, following the rise of unemployment
in the 1980s. The European Commission through the Delors Report
(European Commission 1994) and the Organization for Economic
Cooperation and Development through its Jobs Study (OECD 1994),
recommended a reform of labor market policy to increase job creation.
3
Introduction by the Editors
Researchers from IZA's core staff have contributed to the application
efforts of the matching model (e.g. Sunde 2003).
The research findings of Mortensen and Pissarides also underscore
the reform course initiated by the “Hartz” reform of the German government (strongly supported by IZA through various publications including Zimmermann 2003), which proposes incentives to take up employment and efforts to reduce search costs through a more efficient
job placement process in order to shorten unemployment durations.
Their work fits nicely with the mission and spirit of the Institute for
the Study of Labor. Over its more than ten-year existence, IZA has constantly striven to position itself at the intersection of sound academic
research on labor economics and policy-making.
IZA has been active in the research area that was pioneered by
Mortensen and Pissarides through the Program Area “Labor Markets
and Institutions.” This program area analyzes the effects of institutions on labor market outcomes. A particular emphasis is on the analysis of the institutions on the micro level and their macro­economic consequences for economic performance. Past events include workshops
on wage inequality, labor market institutions, structural change, employment protection, frictions, firm dynamics, and growth. Both Dale
Mortensen and Christopher Pissarides have been active in a number of
IZA events, presenting their most recent research in the IZA seminar
series and conferences. Both Mortensen and Pissarides have been IZA
Research Fellows since 2001.
Over several decades, innovations in the field of labor economics
have been fostered on the microeconomic level through the increased
availability of micro data and the development of better quantitative
methods. Mortensen and Pissarides have forcefully reminded the area
about the strong potentials of macroeconomic thinking by providing
researchers with a convincing and workable quantitative tool. Both the
2005 IZA Prize in Labor Economics and the 2010 Nobel Prize in Economics recognize their contribution in explaining labor markets with
frictions and the applicability of their theory to markets other than the
labor market
4
1
Introduction: The Flow View
of the Labor Market
The flow view of the labor market is fundamental in the theory of
equilibrium unemployment, as expressed in Pissarides (2000). Unemployment is a state which individual workers occupy for a relatively
short period of time, as they seek rewarding long-term employment.
Individual workers flow into the state and others flow out. Offsetting the flow of job-finders and unemployed workers who have given
up job search to take up leisure or other activities outside the labor
force, is a simultaneous flow into the state of unemployment of workers who lost their jobs either through layoff or quit, and new entrants
from a nonparticipation state.
This conception of the labor market has replaced the disequilibrium theory under which employment was determined by the real
wage and labor demand, and unemployment simply reflected excess
supply at a real wage that was above that required to clear the labor
market. The disequilibrium theory, which viewed the labor market
as a spot market that meets every day to allocate jobs to workers,
was inconsistent with the observations regarding the experience
of individual workers over time; in particular the fact that a small
fraction of the labor force flow in and out of unemployment each
year and the typical employment relationship lasts for several years.
Furthermore, available jobs are heterogeneous, offering different
wages, different long-term career opportunities, and various kinds
of non-pecuniary features, such as job security. These facts suggest
that workers might find it in their interest to invest time and effort
in the process of finding a “good” job, one that pays well, that might
offer a challenge, good promotion prospects and consequently have
a chance of lasting for a while.
5
The Flow View of the Labor Market
A natural implication of these ideas, and the empirical fact that
unemployment durations are typically quite short relative to employment durations, is that the labor market should be viewed in
stock-flow terms. Since the early 1970s, this alternative view of unemployment has received empirical support from observations on
worker flows found in the US Current Population Survey, in many administrative data sources collected for the servicing unemployment
insurance systems in Europe, and in new panel data sources that were
becoming available and were to prove influential in labor market research, such as the Michigan Panel Study of Income Dynamics.
In this alternative view of the labor market finding a job is not akin
to waiting on the corner for someone to offer a day's work at the “going wage”. Instead, acquiring an acceptable job is the outcome of an
information gathering process in which the worker exploits contacts,
friends, and neighbors, about job availability. Other more formal
channels of information are also available. “Help wanted” advertising
in newspapers and vacancies posted with the government's employment service inform decisions about where to apply for specific openings. The product of this effort is the prospect of an offer or a sequence
of offers distributed over time. When an employment offer arrives, the
worker has to decide whether or not to accept given the information
gathered to date, the offers previously generated that are still open,
and general information about the availability and generosity of offers that might arise in the future. In other words, the alternative to acceptance for a market participant is not home production, schooling,
or some other non-market activity, although these options are always
available. The default option is normally continued search.
