Worker and job flows
ABSTRACT This paper uses a unique matched panel of firms and workers to describe worker and job flows. It finds a great deal of variation in both worker and job flows by sector: it also finds that job flows are not synonymous with worker flows. Contracting firms continue to hire workers; expanding firms have workers who exit. Older and larger firms have systematically lower rates of job creation and destruction, as well as of worker separations.
- SourceAvailable from: Steven J Davis[show abstract] [hide abstract]
ABSTRACT: This paper investigates how job creation and destruction behavior varies by employer size in the U.S. manufacturing sector during the period 1972 to 1988. The paper also evaluates the empirical basis for conventional claims about the job-creating prowess of small businesses. The chief findings and conclusions fall into five categories: (1) Conventional wisdom about the job-creating prowess of small business rests on misleading interpretations of the data. (2) Many previous studies of the job creation process rely upon data that are not suitable for drawing inferences about the relationship between employer size and job creation. (3) Large plants and firms account for most newly-created and newly-destroyed manufacturing jobs. (4) Survival rates for new and existing manufacturing jobs increase sharply with employer size. Smaller manufacturing firms and plants exhibit sharply higher gross rates of job creation but not higher net rates. Copyright 1996 by Kluwer Academic PublishersSmall Business Economics 02/1996; 8(4):297-315. · 1.55 Impact Factor
- [show abstract] [hide abstract]
ABSTRACT: Previous emperical analyses of job mobility focus on worker rather than firm characteristics. This paper exploits a unique data set on enterprise employment. We describe sectoral difference in turnover rates and in the persistence of turnover. We also present evidence of persistent turnover differends at the level of individual firms--a result that is expected if firms have managers with differing ability to screen workers. When we consider the consequences to the firm of such turnover, we discover that high turnover firms are less likely to survive.Review of Industrial Organization 03/1996; · 0.48 Impact Factor
Worker- and Job-Flows in Mexico
David S. Kaplan*
Departamento de Economía and
Centro de Investigación Económica
Instituto Tecnológico Autónomo de México (ITAM)
Work phone: +52 55 5628 4000, ext. 2915
Fax: +52 55 5628 4058
Gabriel Martínez González
Interamerican Social Security Conference
Department of Economics
* Corresponding author.
Although statistics on worker flows and job flows can be critical for policy evaluations,
relatively few studies generate these statistics for developing countries. Using a census of all
workers in private establishments in the formal sector in Mexico that allows us to track workers
and establishments over time, we present the first statistics on worker flows and job flows for
Mexico. We discuss the policy implications of our results, compare our results to the literature,
illustrate how these statistics change during times of reform and crisis, and present novel
findings that contribute to the broader literature on worker reallocations.
Keywords: Job Flows, Worker Flows, Mexico
We gratefully acknowledge financial support from the Inter-American Development
Bank for the project “Market Institutions, Labor Market Dynamics, Growth and Productivity: An
Analysis of Latin America and the Caribbean.” We also acknowledge financial support from the
Asociación Mexicana de Cultura.
The last twenty years have been a time of great reform and change for Latin America.
Privatization, trade liberalization, and structural reform have induced significant resource
reallocations between and within nearly all sectors of the economy. These changes, however,
can easily be masked by the net employment changes that appear frequently in popular
publications. The purpose of this paper is to examine labor reallocations in Mexico at the micro
level, focussing wherever possible on the policy implications of our results.
Davis and Haltiwanger's (1999) survey of job creation and destruction illustrates that
even if an economy has no net change in aggregate employment, microeconomic data from a
broad range of (developed) countries would reveal large employment expansions and
contractions at the firm or establishment level. Hammermesh et al. (1996) demonstrate that
employment expansions and contractions at the establishment level studied in works such as
Davis and Haltiwanger (1990 and 1992) are only a part of the employment-reallocation process.
In particular, they find that half of all hiring occurs in establishments that are not expanding and
that half of all separations occur in establishments that are not contracting. These results imply
that when we focus on establishment-level employment expansions and contractions (which we
will refer to as job flows) we miss a substantial amount of employee-level turnover (which we
will refer to as worker flows).1
The fact that we observe the identities of individual employees is a crucial advantage
over many of the data sets used to study employment reallocations. As a result of this relatively
rare feature of the data, we can examine both worker flows and job flows. We can therefore
calculate more complete measures of labor reallocations (as in Hamermesh et al. 1996) that
incorporate within-establishment employee turnover as well as labor reallocations across
establishments. We find, for instance, that job flows represent roughly half of total worker flows.
We will also present evidence that examining worker flows might reveal important policy-
induced changes in worker reallocations that statistics on job flows might hide.
Prior to the IDB project “Market Institutions, Labor Market Dynamics, Growth and
Productivity: An Analysis of Latin America and the Caribbean,” of which this paper is a part,
nearly all2 of the literature on worker flows or job flows had focused on developed countries.
