Transcript Calavrezo_COINVEST

Birds of a feather flock together
(or Networks of Mobility)
Oana Calavrezo, Richard Duhautois, Francis Kramarz
Motivation
Most analyses of workers’ mobility episodes have been based on personlevel data sources: the PSID in the U.S. (Abraham and Farber, 1987;
Altonji et al., 1987; Topel, 1991; Buchinsky et al., 2005).
The CWHS or the LEED data are “old” versions of tools that are being
developed systematically now, called longitudinal matched employeeemployer data sources (Topel and Ward, 1992; Anderson and Meyer,
1994).
Longitudinal matched employee-employer data sources give information
on both sides of the labor market. This paper belongs to the new wave
of studies based on such tools (see Abowd, Kramarz and Roux, 2006;
Beffy et al.,2004).
COINVEST Conference, Lisbon 2010
Issue
The purpose of this study is to focus on the characteristics of firms when
analyzing within- and between-industry mobilities (at two-digit level)
between 1991 and 1999.
Three original contributions
First, workers’ mobility within- and between-industries is a topic scarcely
analyzed in the literature.
Second, the data used in this paper represent an original and rich
statistical dataset. One of the most original features of our dataset is that
it permits to examine the origin and destination of a representative
sample of French workers.
And third, we lead an analysis in terms of networks of firms.
COINVEST Conference, Lisbon 2010
Presentation summary
Section I: Introduction
Section II: Data
Section III: Econometric strategy
Section IV: Results
Section V: Conclusion
COINVEST Conference, Lisbon 2010
Introduction
Data
Econometric strategy
Results
Conclusion
Firm networks of mobility
Within- and between- industry mobility
A Simple Framework for Firm Networks of Mobility (1/2)
Starting with Granovetter’s (1973) observations, researchers have
been extremely active in identifying the role of networks on the labor
market. A flurry of papers shows how workers find jobs using friends,
relatives (see for instance Munshi, 2003), or all sorts of ties available to
them.
Indeed most of the ties that were examined are person-based. Very
little attention has been devoted to other links, in particular those
stemming from the employing firm. Not surprisingly, this type of
information is essentially unavailable.
Here, we try to give a comprehensive view of various potential
networks as well as measures of their prevalence and their respective
weights in the mobility episodes.
COINVEST Conference, Lisbon 2010
Introduction
Data
Econometric strategy
Results
Conclusion
Firm networks of mobility
Within- and between- industry mobility
A Simple Framework for Firm Networks of Mobility (2/2)
Types of resources available in the data: Industry resources as measured by
within industry (at two-digit level) mobility; Employer network as measured by the
type of ownership: independent firm versus group (with affiliates); Local resources
as measured by within region or “department” (county) mobility; Type of firm
network as measured by firm size and by economic health.
We define firm networks of mobility as follows: Two firms belong to the same
network if they are similar according to certain attributes (industry, geographic
location, ownership, size or economic health). For example, we consider that two
firms belong to an industry network if they operate in the same industry.
We study how the industry network of mobility is determined by other four
networks of mobility: networks of size, department, ownership and economic
health.
COINVEST Conference, Lisbon 2010
Introduction
Data
Econometric strategy
Results
Conclusion
Firm networks of mobility
Within- and between- industry mobility
Within- and Between- Industry Mobility: Related Literature
Previous research on labor mobility has focused primarily on mobility between
jobs (employers, firms) and on the relation between job seniority and earnings
and worker-firm matching.
There are fewer studies that directly focus on industry mobility:
-
-
Parrado et al. (2007) use the PSID. They only use an industry variable (the
size of industry)
McLaughlin and Bils (2001) also use the PSID. They do not use firm data to
control for industry heterogeneity.
Shin (2007) uses the NSLY79 and he exclusively controls for worker variables.
Topel and Murphy (1987) use the CPS and show that both cyclical and secular
increases in US male unemployment have been accompanied by declines in
between-industry mobility.
Le Minez (2002) studies industry mobilities between 1968 and 1998 and she
does not include firm variables.
COINVEST Conference, Lisbon 2010
Introduction
Data sets
Data
Econometric strategy
Results
Conclusion
Indicators
Data
We used three data sets:
the “Déclarations Annuelles de Données Sociales ” (DADS) from 1991 to 1999: All workers born
in October of an even year are included. The data include all private and semi-public employers.
the “Liaisons Financières” (LIFI) data base. For each head of group, all companies belonging to
the head and most financial relationships between firms can be traced.
the “Bénéfices Réels Normaux” (BRN) files. These data give us measures of employment,
value-added, profits…
By merging the four data bases we finally work with more than 4 millions
observations which cover the 1991-1999 period. We work with a data set
containing observations only for people staying employed on the period of
analysis.