The entry into unemployment is also the result of choices made by
workers and firms, and subject to similar comparisons with the alternatives available elsewhere in the market under imperfect information. Joining the labor market from outside the labor force often
entails starting a process of search similar to the one for existing unemployed workers. Is this a better option than staying outside and
engaging in home production, leisure, or education?
If the worker is already in a job and has the option of staying on,
quitting into unemployment poses a problem that is very similar to
the acceptance problem of the unemployed job seeker. The existing
job is an implicit offer on the table to continue employment, the quitting alternative is a search option for an uncertain outcome. Even
for jobs which have become less profitable than when they were first
6
The Flow View of the Labor Market
opened, there is a non-trivial choice to be made between continuing
employment or leaving the job and seeking new job offers.
A similar choice is faced by the firm. Should it close down jobs that
have become less profitable or continue production? A large part of
entry into unemployment is the result of job closure but this job closure is not necessarily the result of shocks that make the job's overall profits negative instantly and forever. Just as there are reservation
wages for unemployed workers looking for job offers, there are reservation productivities for employed workers and their firms. In good
times a job might be destroyed more readily than in bad times, because the outside options of the worker and firm are better.
This view of the labor market formed the basis for the development
of a coherent theory of worker and employer behavior in a market in
which trading opportunities are generated through search and jobs
are subject to change. The theory is based on solid maximizing microfoundations under rational expectations about the future, when
all gains from private trades are exploited within the constraints
imposed by the market information structure. It builds up from the
microfoundations to give equilibrium macro models with unemployment and employment as distinct states, with large flows between
them. The development of models to explain these processes is the
topic of the papers in this book.
The first attempts at developing such a theory took place in the
1960s, following pioneering work by Stigler (1962), Phelps (1967),
and Friedman (1968).1 But those early attempts were essentially partial models of the labor market, and the problems that subsequent
authors addressed were problems that could be studied within a partial equilibrium framework, such as empirical studies of the distribution of unemployment spells lengths.
The paper by Lucas and Prescott (1974) represents the first attempt to formulate a formal model of search equilibrium and unemployment for an economy as a whole. Their economy is made up of
separate “islands”, a metaphor that had been popularized by Phelps.
Specifically, production takes place at different locations and search
is the time spent transiting from one to another. As wages and employment are determined by the law of “supply and demand” on each
island and workers migrate toward the islands that offer the most
“rent” in a manner which is not completely specified, their model is
very much in the abstract tradition of competitive equilibrium analysis. Hence, their not surprising conclusion that the level of unem7
The Flow View of the Labor Market
ployment attained is constrained efficient follows. Although an important branch of the literature stemming from this source lives on,
its relative lack of current popularity seems to be due to the fact that
the micoeconmics of the search process are not explictly spelled out.
Beginning in the late seventies and early eighties, Diamond and
Maskin (1979) and Diamond (1982a,b), Mortensen (1978, 1982a,b),
and Pissarides (1979, 1984a,b, 1985a) began to formulate matching
models that included in an explicit way search on both sides of the
market. In these papers, the notion of an equilibrium with a “search
technology” was more fully developed. 2 Within the context of these
models, one could formulate and answer questions regarding the
existence and efficiency of the matching structures that arose in
equilibrium, including the “natural” or equilibrium rate of unemployment, the principal topic of the first edition of Pissarides (1990).
In these models, wages, or more generally prices, were viewed as
the outcome of bilateral bargains struck by both parties to matches, rather than dictated by one of the two, as in Phelps (1968) and
Mortensen (1970). These bargains split the rents induced by the fact
that finding alternative matching partners was neither instantaneous nor costless. Naturally, the wage bargains are struck ex post,
after the parties meet. This formulation raised an obvious incentive question: Would the division of match rents that resulted from
bilateral bargaining generate equilibrium structures that were socially efficient in some sense? Although the obvious answer was a
resounding no, the formulation of the problem initiated a new literature on the existence and properties of market equilibrium solutions to fully articulated search models.
Of our early contributions we include here Mortensen (1982a,
Chapter 1) and Pissarides (1985b, Chapter 2). Together with Diamond's (1982a,b) papers they initiated the research that led to what is
known today as the search and matching model of the labor market.