Understanding these gross worker and job flows, however, is especially critical in developing
countries that seem to consistently experience larger shocks. In this paper, we present evidence
that various shocks and policy reforms have indeed had an impact on worker and job flows.
Davis and Haltiwanger (1998) offer twelve reasons why it is important to understand
gross worker and job flows, but there are at least three additional reasons why understanding
these flows in developing countries in general, and Mexico in particular, have important
implications for policy.3 The first reason is that the timing changes in these flows may be linked
to policy reforms, which would provide a way to evaluate the impacts of these reforms. We find,
for example, that worker flows increased considerably during the late 1980s when the General
Agreement on Trade and Tariffs (GATT) was being implemented and when restrictions on
foreign investment were being relaxed. We find no similar trend for job flows in this period,
which highlights the importance of observing within establishment turnover unrelated to changes
in an establishment’s total number of employees. We also find that worker flows and job flows
increased considerably during the period 1997-2001, possibly due to NAFTA or possibly do to a
1997 reform of the pension system that was designed to reduce rigidities and increase
employment in the formal sector.
A second reason why an understanding of worker flows or job flows can have important
policy implications is that this understanding can be used to enhance the efficiency of policies
designed to aid dislocated workers. Mexico's training program PROBECAT (Programa de
Becas de Capacitactión para Trabajadores), for example, uses a nonlinear selection mechanism
to determine eligibility [see Revenga, Riboud, and Tan (1994) for more details], which could be
tailored to reflect demographic and geographic groups most in need of help to find employment.
We find, for example, that people between the ages of 60 and 65 are hired at rates that are only
slightly lower than those for younger workers. That is, although these older workers separate
from their firms at substantially higher rates, many of these people attempt to find new
employment. A program like PROBECAT might be well served by recognizing the re-
employment demands of older workers, especially in light of demographic trends that indicate
that older workers will constitute a growing percentage of the workforce.
A third reason why an understanding of worker flows or job flows can have important
policy implications is that these results can help determine the kinds of policies that might best
foster employment growth. In our study, for example, we find that small firms suffered
disproportionate employment declines during the economic crisis of 1995, during which the
credit market essentially shut down. We also find that, unlike in developed countries, younger
firms in all periods tend to have lower percentage employment growth than older firms. To the
extent that smaller firms and younger firms are more susceptible to poorly functioning credit
markets,4 these results provide microeconomic evidence consistent with Tornell, Westerman, and
Martínez (2004) who suggest that Mexico’s relatively slow growth is related to a lack of credit.
To reflect the needs of policy makers in these areas and to provide the most
comprehensive picture possible of gross worker flows and job flows, we focus on three
important dimensions: firm and worker characteristics, time, and geographic space. Apart from
the result that younger firms exhibit lower net-growth rates than older firms, many of our results
are largely consistent with the stylized facts that emerge from the existing developed-country
literature. For example, older establishments and larger establishments exhibit lower rates of job
turnover and the majority of job creation and destruction is due to expansions and contractions in
continuing establishments. The entry and exit of establishments, however, account for a non-
trivial share of job creation and destruction. Furthermore, the majority of job reallocations occur
within rather than between industries.
We also examine variation in worker flows and job flows over time and across regions.
Perhaps the most important finding is that worker flows and job flows have generally been
increasing over time. We also find striking changes over time in the net-growth rates of specific
groups of firms. For example, we find that large manufacturing establishments and
manufacturing establishments close to the northern border experienced dramatic employment
declines in 2001. This result may be linked to the concurrent significant decline in U.S.
manufacturing activity and is suggestive that opening a small economy to trade with a much
larger economy can expose the small economy to increased volatility. The results from 2001 are
also noteworthy for their stark contrast with the previously mentioned results from the economic
crisis of 1995 when large establishments dramatically outperformed small establishments.
The format of the rest of the paper is as follows. In the next section, we describe how we
created our worker- and job-flows data set from social-security records in Mexico. In section 3,
we describe the methodology we use to calculate worker- and job-flows statistics. In section 4,
we present the statistics. We include results separated by establishment age, establishment size,
gender, and employee age. Many of these results will have important implications for policy. In
section 5 we add some final concluding remarks.
2. Creation of a Job- and Worker-Flows Data Set for Mexico
The raw data come from the Mexican Social Security Institute (Instituto Mexicano del
Seguro Social, or IMSS), which is the agency that manages the social-security accounts for all
private-sector tax-registered workers in Mexico. Since filing with the IMSS has been used as a
criterion for formal sector participation,5 the data can be thought of as a census of formal-sector
establishments in the private sector. The IMSS uses its own 4-digit industry classification system
consisting of 271 separate industries that span all economic activity in the formal sector.
