“How do older displaced workers get by: re-employment ......“How do older displaced workers get...
Transcript of “How do older displaced workers get by: re-employment ......“How do older displaced workers get...
“How do older displaced workers get by: re-employment, early retirement, or
disability benefits?”
Ross Finnie School of International and Public Affairs
University of Ottawa 55 Laurier Ave. East
Ottawa, ON K1N 6N5
David Gray Department of economics
University of Ottawa 55 Laurier Ave. East
Ottawa, ON K1N 6N5 [email protected]
December 2011
JEL codes: J63, J65, J68, J26 Key words: involuntary displacement, social insurance receipt, labour market transitions, retirement
Abstract
The central objective is to investigate the income sources and patterns of prime-age and older workers having a high degree of attachment who suffer a layoff. Using a unique data base that links the event of layoff with data on earnings, we track all of their sources of income over an interval that spans four years prior to the separation to five years after it. We conduct an econometric analysis of the incidence of several alternative destinations following the event of layoff, such as early retirement, re-employment, self-employment, or reliance on social insurance. The most common destination state for prime-age and older workers who have not yet reached normal retirement age are early retirement and continued labour market activity. It is rare for them to draw on social insurance benefits, and we find little evidence that disability benefits and workers’ compensation are functioning as disguised unemployment benefits to any great extent. We also estimate an econometric model of the probability of remaining in the labour force that includes a control group of similar workers who were laid off. We find that older laid-off workers are more likely than their non-displaced counterparts to withdraw from the labour force before reaching the normal retirement age. This effect becomes stronger with the passage of time since the year of layoff.
1
I. Introduction
Due to the demographic phenomenon of an aging labour force in Canada, a
greater share of displaced workers will be over 50 years old. As mentioned in Kuhn
(2003), such workers are likely to be less mobile and ‘…less likely to flow in response to
variety of shocks which affect the economy. Also, it is well-known that involuntary job
turnover becomes more costly with age...’ (p. 9) Some of the contributing factors appear
to be the accumulation of firm-specific, industry-specific, or occupation-specific skills,
the deterioration of alternative skills as the length of prior job tenure increases, and
possible discrimination on the part of would-be employers.
While older displaced workers typically face limited job opportunities relative to
their younger counterparts, they have a wider variety of options; the latter group normally
has no option other than to search for a new job. For the older displaced workers, in
contrast, there are several alternative destination states that involve some form of labour
force withdrawal. These transitions are detrimental to achieving greater labour market
participation among older workers, a policy objective that the OECD (2003,2006)
believes is both attainable and desirable.
The topic of this paper is the post-displacement labor market outcomes of prime-
age and older workers. For both genders we investigate their post-displacement profiles,
which can involve a number of alternative destinations and sequences of states. There is
a large literature that deals with post-displacement outcomes, but many of those studies
deal with outcomes occurring within a short-run time frame, such as the wage of the first
newly found job and the length of joblessness immediately following the layoff. There
are relatively few studies that track the post-displacement labour market profiles for
2
extended time periods in order to measure the longer-term consequences. Those studies
that do involve a long-term time horizon tend to treat the outcomes of wage and earnings
losses. We seek to fill this void by measuring and analyzing the entry of older laid-off
workers into a number of alternative destination states, which can be categorized into
several types, namely i) the receipt of some form of social insurance, ii) early-retirement,
iii) self-employment, and iv) re-employment (at potentially lower earnings).
Post-displacement transitions into the receipt of social insurance benefits, such as
subsequent UI claims, social assistance, and long-term disability have garnered attention
in the US literature. Autor and Duggan (2006) claim that the social security disability
regime has evolved from its original function of insuring workers against earnings loss
attributable to a medical condition to providing long-term income support for the quasi-
unemployable – a phenomenon that they label ‘non-employability insurance’. We search
for similar patterns for the incidence of receipt of the Canadian equivalent to the US
program, namely the Canadian Pension Plan Disability (CPPD) Regime and its Quebec
equivalent, the QPPD.
The empirical approach is based on following cohorts of workers who were laid
off in a given reference year. The outcome variable takes the form of the proportion of
workers that are observed in a given destination state during each of the first five post-
layoff years. In previous studies (Finnie and Gray (2009,2011)), we provide descriptive
statistical analyses based on a typology of post-displacement destination states and
configurations of income from various sources. In this paper we extend the scope by
estimating multinomial econometric models of their relative likelihoods of occurrence
conditional on experiencing a prior layoff. Following the methodological approach of
3
Chan and Stevens (1999,2001) and Ichino et al. (2007), we also examine the impact of
layoff on the post-displacement probabilities of the outcome of strong labour force
attachment relative to the outcomes of control groups of workers who were not displaced.
We draw primarily on the Longitudinal Administrative Database (LAD), which is
a panel of annual data derived from tax declarations. It contains detailed, disaggregated,
and accurate information on the sources of income as well as the respective levels.
Another advantage of this file is that it is very encompassing and is quite representative
of the Canadian working-age population. Attrition via migration to other provinces is not
a problem with the LAD file due to its national scope. It allows one to observe and
follow people over long time periods, facilitating an accurate assessment of pre-
displacement earnings as well as the post-displacement labour market activity profile.
The fourth advantage of the LAD file is its tremendous size.
There are also a number of shortcomings of the LAD file. First, one cannot
observe the event of a job separation, and so it must be imputed. Second, as is common
with administrative data sets, there is no information pertaining to several variables that
are important for labour market outcomes, such as education and skill level, tenure at a
given job, the number of jobs held over an interval, the length of time worked over an
interval, or wage rates. A third disadvantage of the LAD file is that the frequency of the
data is annual. Given that the layoff can occur at any point in time over a calendar year,
this feature raises timing issues as far as the reporting of income and the point of
displacement is concerned.
II. Survey of the Literature
4
The most comprehensive study on worker displacement in Canada is by Abe et al.
(2002), which is based on the Canadian Out-of-Employment Panel (COEP) survey from
the mid 1990s. This file consists of a representative sample of job separators, which are
followed over a fairly short time horizon of about 15 months. These authors estimate re-
employment hazards of jobless workers as well as wage changes for those who are re-
employed.
Morissette et al. (2007) deal with displaced workers in Canada over the interval
from 1988-1997. They focus on the earnings losses of displaced workers based on an
administrative data source called the Longitudinal Worker File. The data allow for the
identification of laid-off workers and for a comparison their post-layoff outcomes to
those observed for a control group of workers who have not been laid off – a strategy that
we adopt for part of our analysis. The results indicate that high-seniority workers who
are displaced from either mass layoffs or total plant closures suffer earnings losses that
are substantial and persistent. Unlike our analysis, they restrict their sample to workers
under 49 years of age at the time of layoff in order to omit any cases of early retirement.
The methodology mentioned in the paragraph above was developed by Jacobson
et al. (1993). That influential US study was one of the first to investigate earnings
outcomes of displaced workers over a time horizon spanning periods long before and
long after the point of displacement. Perhaps more importantly, they included in their
analysis a comparison group of workers who were not displaced, which essentially
amounts to a difference-in-difference approach to estimating wage losses that better
accounts for the counterfactual earnings of displaced workers. While the focus of our
5
paper is not on wage losses, our data set has some of the same features, and we employ a
variation of their approach for analyzing displacement costs.