Chapter 1 (Mortensen 1982a) focuses on the decisions of unmatched
agents to participate in the process of forming matches. This paper is
one of the first papers to deal explicitly with the dual issues of existence
and efficiency of search equilibrium and was also one of the first to address issues on assignment in an environment in which search frictions were important. The model is one of “partnership formation”.
Partnerships are coalitions of two types of individuals, men and women for example, that could produce a joint value that exceeded the sum
of what the two partners could accomplish independently. Potential
8
The Flow View of the Labor Market
partners meet at random. Although all partnerships could be viewed as
identical ex ante, “match quality,” the output specific to the match, is
revealed only after two agents meet. This feature of the model (which
was introduced in the context of labor turnover by Jovanovic 1979),
distinguished it from the case in which agents have fixed “abilities,” or
other characteristics that determined joint output.
The division of the surplus match output generated by the search
frictions and the ex post match quality is a jointly rational division, a
point on the contract curve defined by the available outside options,
continued search. Hence, matches form if, and only if, it is in the interest of both parties. Although there are generally many jointly efficient
divisions of the surplus as in any bilateral bargaining problem, the generalized Nash solution became a focal point of interest in the analysis.
The paper considers two different specific but plausible functional
forms for the matching function, linear and quadratic. In the linear
matching function case, each unmatched agent on either side of the
market can contact a randomly selected unmatched agent on the other side at a chosen frequency. Of course, the same agent might also
be contacted by someone of the other type. Hence, the meeting rate
at which an individual agent meets a prospective partner is equal to
her own search effort plus the average effort of agents on the other
side of the market. The frequency of meetings that a particular agent
could be a party to can be viewed as a Poisson random variable with
expected meeting frequency equal to the meeting rate just defined.
The quadratic matching function is an extension of this simple set
up and can be viewed as one in which individuals cannot tell ex ante
whether a potential partner is matched or not. Hence, the effective
meeting rate with an unmatched agent of the opposite type is the
product of the meeting rate and the probability that the agent met is
not matched. The linear matching function exhibits constant returns
to scale while the quadratic specification implies that the matching
process is subject to “increasing returns” in the sense that a doubling
of the number of unmatched agents of each type, holding the total
number constant, would quadruple the aggregate number of meetings per time period of fixed length.
The rate at which a match produces output is a random variable ex
ante, which is realized when two potential parties to the match meet.
The decision to create a match for the pair involves a joint choice of a
minimally acceptable value of match quality, analogous to the choice
of a reservation wage in the stopping problem formulation of the one9
The Flow View of the Labor Market
sided search model. Given that the two parties divide the match in
a jointly rational manner by selecting a point on the contact curve,
the reservation product is that which equates the sum of the expected
present value of the two parties’ future incomes, were they to form
the match, to the sum of the two option values of continued search.
The rent to be divided is the match surplus, the difference between
the joint value of a match and the sums of the values of continued
search. The optimal search effort choices maximize the difference
between the total expected return and cost of search given the search
effort choices of all other individuals. A symmetric matching equilibrium is simply a pair of search effort choices, one for each agent type,
that maximize each individual’s expected future income given the
search efforts chosen by all other agents in the market. An increase in
the search effort of potential trading partners increases the expected
meeting frequency of every individual, but an increase in the search
effort of one's own type decreases the frequency. These two spill-over
effects are the principal externalities of interest in the welfare analysis
that is discussed in the paper. The paper demonstrates the existence
and uniqueness of equilibrium and establishes that a sharing rule exists, at least in an ex ante sense, which will induce efficient choices.
The second and third papers in the volume, by Christopher Pissarides, are more explicitly macroeconomic and they contain the
foundations for the model in the first edition of his book (Pissarides
1990). The problem that they address is the existence of a natural
unemployment rate that can fluctuate over time. When Friedman
and Phelps launched the idea of a natural unemployment rate it
was implicitly understood that it was a feature of steady-state equilibrium and that over the cycle the rate would be sluggish at best.
Fluctuations in the unemployment rate were then analyzed within
the context of a Phillips curve, essentially driven by misperceptions about prices and wages. But the large changes in unemployment that followed the oil and material shocks of the 1970s and
the productivity slowdown that started in 1973 did not seem to be
consistent with this idea. It was implausible to argue that the massive unemployment hikes, especially in Europe, were due to wrong
expectations about prices. Nor was the idea of a constant natural
rate obviously consistent with the “Beveridge curve,” the negative
relation between vacancies and unemployment discussed at length
in several papers in the 1960s and 1970s, following the pioneering
work of Dow and Dicks-Mireaux (1958).