COINVEST Conference, Lisbon 2010
Introduction
Data sets
Data
Econometric strategy
Results
Conclusion
Indicators
Tree diagram of job mobility
COINVEST Conference, Lisbon 2010
Introduction
Data
Econometric strategy
Results
Conclusion
Data sets
Indicators
Indicators
•For the firm of origin (or the last firm where the worker was employed in),
we construct the following variables:
•Variables of individual characteristics: sex, age, skill, wage,
number of previous mobilities and working time duration.
•Variables of firm characteristics: industry, size, ownership,
economic health, region.
COINVEST Conference, Lisbon 2010
Introduction
Data
Econometric strategy
Results
Conclusion
Data sets
Indicators
Indicators
•For the workers who are mobile during a year, we use information from
firms of origin and destination. We construct the following variables:
•Variables of individual characteristics: number of days of nonemployment between two jobs, skill (sskill), wage (w+ and w-),
working time duration (sduration and duration_sduration)
•Variables of firm characteristics: industry (within), ownership
(indep, indepd, indepf, mmg, m_grp), region (sdep, reg_sdep),
size (ssize), economic health (q3q3, q3q1, q1q3, q1q1).
COINVEST Conference, Lisbon 2010
Introduction
Data
Econometric strategy
Results
Conclusion
Econometric strategy
The idea that factors affecting selection into the sample may
simultaneously affect the binary outcome of interest has been the
motivation for the introduction of the probit sample selection model.
The model we use was initially developed by van De Ven and van
Praag (1981).
COINVEST Conference, Lisbon 2010
Introduction
Data
Econometric strategy
Results
Conclusion
Econometric strategy
The selection equation
The first equation describes the probability of selection: the probability
that a worker changes job within the year.
z are explanatory exogenous variables : sex, age, skill, wage, working
time duration, industry, size, ownership, economic health and previous
number of mobilities.
COINVEST Conference, Lisbon 2010
Introduction
Data
Econometric strategy
Results
Conclusion
Econometric strategy
The within-industry mobility equation – it is defined only if
mobile=1:
The
probability
of
changing
job
within
the
same
industry
The x explanatory variables are: sex, age, skill, wage, working time
duration, industry, size, ownership, economic health, previous number of
mobilities and dummies indicating the way the worker changes job
between the firms of origin and destination (in terms of skill, wage, working
time duration, firm size, firm ownership, firm economic health and firm
location)
COINVEST Conference, Lisbon 2010
Introduction
Data
Econometric strategy
Results
Conclusion
Econometric strategy
δ and β are suitable vectors of unknown regression parameters
 (ε1, ε2) is a zero-mean unit-variance bivariate normal random variable with
corr (ε1, ε2) =ρ.
The contribution to the likelihood of the i-th unit of the sample can be written as
follows:
COINVEST Conference, Lisbon 2010
Introduction
Descriptive results
Data
Econometric strategy
Results
Conclusion
Estimation results
Descriptive results (1/2)
32,6 % of individuals are mobile at least once in any given year.
Mobility occurs more often in services industries.
Workers employed in independent (resp. group) firms have a
tendency to move to independent (resp. group) firms.
Workers tend to move more often in the Province rather than in the
Paris region.
COINVEST Conference, Lisbon 2010
Introduction
Descriptive results
Data
Econometric strategy
Results
Conclusion
Estimation results
Descriptive results (2/2)
Mobility is decreasing with age and skills.
30% of moves between firms have no non-employment spell between
the two jobs and another 30% need at least 6 months to find another job.
Two-third of all episodes are made by high-mobility workers (an
average of four episodes).
COINVEST Conference, Lisbon 2010
Introduction
Description of the sample
Data
Econometric strategy
Results
Conclusion
Estimation results
Selection regression estimates for being mobile
COINVEST Conference, Lisbon 2010
Introduction
Description of the sample
Data
Econometric strategy
Results
Conclusion
Estimation results
Main equation: estimates for within-industry mobility
COINVEST Conference, Lisbon 2010
Introduction
Data
Econometric strategy
Results
Conclusion
Conclusion
Our analysis permits to control, in the same time for individual and firm
characteristics.
Robustness tests:
On different subsamples obtained by focusing at one year at the time
or on extracting randomly from them 10% of their observations.
Results are identical
Our main result shows that the probability of being mobile within the same
industry increases whether the worker changes job between two similar
firms in terms of size, region, economic health or ownership.
COINVEST Conference, Lisbon 2010
Thank you very much for your attention!
COINVEST Conference, Lisbon 2010