Another recent Canadian study is Neill and Schirle (2008), who estimate earnings
losses of displaced workers between the ages of 50 and 65 years based on longitudinal
data drawn from the Survey of Labour Income Dynamics (SLID) over the interval 1993-
2005. Unlike many of the studies in the literature, they account for the outcome of early
retirement. Their estimates suggest that male workers of all ages suffer substantial and
persistent earnings losses stemming from displacement, of which the length of tenure at
the prior job plays a more important role than does the worker’s age per se.
Chen and Morissette (2010) is the most recent study treating displaced workers in
Canada. Their estimating sample covers the period between 1979 and 2004 and is
restricted to workers between 50 and 54 years old. Their objective is to determine the
extent to which post-displacement employment patterns – including employment rates
and average earnings losses - of older displaced workers evolved over that period. This
study does not involve any analysis of counterfactual outcomes.
To our knowledge, a series of articles by Chan and Stevens (1999,2001,2004) are
the mostly closely related studies from the US literature. Based on a specialized survey
drawn from the Health and Retirement Study, with three waves dating from 1992, 1994,
and 1996, these authors track the post-displacement employment probabilities of older
displaced workers over extended time horizons. This framework allows them to take
account of subsequent events such as job instability and early retirement. They estimate
hazard models for exiting from the spell of joblessness following displacement as well as
exiting from a new, post-displacement job. Their equations control for age and the
6
number of months already spent in the respective initial states. They discern sharp
differences in these post-displacement profiles between those who were laid off at a
given point of time and a control group of counterparts who were not. For example,
experiencing job loss after age 50 roughly doubles the annual probability of retirement in
subsequent time periods. There are substantial and lasting gaps in employment rates
between the two groups, reflecting both the initial period of non-employment after job
loss and the subsequent instability of post-displacement jobs. They also conclude that
while displaced older workers face a reduced incentive to work and a greater incentive to
retire following job loss (through the channel of pensions), the barriers to re-employment
(through the channel of unattractive labour market opportunities) are the primary
explanatory factor for relatively low re-employment rates.
Based on Austrian administrative data, Ichino et al. (2007) model the post-
displacement employment rates of workers, comparing their outcomes with those of a
control group of matched non-displaced workers. They determine that both displaced
workers of prime-age and of older ages have significantly lower employment rates
relative to their respective control groups. We estimate a somewhat similar equation in
the second component of our econometric analysis.
Drawing on data from four countries contained within the European Community
Household Panel, Tatsarimos (2010) estimates hazard models (based on inflow samples
of jobless older workers) in which the effect of job displacement on transitions into
retirement and into the state of re-employment within a competing hazards framework. .
He distinguishes between the short and the long-run effects of displacement, and he pays
7
particular interest to the role of institutions such as government-financed early retirement
regimes.
III. Data Issues and Methodology
III.1 Sample Selection
The population consists of prime-age and older workers having a relatively high
degree of labour market attachment. This group’s employment pattern has been stable
and uninterrupted over a relatively long time period; we label it a ‘clean’ employment
record prior to the event of a layoff. Given the breadth of the LAD’s coverage, there are
many subjects whom we omit from the working sample, particularly those with a low
degree of labour market attachment as well as those working in fragmented, seasonal,
and/or part-year jobs. More specifically, we only include person-year observations for
which the subject is between the ages of 45 and 64 years for all of the years in the
sampling window. We omit all observations for which the subject is a full-time student,
which can be readily identified.1 Because our focus is on displaced workers, we omit
those subjects who appear to be primarily self-employed by deleting all those
observations that involve a non-trivial amount of self-employment income, defined as
that exceeding $1,000 expressed in constant 2005 dollars. The observations for non-filers
are also deleted.
In order to identify workers with a ‘clean’ employment record, we have to
observe their activity over an extended period of time. To this end, we construct four-
year windows of data. The LAD file became very representative of the labour force
starting in 1992, before which many low-income individuals did not have a strong 1 Given the age-related criteria, this case should not cause many observations to be deleted.
8
incentive to file a tax return, and thus were not necessarily sampled during every year.
With the implementation of the GST tax credit in 1991, however, almost all working-age
adults have an incentive to file a tax return. Given this seam in the LAD file, we
commence the sample selection process in 1992.2 We subsequently follow the cohort of
all subjects that are observed in the LAD file in 1992 (and met the age-related selection
criteria) for four consecutive years up to and including 1995. Once these years of data
have been observed for an individual, that four-element vector for the first cohort is
retained. We repeat this procedure for 1993, for which we will sample new entrants in
addition to all of those subjects from the 1992 cohort that are observed in 1993. This
window of observation closes in 1996, and we thus form the second cohort of subjects.
These two cohorts share most of their subjects. This procedure is then repeated for each
year from 1992 until 1998, which yields a total of 7 cohorts, the last of which has an
employment record from 1998 to 2001. While we continue to follow subjects over the
years 1998-2005, no new entrants are added to the working sample after that cohort.
While we could form later cohorts, it is not possible to observe the post-displacement
outcomes for adequately long periods.
At this stage of the sample selection process, the working sample consists of four-
year vectors of observations for most subjects appearing in the LAD file.3 The next step
is to inspect each four-year vector of observations that we have for a worker in order to
select those with ‘clean’ employment records, who by construction have a stable
2 There is another reason related to reporting issues for which we do not deal with data before 1992. The details for that choice are given in the appendix. 3 For any subject having four or more consecutive years of observations within our interval of 1992-1998, there will be potentially several vectors, some of which will overlap. He/she will be sampled multiple times, but each case involves a different window of observation. It is also possible for an individual to have several spells of employment activity separated by an episode of labour force withdrawal. In such a case, the subject will be sampled more than once, and the windows drawn from the two periods will be disjoint.
9
employment history over the window of observation. To this end we scan the four
consecutive annual observations that we have for him/her for certain outcomes that lead
to deletion. An annual observation is flagged if the earnings fall below a level of $15,000
(2005 dollars). That threshold was chosen because it represents a quasi lower bound on
labour income that would be earned by a full-time, full-year worker. We also flag any
annual observation if income from any of the following sources is reported: social
assistance benefits, employment insurance (EI) benefits, Canada Pension Plan (CPP)
benefits (both pension and disability income), or workers’ compensation benefits. We
then delete any four-year vector in which any flag appears. This procedure should cause
repeat users of EI to be dropped.
After the cohorts of individuals with ‘clean’ four-year employment records have
been formed, the unit of observation thus becomes an individual having that
characteristic. Those cohorts comprise the risk set for the event of experiencing a layoff
in the following year. We label the first year immediately following the ‘clean’ spell of
employment year T, which becomes the reference year for observing and recording the
event of displacement. If a layoff occurs during that year, we then follow the individual
over the subsequent five-year interval. It is over that ex post window of observation that
the post-displacement outcomes are observed and the analysis of their income sources is
carried out. The structure for the data set is summarized in text Table 1.