10
The Flow View of the Labor Market
Pissarides' first paper in this volume (Pissarides 1985b, Chapter
2) borrowed ideas from Nash bargaining theory, including those in
Mortensen's work (Chapter 1 of this volume), to argue that the bilateral wage bargains that share the rents in a search equilibrium imply
some wage stickiness. The reason is that the wage equation that results from the generalized surplus-sharing rule is a weighted average
of the worker's marginal product and her non-market returns, such
as unemployment compensation, value of leisure, and home production. If the model is then subjected to a real productivity shock (or a
material one, or even a nominal misperceptions one), and the worker's non-market returns are not responsive to such shocks, wages will
not fully respond to the shock and so the “natural,” or “equilibrium”
unemployment rate will change.
The contribution of the paper was to embed this idea into a fully
developed equilibrium model of the labor market that behaved, in
all other respects, as a neoclassical Solow model with unemployment. The key building blocks were (1) the arrival rates of matches
that depended on one of the model's key unknowns, “tightness,”
the ratio of vacancies to unemployment, (2) the job vacancy equation, and (3) the Nash wage equation. The arrival rates of matches
could be derived from an aggregate matching function with constant returns to scale, which make the model compatible with a
Solow growth model (although no growth was explicitly considered). This assumption makes tightness the key unknown in this
class of models. The firm's decision to create vacancies was modeled as a zero-profit condition on the marginal vacancy, which
gave rise to an equation virtually identical to a dynamic demand
for labor equation with hiring costs.
The transmission of shocks to unemployment was then as follows.
Let a positive shock hit the marginal product of labor. Each firmworker pair now make more profit and this extra profit is shared between the firm and the worker according to the Nash wage equation.
Because of the influence of non-market returns on the wage equation,
the wage rate does not absorb all the rise in labor's marginal product,
and profits also rise. Firms, anticipating more profit per job, now open
more job vacancies and enter the market to look for a match. The aggregate matching function gives a path for the fall in unemployment,
as more matches take place. The model, at least qualitatively, reproduced the Beveridge curve with anti-clockwise loops around it and
a “sticky” wage. It thus delivered cyclical unemployment dynamics
11
The Flow View of the Labor Market
without any reference to inflation or the Phillips curve, and with fully rational maximizing behavior under rational expectations.
The model, however, could also be reinterpreted as a static equilibrium model of the labor market. The zero-profit condition for vacancy entry was essentially a demand for labor equation with similar
properties to conventional demand curves, whereas the supply of
labor, which was the controversial relation in neoclassical models
of employment fluctuations (such as the intertemporal substitution
model), was replaced by the Nash wage equation. The conventional
labor supply equation in the model was a vertical line, with employment given inside the line according to the parameters of the matching function and the zero-profit condition.
The model had one more feature which gave richer dynamics. As in
the Jovanovic (1979) and Mortensen (1982b) papers, the ex post productivity of a match varied across matches, although ex ante there was
homogeneity across workers and jobs. This assumption introduced the
idea of a reservation match quality, which reflected a reservation wage
and a reservation profit level. The firm and worker agreed, because of
the Nash sharing rule that applied continually, which job matches to
form (after a meeting) and which to reject (after the arrival of a shock).
The match acceptance decision was identical to the one previously
studied in a similar equilibrium setting (Pissarides 1984b). The job destruction decision anticipated in one respect the later findings of Davis, Haltiwanger and Schuh (1996) and the model of Mortensen and
Pissarides (1994, see Chapter 4 in this volume). In particular, if an aggregate negative shock hit jobs, many job matches close to the reservation were instantly destroyed; no such mass creation took place when
a positive shock hit, as matches still had to be formed through the slow
matching process. So there was an asymmetry between rises and falls
in unemployment due, as the later analysis made clear, to the cyclical
asymmetry in job creation and destruction.
Chapter 3 (Pissarides 1986a) has to be seen in the context of the
changes taking place in labor markets in Europe in the 1970s and
1980s. But the more lasting contribution of the paper was to provide
the empirical foundations for the flows approach to labor market
analysis, and more specifically for the building blocks of the model
in Pissarides (1985b).