Unfortunately, if an employee leaves the formal (tax-registered) sector, we are unable to observe
if the employee becomes unemployed or finds a job in the informal sector.
Individual records in the raw data contain an identifying number for the person, an
identifying code for the establishment, the daily wage, the date when the information of this
record became valid, and the date when the information stopped being valid. If the worker leaves
the establishment, the old record is closed. If the worker’s salary changes, the old record is
closed and a new record is opened with the updated wage information but with the same
identifier for the establishment. Importantly, we have both an establishment identifier and a
person identifier that are consistently coded over time. Our first step was to convert this
information into annual information. We chose December 31 as the date for which we would
extract the relevant information each year from 1985 to 2001.6
For each December 31 of the 17 years for which we have data, we selected the records
that were applicable to the particular date. If a person had two apparently applicable records from
the same establishment, we chose the record with the later start date. If a person had two
applicable records from different establishments, we assumed the person really was working in
both establishments. We only selected workers with strictly positive wages. This restriction
mainly excludes students from the database, many of whom are insured by the IMSS although
they are not really employees.
The files mentioned above include wage and employment histories of all workers
registered with the IMSS. The files also contain the age and gender for nearly all workers. We
also merged in industry and location information of the establishment using separate files
provided by the IMSS. The match rate was nearly 100%.
Since we are using a new data set, we believe it is useful to look at some simple statistics
and compare them to official statistics of the IMSS. We do this comparison in table 1, although it
should be noted that we made no attempt to replicate the precise methodology used in the
generation of these official statistics. In fact, we do not know the precise methodology used by
the IMSS although we are using the same raw data.
The first employment figures in table 1 are official IMSS statistics on cotizantes on
December 31 of each year. 7 Cotizantes are employees who pay social-security taxes or for
whom social-security taxes are paid. Of the official statistics we have found, we suspect that
these statistics use the definition that most closely matches our definition of all individuals who
receive positive salaries. Unfortunately, however, we could not find statistics on cotizantes
(Table 1 here)
The second set of employment observations presents our calculations of employment on
December 31 of each year. The figures match up fairly well, although the net-growth percentages
in our data tend to be slightly smaller than in the official statistics. The third set of statistics gives
our counts of “jobs” which will correspond to our worker- and job-flows statistics. The
difference between the statistics on jobs and employment is that one employee may have more
than one job.
Our data represent all sectors of the Mexican economy, but, as an additional check, we
also compared our 1993 average employment in manufacturing with the 1993 average total
employment in the 1993 Mexican Industrial Census. One would expect the majority of
employees in manufacturing to be formally registered, implying that manufacturing employment
registered with the IMSS should be similar to manufacturing employment recorded in a
manufacturing census. Our 1993 manufacturing employment is 2,836,277 and the 1993 Census
manufacturing employment is 3,246,039, suggesting that our data cover about 87.4% of total
manufacturing employment. Based on these comparisons, we believe that our data are reliable.
We now turn to our methodology for studying job and worker flows. To facilitate
comparison with the developed countries that dominate the existing literature, we use established
definitions of both job flows and worker flows. We begin with the methodology for our worker-
flows statistics. When an establishment hires a new employee, we refer to this event as an
accession. For a given year, we define the accession percentage according to the following
acc, is the number of employees in establishment j in year t who were not working in
establishment j in year
empl, is the number of employees of establishment j in year t,
is the number of employees of establishment j in year
. Similarly we define
the separation percentage as
sep, is the number of employees in establishment j in year 1
who were not working
in establishment j in year t. It is now natural to define the net-growth percentage in
employment, which is simply
Our two statistics on worker flows,
sepper give us information of
reallocations of people within and across establishments. As we mentioned in the introduction,
however, it is also common to examine reallocations of jobs across establishments.8 Job flows
statistics give us information about establishment-level changes in employment without taking
into consideration the identities of the employees. For example, consider an establishment in
which five employees have left since the last year and were replaced by five new employees. We
would say that this establishment experienced worker flows in the form of five accessions and
five separations. Since total employment has not changed, however, we would say that the
establishment neither created nor destroyed jobs.
More precisely, define net employment growth in establishment j and period t as
Now denote job creation in establishment j and period t as
and denote job destruction in establishment j and period t as
We can now define the job-creation percentage and job-destruction percentage in period t as
It should be clear that statistics on job flows and statistics on worker flows are related. If
an establishment increases its total employment by one, at least one current employee must be
new. If an establishment reduces its total employment by one, at least one employee must have
left. In this sense, statistics of job flows give us a lower bound on our worker-flows statistics.