Text Table 1 – Structure of the cohorts of workers with uninterrupted employment records
Cohort number Window of
observation for ‘clean’ four-year employment record
Year of potential layoff
Years of post-layoff observations – five-year intervals
1 1992-1995 1996 1997-2001
10
2 3 4 5 6 7
1993-1996 1994-1997 1995-1998 1996-1999 1997-2000 1998-2001
1997 1998 1999 2000 2001 2002
1998-2002 1999-2003 2000-2004 2001-2005 2002-2005 2003-2005
More details on the construction of the sample are provided in appendix Table 1,
the columns of which correspond to each calendar year, while each row indicates the
number of remaining observations after each cut is made. For each year, we commence
with over 4 million person-year data points in the full LAD file. We then retain the
approximately 61 % who filed a return. Next we delete all observations for which the
age of the subject falls outside of the 45 to 64 year range. A few observations are
deleted because either the subject died or left Canada. The exclusion of the self-
employed observations results in the deletion of 3 to 4 % of the observations from the
original sample. The earnings restriction that we apply in order to form the ‘clean
employment’ record nearly cuts the remaining sample in half. Approximately one-third
of these workers that meet the earnings criterion are dropped because there was some
receipt of social insurance income. A few additional observations are dropped because
the subject was a student or it turned out that he/she received special (i.e. sickness,
maternity, or parental) EI benefits. The estimating samples are then selected as a 10 %
draw, leaving estimating samples of between 31,000 and 43,000 observations for each
year.
III.2 Identification of a Separation
This task requires that our sample selected from the LAD file be linked to an
administrative data set generated by Human Resources and Skills Development Canada
11
called the Status Vector and Record of Employment file (STVC/ROE). The flag for the
event of a layoff is the observation of the act of collecting income from the Employment
Insurance (EI) system, as reported in the LAD file. It is necessary, however, to
distinguish between the various types of EI benefits that might be reported in the LAD
file. If an individual declares receiving EI benefits on his/her tax return, it could reflect
‘regular’ benefits stemming from an unemployment spell experienced after a layoff, or it
could reflect ‘special’ benefits, such as maternity, parental, and sickness benefits that are
unrelated to a layoff. In order to verify that the EI income stems from a layoff, we use
the merged file. There is a line in the STVC/ROE record that identifies the ‘type’ of
benefit in the EI claim, and thus makes the distinction between ‘regular’ and ‘special’
benefits. We identify the subject as displaced if he/she declares EI/UI income during that
year, and if cross-verification with the STVC/ROE file indicates that these were ‘regular’
EI/UI benefits.4
We note that this procedure for identifying the event of displacement will capture
only a subset of all displaced workers. For instance, there are some displaced workers
who either suffer no unemployment at all, and others who experience only brief spells of
it. They therefore will not file an EI/UI claim, and the layoff will not be observed. In
order for an individual to be identified as displaced in our sample, the jobless spell
following displacement had to have lasted at least three weeks (due to the universal
4 If the EI benefit period spans two contiguous calendar years, then some EI income will be declared for both years. In this instance, the separation must have occurred in the prior year. This ambiguity does not affect our measurement of a layoff, however, because during the earlier year of this pattern, the subject would not have fulfilled the requirement of a ‘clean employment’ record. Our sample selection procedure thus ensures that the separation occurs during the same year that we flag the record due to the receipt of EI income. We could certainly observe EI income during the subsequent year.
12
application of a two-week waiting period for EI benefits).5 Note that the problem of
missing very short spells of joblessness also exists in the articles in the literature on
displaced workers that are based on retrospective surveys such as the US Displaced
Worker Survey.
There is also the case of displaced workers who do experience a non-trivial spell
of unemployment but who do not qualify for EI benefits. In that case the layoff will not
be observed. Most of the displaced workers literature is, however, oriented around
workers with a high degree of attachment to the labour force, and most such workers
would qualify for EI benefits. This conventional practice applies to our sample in light of
the selection criteria that we apply, and therefore we observe most workers of this type
who experience a layoff.6 In the case of seasonal workers, who typically do file claims
for UI/EI benefits, most of them are excluded from the working sample because they did
not satisfy the sampling criterion of a ‘clean employment’ record.
Another pertinent case is an individual who is temporarily laid off from his/her
position. Unless the jobless spell is quite short, this worker has a strong incentive to file
a claim for EI/UI benefits, and given our sampling criteria, is likely to qualify. The
sampling criteria should ensure that he/she is an individual with a stable employment
history for whom a layoff is an infrequent event. In the event that he/she is recalled to
the former employer, we will observe employment income in subsequent time periods.
As our data set contains no information on the employer, we cannot distinguish between
5 Most of the displaced workers who meet that criterion and who qualify for EI benefits do have a strong incentive to file a claim. For instance, consider the case of a worker electing to accept a ‘buyout’ or an early retirement package from his/her employer that qualifies for EI benefits. Both parties in this arrangement have an incentive for a third-party – namely the Federal government – to contribute. 6 Any laid-off worker in Canada qualifies for UI/EI benefits if they have 20 or more weeks of employment activity over the 52-week period preceding the layoff. In high-unemployment areas, one qualifies with even shorter employment spells. Virtually all individuals that are selected into our risk set easily meet that criterion
13
the cases of recalls and new hires. Temporary layoffs are therefore treated in the same
way as permanent layoffs. The estimated proportion of all person-year observations (the
risk set for layoff) during which a layoff occurred in 1.92 %.
We note that like much of the existing literature on displaced workers, we take
layoffs to be an exogenous event.7 We do not take account of any selection issues that
would reflect unobservable attributes of workers, with presumably the less able ones
being targeted by firms for layoff. On the other hand, in the interests of reducing the
social and economic costs of mass layoffs, on some occasions workers selected for layoff
are those that have the most generous early retirement benefits or severance pay. They
could also be those with the most attractive outside opportunities in the form of
alternative employment in the labor market. We are thus uncertain of the impact of
possibly endogenous layoffs on our estimates.
To the extent to which there are time patterns in the unobserved attributes of our
sample, the compositional changes will be partially captured in year-specific effects, but
these estimated coefficients could also reflect changes in global labour market conditions,
and thus it is not possible to empirically isolate either pattern.
IV. Empirical Strategy
Our empirical approach uses a typology framework. The first step is to enumerate
the post-layoff patterns and profiles of income sources. We observe all of the income and
its numerous potential sources that are reported during that 5-year period commencing in
year T + 1, which is the first year after the layoff. Based on a tabulation of the flows of
older displaced workers transiting from layoff to the state of receiving income from a 7 One notable exception is Tatsarimos (2010).
14
particular source or from a combination of sources, we develop a typology of income
profiles and configurations that have economic significance and policy relevance. For
each cohort, which is identified by the year of separation, the number of cases for receipt
of each of the types of income (described below) is tabulated for each of the five
subsequent calendar years.8 All of the person-year observations are pooled together
regardless of the year of layoff. At this stage the outcome variable is the raw frequency
of the number of laid off workers that received income from this particular source. This
figure is subsequently divided by the size of cohort in order to generate proportions that
are interpreted as incidence rates. Since many subjects receive income from several
sources simultaneously, these categories are not mutually exclusive, and the proportions
will not sum to unity. With the exception of sources with trivial amounts, we enumerate
every source of income that is received by any subject, so the coverage should be
exhaustive. These potential sources are listed in text table 2.