The paper was commissioned by the editors of Economic Policy,
at that time a new journal whose objective was to do for European
economics what the Brookings Papers in Economic Activity was do12
The Flow View of the Labor Market
ing for US economics, to apply the ideas in Pissarides (1985a,b) to the
rise of unemployment in Britain. In 1973, unemployment in Britain
was about 2% but at the cyclical peak of 1978/79 it only fell down to
5% and by 1983 it had reached 11%. It was clear from the data that
most of the rise in unemployment was due to a rise in the natural rate,
which existing models could not explain. Models with real shocks
quickly appeared to address the problem (most influential among
them were models by Bruno and Sachs 1985, and Layard and Nickell 1986a), relying on non-market clearing wages or union bargains.
The claim made in the Economic Policy paper was that we could learn
more about the dynamics of unemployment and the natural rate by
studying the flows in and out of unemployment and the dynamics of
job vacancies, because these statistical series contained a lot of useful
information that traditional models neglected.
The paper uses quarterly British data to address a number of issues
that came up in the first generation of search and matching papers. In
particular, the four issues that it focuses on are, (1) are the unemployment dynamics driven by the inflow or the outflow rate, (2) is there a
well-behaved matching function and is it characterized by constant
or increasing returns to scale, (3) is the search equilibrium efficient,
and (4) is there a well-behaved “vacancy supply” equation?
The answer to (1) was overwhelmingly in favor of the outflow from
unemployment, with the inflow making a trivial contribution, except
possibly for the two years of very fast rise in unemployment in 1980–
81. This provided a justification for focusing on the much simpler equilibrium model with constant job separation rate. The answer to (2) was
also unequivocal. The time series data overwhelmingly supported constant returns to scale. Two versions of the matching function were estimated, a linear one and a log-linear (Cobb-Douglas) one. Both strongly
supported constant returns, with unemployment elasticity about 0.7.
The answer to (3) was more speculative. It has always been known
that the search equilibrium suffered from externalities that implied inefficiencies (Diamond 1982b; Mortensen 1982a,b; Pissarides
1984a,b). The externalities have both positive and negative components: when one more worker enters the market to look for a job, firms
with job vacancies are made better off but workers looking for jobs
are made worse off. Diamond (1982b) had already hinted that there
might be an internalized Nash wage rule and Pissarides (1984b) derived explicitly the share of labor in the wage bargain that internalizes the externalities, what became later known as the “Hosios rule”
13
The Flow View of the Labor Market
(Hosios 1990). Efficiency required constant returns and the share
of labor in the Nash wage bargain set equal to the elasticity of the
matching function with respect to unemployment. In the empirical
application, Pissarides (1986a, see Chapter 3 in this volume) speculates that the share of labor in the symmetric Nash bargain is 1/2, but
also makes use of another bargaining solution suggested by Binmore,
Rubinstein and Wolinski (1986), that makes the share of labor equal
to v/(u + v); where v are vacancies and u is unemployment. He then
argued that in both cases, if the share of labor was to be 0.7, tightness
(v/u) needed to be far higher than observed in his data.3
Finally, the paper makes an attempt to “close the model” and estimate an econometric version of the zero-profit condition that gives the
supply of job vacancies. Combined with the matching function and a
constant job separations rate, the two equations are used to make predictions about the factors that influence unemployment. The supply
of vacancies equation is, however, essentially a demand for labor equation, with the wage rate substituted out from the Nash wage equation.
The equations fitted the data well and strong predictions were made
about the factors that contributed to the rise in British unemployment.
The joint paper by Mortensen and Pissarides (1994), reproduced in
this volume as Chapter 4, is a complete statement of the equilibrium
search and matching model of the labor market with endogenous job
creation and job destruction. Although job destruction played a small
role in the earlier model of Pissarides (1985b, see Chapter 2 in this volume), the dynamics of unemployment in the then canonical model
were essentially driven by the job creation rate. In the meantime,
influential data compiled by Davis and Haltiwanger (1992) showed
that the job destruction rate in the US manufacturing sector had at
least as much variance as the job creation rate. Subsequent work in
other countries reached similar conclusions. In the Mortensen-Pissarides (1994) model the search and matching (job creation) side of
the model is similar to the one in Pissarides (1985b and 1990) but job
destruction is now a decision variable and the job destruction rate becomes volatile in response to shocks. Once the match is made there
are idiosyncratic productivity shocks that move the job's productivity around. The firm and worker have the choice of continuing at the
new productivity realized after a shock, or of destroying the job and
separating. The authors show that the job destruction decision is governed by a reservation productivity; if the job's productivity goes below the reservation, the job is destroyed.