Along these lines, we will now explain our decomposition of worker flows into two
components: the component explained by job flows and the “excess” component. First, we will
define the sum of worker flows
as our summary measure of worker flows. Similarly, we will define our summary measure of job
As we mentioned earlier, the sum of job flows (
sumjf ) can be thought of as a component of
worker flows (
sumwf ). Our definition of “excess” worker flows will simply be
In words, excess worker flows are the worker flows not accounted for by job flows.
One common practice in the literature on job flows is to separate jobs created by births
(establishments that had zero employment in the previous year) from jobs created by expansions
(establishments that had positive employment in the previous year and expanded). Similarly, it is
common to distinguish jobs destroyed by deaths (establishments whose employment fell to zero)
from jobs destroyed by contractions (establishments that reduced employment but continue to
employ at least one employee). Our data are particularly well suited for studying births and
deaths because we observe all establishments, no matter how small they are. We do not, for
example, only observe establishments only when they cross some employment-size threshold.
It is also common to decompose the sum of job creation and destruction (
sumjf ) into an
aggregate component, an industry component, and an idiosyncratic component. The aggregate
component is simply the absolute value of the net-growth percentage. If the economy-wide
employment increases (decreases) by five percent, we know the job-creation percentage (job-
destruction percentage) must be at least five percent.
The industry component of the decomposition is the component of job creation and
destruction that can be explained by industries simultaneously expanding and contracting.
Specifically it is the average across industries of the absolute value of each industry’s net-growth
percentage, weighted by the industry’s average employment, minus the absolute value of the
economy-wide net-growth percentage. If all industries are expanding or if all industries are
contracting, this component is equal to zero.
The final component of the decomposition, the idiosyncratic component, is the
component of job creation and destruction that arises from establishments simultaneously
expanding and contracting within the same industry. It can be defined as the sum of job flows
minus the aggregate component minus the industry component. In our data this is always the
largest component of the job-flows decomposition.
4. Results on Job and Worker Flows
Finally we are ready to see some results. Table 2 presents all of the statistics on worker
flows and job flows discussed in the methodology section for each year 1986-2001. Table 2 is
the central table of our paper and we will discuss its implications quite thoroughly.
(Table 2 here)
Job flows in Mexico on average appear to be not very different from job flows in the
U.S., although both job creation and job destruction are somewhat higher in Mexico. For
example, using data from West Virginia, Spletzer (2000), finds an annual job-creation
percentage of 15.8%, which is lower than the 19.0% we observe on average in our data. Spletzer
reports an annual job-destruction percentage of 14.4%, which is slightly lower than our figure of
If one compares the manufacturing sector in Mexico to the results from Davis and
Haltiwanger (1992), we again find that both job creation and job destruction are higher in
Mexico. If we restrict the sample to manufacturing establishments, we get an average job-
creation percentage of 16.4%, substantially higher than the figure of 9.2% from Davis and
Haltiwanger. The average figure for job destruction in Mexico is 12.8%, marginally higher than
the 11.3% from Davis and Haltiwanger. Tables analogous to table 2 calculated separately for the
manufacturing and non-manufacturing sectors are available upon request.
As in Hamermesh et al. (1996) and Abowd et al. (1999), we find that a substantial share
of worker flows cannot be accounted for by job flows. As described in the previous section, we
can summarize worker flows by using the sum of the accession percentage and the separation
percentage. The average of this statistic in our data is 71.3%. We can similarly summarize job
flows by using the sum of the percent of jobs created and the percent of jobs destroyed. The
average of this statistic in our data is 33.8%. Job flows therefore account for slightly less than
half of total worker flows.
We now turn to the changes in job flows and worker flows over time. The period we
study encompasses several important reforms, policies, and economic events in Mexico. Mexico
joined the General Agreement on Tariffs and Trade on January 1, 1986 and implemented deep
tariff cuts. A peso devaluation in 1987 was followed by an economic "Solidarity Pact" that
effectively reduced inflation from over 100% per year. Foreign investment laws were liberalized
in 1988, 1989, and 1990 and the new laws induced a rapid inflow of foreign capital. In 1990,
Mexico announced it was pursuing a free trade agreement with the United States (with Canada to
join the negotiations soon thereafter). The North American Free Trade Agreement was signed in
1992 and went into effect in January 1994. The peso crashed in December 1994 and was
followed by a very deep, but relatively brief, recession that was followed by a four-year
recovery. We consider our results in the context of these changes.
It is interesting to note that the pace of job flows and worker flows has been increasing
over time, although not in a linear fashion. From roughly 1986-1990, the pace of worker flows
was accelerating. In fact, both the accession percentage and the separation percentage increased
from 1988 to 1989, and again from 1989 to 1990. The pace of job flows was fairly flat over this
period, which highlights the importance of observing worker flows, which are a more complete
measure of reallocations than job flows.9 The timing of these accelerations in worker flows is
consistent with the hypothesis that inflows of foreign capital and the implementation of GATT
led to an increase in worker turnover.