Text Table 2 – Possible sources of income and other destinations for laid-off workers
1) labour market earnings 2) self-employment income 3) employment insurance 4) social assistance 5) workers’ compensation 6) CPP/QPP and/or OAS and/or GIS (public pension income) 7) private pension income 8) CPP – Disability regime 9) other type of income (property income, capital income) Missing or censored observations 10) zero income 11) non-filer
8 With the exception of the latest two cohorts, we can follow all of them for five years after the point of layoff. Cohorts # 6 and # 7 can be followed for 4 and 3 years, respectively, after the point of layoff, until 2005.
15
12) attrition from sample due to age restrictions or FT student status
The values for these incidence rates are presented in Tables 2 and 3 of Finnie and
Gray (2009). This procedure generates a great deal of empirical detail; many workers
exhibit complex profiles. For any given post-separation year, there are potentially many
distinct states resulting from combinations of income received simultaneously from those
nine different sources. If one were to enumerate all of the possible combinations, the
permutations would number in the thousands. In the interests of econometric tractability
and ease of exposition, these states of receiving a certain type of income are collapsed
into a typology of income configurations.
For the typology analysis, the central thrust is the following question: during the
post-displacement years, what was the principal source of income? How is the worker
getting by, if indeed he/she has managed to replace a substantial share of his/her pre-
displacement income? The principal source of income is defined as the source that
accounts for 50 % or more of the subject’s total income, provided that such a source
exists. For each type of income, we calculate the proportion of the group of laid-off
workers that received it as their primary income source. For any subject-year observation
for which there are either one or two sources of income, there must necessarily be a
principal source.
The nine categories of income listed above are not mutually exclusive for the
purposes of the typology analysis, as there are many cases in which the income sources
are diversified such that no single source accounts for more than 50 % of the subject’s
total income during that year. Out of all the residual observations that do not fit into one
of those categories according to the 50 % criterion, we define further types that are
16
defined according to whether the following combinations of sources when combined
account for 50 % or more of the total income: i) earnings plus self-employment income
plus any combination of social insurance benefits (EI, SA, WC, CPPD), and ii) public
pension plus private pension. Finally, there is a residual category. The observations are
assigned into these categories in the sequence that is listed in text table 2, noting that the
composition of the groups, and especially the catch-all, residual category, is sensitive to
that order. All of the possible configurations are partitioned into 12 different types (listed
in text table 3). The types are constructed to be mutually exclusive and exhaustive, and
thus the shares sum to unity.
The residual category accounts for between 0 and 4 % of all observations
(depending on the year), but in most years, fewer than 1 % of the observations did not fall
into the 11 specific classifications. Observations are omitted, and are thus not assigned a
type, for any of the three following reasons: they have total income falling below the
$5,000 threshold, they did not file a return that year, or they aged out of the working
sample.
Text Table 3: Types of Income Patterns
Configurations of income received, defined by 50 % of more of the total received from: 1. earnings 2. EI 3. public pension 4. private pension 5. workers’ compensation (WC) 6. CPPD 7. social assistance 8. investment income 9. earnings and/or self-employment 10. earnings and/or self-employment plus any combination of EI, SA, WC, and/or CPPD (mostly EI) 11. any combination of pension income, public and/or private
17
12. residual category
We extend the scope of the descriptive typology analysis presented in Finnie and
Gray (2011) by estimating an econometric model of the shares of subjects relying on a
given source of income. A multi-variate equation can account for some of the
compositional effects that would confound uni-variate analysis, in particular the impact
of age, which is critical in the determination of retirement behaviour. We are particularly
interested in the influence of three variables, namely the number of years elapsed since
layoff, the calendar year, and regional labour market conditions. The coefficients for the
first set of variables should be identified through variation of years elapsed since layoff
across individuals of the same age. The second set of regressors is specified as a flexible
form of year-specific binary variables, while the effect of age is specified in flexible form
as a set of binary indicator for each year between the ages of 45-63).
In the interests of econometric tractability, we estimate the multi-nomial logit
model based on fewer and hence broader categories than the twelve that are enumerated
above. The outcomes for the dependent variable (collapsed from the twelve types
enumerated above) are reliance on: i) social insurance (3.4 % of person-year
observations), ii) earnings (69.6 % of cases), iii) earnings plus self-employment (2.3 % of
cases), iv) public and/or private pension (13.7 % of cases), and v) the residual group (2.8
% of cases).9 The post-displacement profiles are captured by a set of binary variables for
the number of elapsed years since layoff (T2, T3, T4, T5 for years T + 2, T + 3, T + 4,
and T + 5, respectively). The remaining exogenous variables are listed in text table 4.
Text Table 4 – List of Covariates for the Regression equation
9 6.5 % of the observations are not included in the regression analysis because they are non-filers, and 1.7 % of cases are excluded because their total incomes fell below $5,000.
18
Variable Description and Categories Calendar year 1997-2004 (1997 omitted) Age at point of layoff in years Separate effects for 45-63 years (45
omitted) Gender Separate equations Family structure Single (omitted), married no children,
married with children, lone parent Union membership before layoff Binary indicator SMA size (area size of residence) greater than 500,000 (omitted), 100,000-
499,999, 30,000-99,999, 15,000-29,999, small urban area, rural area
English/French speaker (minority language status)
English in Quebec, French outside of Quebec, majority language (omitted)
Province 10 provinces (Ontario omitted) Regional unemployment rate Continuous value for each of 51 regions
The second component of the econometric analysis seeks to assess the impact of
the event of layoff on a post-displacement outcome utilizing a treatment group of
displaced workers and a control group of counterparts who were not laid off during the
first year of the respective cohort. This specification is somewhat similar to the one in
Ichino et al. (2007), in which the dependent variable is binary and is related to post-
displacement employment activity. Whereas those authors model the employment status
that is observed during a quarter, we address the somewhat narrower question of whether
the worker relied primarily on employment activity instead of receipt of income from
alternative sources) during a post-displacement year. If he/she derived over half of their
total income from employment activity, this variable assumes a value of unity.
As mentioned in our previous work, approximately 2 % of the total worker-year
observations in any calendar year represent layoffs. Although the values may well be
higher during recession years, during our interval the event of layoff for a highly attached
worker is low. The inclusion of all the observations that do not involve displacement
19
results in huge estimating samples: almost 6,000,000 for men and approximately
4,700,000 for women. By sample selection procedures, the treatment group and the
comparison group have labour market histories (before the reference year) that are
roughly similar, albeit according to our broad definition of exhibiting the ‘clean
employment window’. Because there is no difference in the outcome variable (which is
essentially labour market attachment) during the time period prior to the reference year,
we do not estimate a difference-in-difference specification. In order to control to some
extent for pre-displacement variation in labour market activity, we include the level of
prior earnings (using a set of categorical variables) observed during the 4-year interval
leading up to the year of displacement.10 The impact of layoff on post-displacement
income sources is identified by a comparison of the conditional mean values of the
dependent variable between the two groups. The indicator is a binary variable assuming
a value of unity if the worker was laid off during the first year of his/her cohort, and zero
otherwise. This dummy variable enters as an intercept shift effect, but it is also interacted
with the number of elapsed years since the point of layoff in order to discern whether the
impact of being laid dissipates along that dimension. The layoff dummy variable is also
interacted with the age-bracket of the worker at the point of layoff. We estimate both the
linear probability model via least squares and the logit model.