14
The Flow View of the Labor Market
Because job destruction can be instantaneous (barring any job destruction costs, e.g., associated with “employment protection legislation” that slow down the agents' response to shocks), it leads the cycle.
When cyclical shocks arrive job destruction changes first and job creation follows as the matching process gathers momentum. Also, when
negative cyclical shocks hit, many jobs with productivity close to the
reservation now become unprofitable, and close down. This mass destruction following negative shocks explains the spikes in the job destruction data of Davis and Haltiwanger (1992). Mortensen and Pissarides show that reasonable parameter restrictions explain the dynamic
properties of job flows well. They apply their model to the controversy
about the origin of cyclical shocks and argue that the job creation and
destruction data support aggregate shocks as the driving forces of the
cycle rather than sectoral shocks as in Lilien's (1982) model.
The model of Mortensen and Pissarides (1994) is characterized by
productivity differences across firms and wage differentials. It offers itself for an analysis of worker flows and job to job movements,
through on-the-job search, which was absent in the original model. These issues were studied within the context of the model by
Mortensen (1994) and Pissarides (2000, Chapter 3). As in the partial
equilibrium model of Burdett (1978), the optimal policy when there
is on the job search is one characterized by two reservation wages.
Jobs whose productivity drops below the low reservation wage are destroyed, whereas workers in jobs with productivity between the two
reservations search on the job for a better match. If jobs have higher
productivity, there is no on-the-job search.
The wage inequality in the Mortensen-Pissarides model, however,
is entirely due to exogenous productivity differences across jobs.
Richer models of wage inequality for identical workers were examined by Mortensen (1990b), reproduced as Chapter 5 of this volume.
Mortensen considers wage dispersion, defined as sustainable differences in wages for identical workers, in models where the firm sets the
wage. These models have become known as “wage posting” models,
in contrast to the “vacancy posting” models considered so far, where
the wage is determined by the Nash bargain solution. In their seminal
paper, Burdett and Judd (1983) demonstrated that pure and permanent price dispersion was a theoretical possibility when price information is costly. The Burdett-Judd proof that equilibrium dispersion
can persist built on an earlier paper by Diamond (1971) set in a product market context. In a context in which many profit maximizing
15
The Flow View of the Labor Market
sellers set prices and buyers engage in costly sequential search, Diamond had shown that a non-cooperative Nash equilibrium solution
to the game exists in which there is no dispersion when all buyers are
identical. The single price is the buyer's common reservation price.
The essential idea of the proof is straightforward. If no buyer expects price dispersion, he has no incentive to pay the cost of gathering
more than one price quote. But if every buyer gathers only one offer,
the seller's dominant strategy is to charge the buyer's reservation price.
Interpreted in a labor market context where employers and workers are
identical and workers search only while unemployed, there is only a
single equilibrium wage offer, which is the monopsony wage. Albrecht
and Axell (1984) extend the analysis by considering the case of different types of workers. They show that if there are two groups of workers
who have different values of non-market time and if labor productivity
varies across employers with sufficient dispersion, a two-wage equilibrium offer distribution exists. Although the two equilibrium offers
equal the endogenously determined reservation wage rates of the two
worker types, the lower offer generally exceeds the monopsony case.
The assumption that workers do not receive job offers while employed has the consequence that every equilibrium wage offer must
be the reservation wage of some worker type. Allowing for on-the-job
search, Burdett (1990) and Mortensen (1990b, Chapter 5 in this volume) show that an equilibrium is a non-degenerate continuous distribution of wage offers, with lower support equal to the common reservation wage when workers are identical and offer arrival frequencies
are independent of employment status.
The paper in this book extends the previous analysis to the case of
heterogeneous workers and employers. In the case of workers, the model is closely related to and, indeed, an extension of Albrecht and Axell
(1984). The paper shows that the equilibrium distribution is characterized by a continuous density on an interval that strictly contains all the
reservation wage rates when employed workers search. Extending the
analysis to the case in which productivity differs across employers is
the second contribution of the paper. As in the case of worker heterogeneity, only a finite number of employer types are considered. Although
the equilibrium distribution of wage offers is continuous, the form of
the distribution function is rather complicated. The paper concludes
with a short discussion on identifying the structural parameters of
the model, using micro panel data on worker spells of unemployment,
wages earned, and employment spell lengths.
16