VI. Results The results of a typology analysis consisting only of descriptive statistics are
presented in Finnie and Gray (2008,2011). As a lead-in to our multi-variate analysis, a
10 The set of categorical variables corresponds to the following earnings brackets denominated in constant 2005 $ (annual values averaged over the 4-year period): below 25,000, 25,000-40,000, 40,000-55,000, 55,000-70,000, 70,000-85,000, 85,000-100,000, above 100,000.
20
brief synopsis is presented here. During the estimation interval, approximately two-thirds
of the 45-59 year olds and but less than one-third of the over-59 group gained over 50 %
of their income from the labour market. As might be expected, these proportions
diminish with the elapsed time since layoff. The values for women are slightly lower
than those for their male counterparts. While a significant minority of workers does file
at least one EI claim subsequent to the initial layoff, in any given year, less than 2 % of
the group becomes dependent on the EI regime as their primary source of income.
Hardly any subject meets our criterion of dependence on the social insurance programs of
social assistance (SA), workers’ compensation (WC), or Canada Pension Plan Disability
(CPPD). Only about 1 % of the group of laid-off workers becomes dependent on the
CPPD regime, although the incidence is slightly higher among those over 59 years of age.
These findings are not consistent with the literature drawn from the USA that was cited
above, but the eligibility conditions for the Canadian program are extremely stringent.
Instead of that option, the displaced workers in our sample are far more likely to be
dependent on either early retirement income financed by the public pension schemes (up
to 20 % and 28 % of the older group of men and women, respectively), or by private
pension schemes (up to 39 % and 37 % of the older group of men and women,
respectively). A further 1 to 10 % derives more than half of their post-layoff income
from a combination of public and private pension benefits.
The primary regression results for the multi-nomial logit model (those consisting
of the estimated coefficients) are available from the authors. The calculations for the
more intuitive marginal probability effects are reported in Tables 1 and 2 for men and
women, respectively. The values are interpreted as absolute deviations of probabilities
21
relative to the average caused by a unit increase in that regressor. Each column
corresponds to one of the five possible types, and these marginal probabilities
horizontally sum to zero. These magnitudes can be compared to the univariate shares for
each type that were listed above: 0.034 for social insurance, 0.696 for earnings, 0.023 for
self-employment income, 0.137 for pension income, and 0.028 for the residual category.
Due to the abundance of estimates, our discussion is limited to the more marked
patterns that we discern. Relative to childless single workers, married couples are less
likely to rely on social insurance income and more likely to rely on self-employment
income. The impact of the age variable for both genders is essentially as expected; the
younger workers are more likely to rely on earnings or self-employment income, but are
less likely to rely on pension income. Starting at around age 52, a quasi-monotonic trend
of decreasing probabilities for that outcome is discerned for both genders, and the
magnitude is higher among women. Starting at around age 48, the opposite pattern is
discerned for the type of relying on pension income, and for many ages the magnitudes
are almost offsetting. The estimated coefficients pertaining to the social insurance
outcome (not including pensions) are mostly insignificant among both men and women.
Many of the point estimates for the area-size-of-residence variables are significant
for the outcomes of reliance on self-employment income and pension income. Relative
to regions with more than 500,000 people, being situated in a smaller SMA is associated
with a lesser reliance on self-employment income, and either a greater reliance on
pension income or on labour market income (not statistically significant). No patterns are
discerned for reliance on social insurance income. The provincial effects are most
notable for Quebec and to a lesser extent, Alberta and BC. Relative to Ontario, both men
22
and women in Quebec are less likely to rely on either earnings or self-employment
income, and more likely to rely on pension income. In BC the opposite pattern applies,
and in Alberta the propensity to retire on either public or private pensions is relatively
low. The rate of unemployment prevailing in the economic region has the expected
positive impact on reliance on social insurance and retirement income, and the expected
negative impact on reliance on self-employment income for both genders. On the other
hand, reliance on labour market earnings surprisingly increases with the unemployment
rate, and is even significant for men.
The impact of union representation is very similar for both genders. Relative to
their non-union counterparts, unionized workers are less likely to rely on income received
from either social insurance programs, pension regimes, or self-employment, but are
more likely to rely on employment earnings. This pattern is suggestive of a higher degree
of labour force attachment among previously unionized workers for reasons that are not
apparent. A possible interpretation is that some of them were recalled to their previous
jobs.
We do not expect to capture much in the way of cyclical effects from the calendar
year covariates, as the period from 1997 to 2005 was characterized by favourable
macroeconomic conditions, with the exception of a slowdown in 2002. There are certain
time trends that are discerned which are interpreted as deviations from 1997. Among
women there appears to be a declining trend in reliance on social insurance income over
this time period, and among men to a lesser extent. For both genders there is a more-or-
less monotonic trend towards greater reliance on earnings over this period, although this
23
is often estimated imprecisely. Among men there is a monotonic trend toward less
reliance on pension income, but for women this pattern emerges only after 2001.
The rows at the bottom of Tables 1 and 2 contain the point estimates for the
regressor of the elapsed years since layoff, which are interpreted as deviation from year T
+ 1 (the year after the layoff). Net of age effects, for both genders there is a tendency to
rely less on social insurance income during years T + 2 through T + 5, but there is very
little estimated difference between those years. The probability of that outcome does not
rise with elapsed time since the point of layoff. For both genders, reliance on earnings is
greater two years after layoff, but that effect reverts back to the values in year T + 1
thereafter. There is some evidence of both men and women turning to self-employment
during years T + 2 through T + 5. As one might expect, we discern a steadily increasing
tendency to rely of pension income for both men and women.
The regression results for the equations that include a control group are presented
in tables 3-5. The full set of estimates for the logit model as well as the marginal
probability effects are listed in table 3, while those for the linear probability model are
listed in table 4. The dependent variable is the event of relying primarily on either
earnings or self-employment income during the reference year (we label this state
‘working’ below). The treatment variable is the layoff indicator, which assumes a value
of unity if the worker was laid off during the cohort year and a value of zero otherwise.
In the specifications listed in columns 3,4,7, and 8 of Table 3, it is interacted with the
variables of the number of years elapsed since layoff as well as age. For the sake of
parsimony, the age variable is collapsed into four categories: 40-44 (omitted), 45-49, 50-
54, and 55-59.
24
With the exception of the age variables, the results from the logit model are
qualitatively and (in the case of the marginal probability effects) often quantitatively
similar to those of the linear probability model. Controlling for prior earnings has little
impact on the estimates, but it does raise the degree of explanatory power slightly. We
first mention the estimated coefficients not involving the layoff variable, which are
identified overwhelmingly by the non lay-off group. For men, all other (pre-layoff)
earnings categories other than the omitted category of 39-52 K are less likely to work,
with the largest impact between 55-100 K. Among women there is a similar pattern, but
the magnitudes are higher. The pure (non interacted) age effects are monotonic and as
expected; working becomes less likely at older ages, and the magnitude is almost
identical across genders. The impact of area size of residence is usually increasing; the
higher the density, the more likely they are to be working. The impact of residing in
Western Canada relative to Ontario tends to be positive, but it is remarkably negative in
Quebec and Newfoundland. The unemployment rate has the expected negative impact,
although the magnitude is low. The other regional indicators (i.e. province and area size
of residence) might be capturing some of the impact of local labour market conditions.
The impact of having a unionized job is positive, but that is a partially mechanical effect
given the predominance of non-laid off workers in our sample. The trait of speaking
English in Quebec has high positive impact, while the trait of speaking French outside of
Quebec has the opposite effect. The effect associated with the calendar year is gradually
increasing relative to the baseline year of 1996. The monotonic pattern is similar across
genders.
25
The effects associated with the event of lay off are summarized in table 5. All of
these values are derived from figures contained in Tables 3 and 4. The intercept layoff
parameter – corresponding to the year after layoff (T + 1) - has a strong negative impact.
The marginal probability effect is estimated to be -0.39 for men and -0.47 for women in
the logit equation and -0.38 and -0.49 in the linear probability model. This point estimate
is much higher when we include all of the interacted terms.
The interacted effect with elapsed time since layoff (T2,T3,T4, and T5) is relative
to year T + 1. For the laid-off group (columns 3-4 and 7-8, bottom panel), the probability
of working in year T + 2 is higher than it is in year T + 1. From year T + 2 and
thereafter, we discern monotonic decreases in the probability of working for both genders
across both specifications. The point estimate in year T + 5 is close to the baseline value
in year T + 1. The gap between the layoff group and the non-layoff group, as captured in
the interacted term, grows wider and wider with elapsed time. The patterns among
women are somewhat sharper
The interacted effects for age groups are relative to the oldest category of 60-64
years. The oldest group of laid-off workers is the least likely to be working relative to
their non-laid off counterparts, and for both genders effects are monotonically decreasing
in age. The gap between the layoff and the non layoff group, actually becomes negative
in the case of 45-49 year-old group; the laid-off workers of this age group are actually
estimated to be more likely to be working than their non laid-off counterparts in the
baseline year of T + 1. For the laid-off group (columns 3-4 and 7-8, top panel), the
negative impact of that event emerges after workers have turned 50. The gap between
26
laid-off and non laid-off workers is highest for the 50-54 years old group with the
exception of the logit equation based on the male sample.
V. Conclusions
The central objective of this study is to investigate the income sources and
income-receipt patterns of prime-age and older workers after having suffered a layoff.
We focus on a set of cohorts comprised of both male and female workers who are
deemed to have a high degree of attachment to the labour force and uninterrupted
employment activity preceding that event. Using a unique administrative data base, we
track all of their sources of income over a five-year interval after the layoff. This
provides ample information on the sources of income which are typically not considered
in the existing displaced workers literature, such as pension income, social insurance
payments, self-employment income, and labour market earnings.
The most common destination states for prime-age and older laid-off workers
who have not yet reached normal retirement age are reliance on private pension benefits
and continued labour market activity. It is relatively rare for them to draw heavily on
social insurance benefits. In contrast to the case of the USA, we find little evidence that
disability benefits and workers compensation are functioning as disguised unemployment
insurance benefits.
A multinomial logit regression analysis of a typology of the principal source of
income for the laid off worker revealed a number of empirical patterns. Workers situated
in larger SMAs are more likely to rely on earnings and self-employment activity than are
their counterparts in more rural areas. Workers situated in Quebec are less likely than
27
their Ontario-based counterparts to rely on earnings or self-employment income, and are
more likely to depend on retirement income, while the opposite pattern applies to the two
western-most provinces. Higher regional unemployment rates are associated with a
higher reliance on social insurance income, but no significant relationship was discerned
for the outcomes involving earnings or early retirement. Unionized workers are more
likely than their non-union counterparts to rely on earnings, and less likely to rely on
income received from either pensions or social insurance regimes. Over the interval from
1997 to 2005, we discerned monotonic trends of an increasing degree of reliance on
earnings and self-employment income, and a decreasing reliance on social insurance
income. There is no evidence of a tendency to rely more on social insurance income with
the passage of time since layoff, but for the period between 3 to 5 years after layoff, there
is a marked decline in reliance on earnings coupled with an increase in the degree of
reliance on pensions.
The existing literature on older and middle-aged displaced workers has addressed
the evolution of their employment rates and pathways into early retirement for the United
States and several European countries. Our analysis that includes a comparison group of
non-displaced workers indicates that laid-off workers are much less likely to be working
in the (baseline) year right after the year of lay off: the marginal probability effect is
approximately – 0.4 men and almost - 0.5 for women. Two years after the point of
layoff, the marginal probability effect (relative to the baseline year) is positive, but five
years after the year of layoff, they are no more likely to rely on labour market or self-
employment income than they were during the baseline year. The discrepancy in the
employment rates between the laid-off group and the comparison group widens over
28
elapsed time since layoff. Once workers turn 50, the event of suffering a layoff renders
them less and less likely to work as they age. Workers who are laid off between the ages
of 45-50, however, are actually more likely to be working than their counterparts who
were not laid-off.
29
References
Abe, P. Kuhn, Y. Higuchi, M. Nakamura, A. Sweetman (2002) “Worker Displacement in Japan and Canada” in Kuhn, P., editor, Losing Work, Moving On: International Perspectives on Worker Displacement W.E. Upjohn Institute for Employment Research
Autor, D. and M. Duggan (2006) “The Growth in the Social Security Disability Rolls: A
Fiscal Crisis Unfolding” Journal of Economic Perspectives 20, 3, pp. 71-96 Chan, S. and A. Huff-Stevens (1999) “Employment and Retirement Following a Late-
Career Job Loss” American Economics Association Papers and Proceedings, Vol. 89, May, pp. 211-16
Chan, S. and A. Huff-Stevens (2001) “Job Loss and Employment Patterns of Older
Workers” Journal of Labor Economics Vol. 19, 2, pp. 484-521 Chan, S. and A. Huff-Stevens (2004) “How Does Job Loss Affect the Timing of
Retirement?” Berkeley Journal of Economic Analysis and Policy, Vol. 3, 1, article 5
Chen, W-H and R. Morissette (2010) “Have Employment patterns of Older Displaced
Workers Improved Since the Late 1970s?” Canadian Labour Market and Skills Researcher Network, WP No. 61, May
Finnie, R. and D. Gray (2009) “Displacement of older workers: re-employment, hastened retirement, disability, or other destinations?” Canadian Labour Market and Skills Research Network working paper, No. 15
Finnie and Gray (2011) “ Labour-Force Participation of Older Displaced workers in Canada: Should I Stay or Should I go?” IRPP (Institute for Research on Public Policy) Study No. 15, February
Ichino, A., G. Schwerdt, R. Winter-Ebmer, and J. Zweimuller (2007) “Too Old to Work,
Too Young to Retire?” Institute for the Study of Labour (IZA) Working Paper 3110
Jacobson, L., R. Lalonde, and D. Sullivan (1993) “Earnings Losses of Displaced Workers” American Economic Review 83, 4, pp. 685-709
Kuhn, P. (2003) “Effects of Population Aging on Labour Market Flows in Canada: Analytical Issues and Research Priorities” Issues paper, Skills Research Initiative Partnership – Expert Roundtable on Labour Market Adjustments due to population aging in Canada”
30
Morissette, R., X. Zhang, and M. Frenette (2007) “Earnings Losses of Displaced Workers: Canadian Evidence from a Large Administrative Database on Firm Closures and Mass Layoffs” working paper No. 291, Statistics Canada, Business and Labour Market Analysis Division
Neill, C. and T. Schirle (2008) “Remain, Retrain, or Retire: Options for Older Workers
Following Job Loss” in Abbot, Beach, Boadway, and MacKinnon, eds. Retirement Policy Issues in Canada Montreal: McGill-Queens’ University Press, pp. 277-308
OECD Employment Review (2003) Transforming Disability into Ability; Policies to Promote Work and Income Security for Disabled People Paris: Organization for Economic Cooperation and Development
Tatsiramos, K. (2010) “The Effect of Job Displacement on the Transitions to
Employment and Early Retirement for Older Workers in Four European Countries” European Economic Review 54,4, pp. 517-535
31
Table 1
Regression Results: Multinomial Logit Model of Principal Source of Income Marginal Probability Effect, Men
reliance on income earned from
social
insurance earnings Self-empl.
Pension income
Other sources
Family Status
Single (omitted category)
Couple no Kids -0.033*** 0.04 0.010*** -0.009 -0.008***
[0.004] [0.031] [0.002] [0.008] [0.003]
Couple with Kids -0.029*** 0.045 0.014*** -
0.023*** -0.007***
[0.004] [0.044] [0.003] [0.004] [0.002]
Lone Parent -0.002 -0.018 0 0.018 0.002
[0.013] [0.107] [0.007] [0.020] [0.009]
Age Category
45 (omitted category)
Age40 -0.006 0.017 0.002 -0.007** -0.005
[0.006] [0.095] [0.006] [0.003] [0.003]
Age41 -0.007 0.014 -0.008 -0.002 0.003
[0.006] [0.096] [0.005] [0.004] [0.005]
Age42 -0.002 0.013 0 -0.007* -0.005
[0.007] [0.095] [0.006] [0.003] [0.003]
Age43 0.004 0.003 -0.004 -0.001 -0.002
[0.007] [0.094] [0.006] [0.004] [0.004]
Age44 -0.002 0.005 -0.002 -0.002 0.001
[0.007] [0.095] [0.006] [0.004] [0.004]
Age46 0 -0.001 0.001 0 0.001
[0.007] [0.094] [0.006] [0.004] [0.004]
Age47 -0.002 0 -0.002 0.005 -0.001
[0.007] [0.097] [0.006] [0.005] [0.004]
Age48 0.005 -0.016 -0.008 0.017** 0.001
[0.008] [0.095] [0.006] [0.007] [0.004]
Age49 0.01 -0.05 0.002 0.028*** 0.010*
[0.008] [0.089] [0.007] [0.009] [0.005]
Age50 0.006 -0.052 -0.009* 0.046*** 0.009*
[0.008] [0.089] [0.006] [0.011] [0.005]
Age51 0.002 -0.097 -0.008 0.091*** 0.012**
[0.007] [0.082] [0.006] [0.016] [0.006]
Age52 0.008 -0.122 -0.013** 0.112*** 0.015**
[0.008] [0.079] [0.005] [0.018] [0.006]
Age53 0.009 -0.159** -0.012** 0.144*** 0.018***
[0.008] [0.074] [0.006] [0.021] [0.006]
Age54 0.007 -
0.196*** -0.012** 0.175*** 0.026***
[0.008] [0.068] [0.006] [0.024] [0.007]
Age55 0.020** -
0.228*** -
0.022*** 0.207*** 0.023***
32
[0.009] [0.067] [0.005] [0.027] [0.007]
Age56 0.004 -
0.277*** -
0.028*** 0.269*** 0.031***
[0.008] [0.063] [0.004] [0.033] [0.008]
Age57 0.004 -
0.318*** -
0.022*** 0.297*** 0.040***
[0.008] [0.058] [0.006] [0.036] [0.009]
Age58 0.007 -
0.362*** -
0.028*** 0.346*** 0.037***
[0.008] [0.054] [0.005] [0.040] [0.009]
Age59 0.009 -
0.420*** -
0.033*** 0.394*** 0.050***
[0.009] [0.048] [0.005] [0.045] [0.010]
Age60 0.011 -
0.467*** -
0.033*** 0.431*** 0.058***
[0.009] [0.042] [0.005] [0.047] [0.011]
Age61 0.009 -
0.421*** -
0.035*** 0.386*** 0.062***
[0.010] [0.056] [0.005] [0.049] [0.013]
Age62 0.017 -
0.464*** -
0.040*** 0.440*** 0.047***
[0.012] [0.062] [0.005] [0.066] [0.013]
Age63 0.016 -
0.453*** -
0.043*** 0.401*** 0.080***
[0.015] [0.083] [0.004] [0.074] [0.021]
Area Size of Residence
500 000+ (omitted category)
100 000 499 999 -0.003 -0.004
-0.012*** 0.020*** -0.002
[0.004] [0.039] [0.003] [0.007] [0.003]
30 000 - 99 999 0.004 -0.007 -
0.014*** 0.022** -0.006**
[0.005] [0.048] [0.003] [0.009] [0.003]
15 000 - 29 999 -0.002 0.006 -
0.015*** 0.009 0.003
[0.007] [0.077] [0.005] [0.013] [0.005]
1 000 - 14 999 -0.001 -0.001 -
0.013*** 0.016* -0.002
[0.005] [0.051] [0.003] [0.009] [0.003]
Less than 1000 0.003 -0.02 -0.005* 0.018** 0.005
[0.004] [0.040] [0.003] [0.007] [0.003]
Province
ON (omitted category)
NF -0.003 -0.075 0.01 0.051** 0.018*
[0.010] [0.130] [0.013] [0.026] [0.010]
PE -0.013 -0.064 0.02 0.039 0.018
[0.014] [0.218] [0.026] [0.038] [0.015]
NS -0.008 -0.067 -0.006 0.071*** 0.011**
[0.006] [0.073] [0.005] [0.017] [0.005]
NB 0.005 -0.034 -0.003 0.024 0.008
[0.008] [0.097] [0.007] [0.016] [0.006]
PQ 0.009** -0.054 -
0.007*** 0.036*** 0.015***
33
[0.004] [0.038] [0.003] [0.007] [0.003]
MB 0 -0.004 -
0.015*** 0.021 -0.003
[0.007] [0.087] [0.005] [0.015] [0.004]
SK -0.005 0.016 -0.005 -0.012 0.006
[0.008] [0.107] [0.007] [0.015] [0.006]
AB -0.012*** 0.039 -
0.013*** -0.018** 0.003
[0.004] [0.057] [0.003] [0.007] [0.003]
BC 0 -0.002 0.016*** -
0.018*** 0.004
[0.003] [0.041] [0.004] [0.006] [0.003]
Language
Majority Lang. (omitted category)
English in Qc 0.001 0.03 -0.007 -0.027** 0.003
[0.006] [0.079] [0.005] [0.011] [0.005]
French out. QC. -0.007 -0.013 -0.012 0.004 0.028*
[0.013] [0.174] [0.010] [0.028] [0.016]
Regional labour market
unemployment rate 0.006*** 0.021**
-0.002*** 0.003** 0.002***
[0.001] [0.008] [0.001] [0.001] [0.000]
Union Status
Non-Union (omitted category)
Union -0.057*** 0.208*** -
0.049*** -
0.086*** -0.017***
[0.002] [0.047] [0.002] [0.005] [0.002]
Year
1997 (omitted category)
1998 -0.003 0.021 0.003 -0.004 -0.016***
[0.006] [0.039] [0.003] [0.009] [0.004]
1999 -0.009* 0.038 0 -0.015 -0.014***
[0.005] [0.046] [0.004] [0.010] [0.004]
2000 -0.012*** 0.062 -0.001 -
0.037*** -0.012***
[0.004] [0.052] [0.004] [0.010] [0.004]
2001 -0.014*** 0.082 0 -
0.056*** -0.012***
[0.004] [0.053] [0.004] [0.010] [0.004]
2002 -0.013*** 0.093 0 -
0.063*** -0.017***
[0.005] [0.057] [0.004] [0.010] [0.003]
2003 -0.012** 0.105* -0.002 -
0.073*** -0.017***
[0.005] [0.060] [0.004] [0.009] [0.003]
2004 -0.007* 0.110* 0 -
0.090*** -0.013***
[0.004] [0.067] [0.005] [0.010] [0.004]
2005 -0.008* 0.124* 0 - -0.012***
34
0.104***
[0.004] [0.073] [0.005] [0.010] [0.004]
T2 -0.082*** 0.052*** 0.007*** 0.040*** -0.017***
[0.002] [0.018] [0.001] [0.003] [0.002]
T3 -0.079*** 0.036 0.008*** 0.054*** -0.018***
[0.002] [0.022] [0.002] [0.004] [0.002]
T4 -0.076*** 0.012 0.006*** 0.076*** -0.018***
[0.003] [0.027] [0.002] [0.006] [0.002]
T5 -0.074*** -0.017 0.007*** 0.102*** -0.018***
[0.003] [0.029] [0.002] [0.008] [0.002]
Standard errors in brackets
*** p<0.01, ** p<0.05, * p<0.1
35
Table 2
Regression Results: Multinomial Logit Model of Principal Source of Income: Marginal Probability Effect, Women
reliance on income earned from
social
insurance Earnings Self-empl.
Pension income Others
Family Status
Single (omitted category)
Couple no Kids -0.011*** -0.007 -0.001 0.013* 0.005**
[0.004] [0.033] [0.002] [0.008] [0.003]
Couple with Kids -0.014*** 0.01 0.003 0.002 -0.002
[0.005] [0.053] [0.003] [0.009] [0.003]
Lone Parent 0.020** 0.017 0.002 -0.037** -0.002
[0.010] [0.075] [0.004] [0.015] [0.006]
Age
45 (omitted category)
Age40 -0.002 0.009 0.01 -0.007* -0.010***
[0.009] [0.118] [0.006] [0.004] [0.004]
Age41 0 0.011 0.007 -0.006 -0.011***
[0.009] [0.121] [0.006] [0.004] [0.004]
Age42 -0.002 0.004 0.007 -0.006 -0.004
[0.009] [0.121] [0.006] [0.004] [0.005]
Age43 -0.007 0.013 0.004 -0.004 -0.007
[0.009] [0.122] [0.006] [0.004] [0.004]
Age44 0.001 -0.001 0.001 0.005 -0.006
[0.009] [0.119] [0.005] [0.006] [0.004]
Age46 0.002 -0.008 0.005 0.005 -0.004
[0.009] [0.116] [0.006] [0.006] [0.005]
Age47 0.01 -0.022 -0.001 0.011* 0.001
[0.010] [0.114] [0.005] [0.007] [0.005]
Age48 0 -0.027 0.005 0.022*** 0
[0.009] [0.113] [0.006] [0.008] [0.005]
Age49 0.006 -0.055 0.001 0.035*** 0.013*
[0.009] [0.109] [0.005] [0.010] [0.007]
Age50 0.008 -0.089 -0.001 0.072*** 0.01
[0.010] [0.101] [0.005] [0.015] [0.007]
Age51 0.004 -0.164* -0.005 0.156*** 0.009
[0.009] [0.089] [0.005] [0.024] [0.006]
Age52 0.001 -0.192** -0.006 0.179*** 0.019**
[0.009] [0.086] [0.005] [0.027] [0.008]
Age53 0.007 -0.230*** 0 0.204*** 0.018**
[0.009] [0.074] [0.005] [0.028] [0.007]
Age54 0.012 -0.262*** -0.002 0.222*** 0.031***
[0.010] [0.072] [0.006] [0.030] [0.009]
Age55 0.005 -0.328*** -0.007 0.291*** 0.039***
[0.009] [0.063] [0.005] [0.036] [0.010]
36
Age56 -0.001 -0.381*** -0.017*** 0.357*** 0.042***
[0.009] [0.059] [0.004] [0.044] [0.011]
Age57 0.002 -0.407*** -0.010* 0.371*** 0.044***
[0.009] [0.056] [0.005] [0.045] [0.011]
Age58 -0.001 -0.483*** -0.017*** 0.436*** 0.066***
[0.009] [0.047] [0.004] [0.051] [0.013]
Age59 0.005 -0.542*** -0.024*** 0.518*** 0.043***
[0.010] [0.041] [0.003] [0.063] [0.011]
Age60 -0.016** -0.550*** -0.020*** 0.530*** 0.056***
[0.008] [0.041] [0.004] [0.065] [0.013]
Age61 -0.017* -0.533*** -0.025*** 0.516*** 0.059***
[0.009] [0.057] [0.004] [0.079] [0.016]
Age62 -0.015 -0.551*** -0.020*** 0.513*** 0.072***
[0.011] [0.060] [0.006] [0.085] [0.020]
Age63 0.013 -0.520*** -0.024*** 0.486*** 0.045**
[0.021] [0.104] [0.006] [0.122] [0.022]
Area Size of Residence
500 000+ (omitted category)
100 000 499 999 0.006 -0.027 -0.004* 0.030*** -0.005
[0.005] [0.046] [0.003] [0.009] [0.003]
30 000 - 99 999 0.008 -0.032 -0.012*** 0.040*** -0.003
[0.006] [0.056] [0.002] [0.012] [0.004]
15 000 - 29 999 0.011 -0.032 -0.006 0.031 -0.004
[0.010] [0.091] [0.005] [0.019] [0.006]
1 000 - 14 999 0.009* -0.046 -0.004 0.040*** 0
[0.006] [0.050] [0.003] [0.012] [0.004]
Less than 1000 0.014*** -0.064 0.002 0.043*** 0.005
[0.005] [0.041] [0.003] [0.010] [0.004]
Province
ON (omitted category)
NF 0.005 0 -0.014* 0.014 -0.006
[0.021] [0.227] [0.008] [0.033] [0.012]
PE -0.023 0.082 -0.011 -0.033 -0.014
[0.020] [0.379] [0.012] [0.042] [0.015]
NS 0.006 -0.013 0 0.01 -0.002
[0.010] [0.104] [0.006] [0.017] [0.007]
NB -0.003 0.015 -0.003 0.004 -0.012*
[0.012] [0.151] [0.007] [0.022] [0.007]
PQ 0 -0.090** -0.007*** 0.106*** -0.009***
[0.005] [0.043] [0.002] [0.011] [0.003]
MB -0.007 0.021 -0.005 0.004 -0.014***
[0.009] [0.107] [0.005] [0.016] [0.005]
SK -0.013 0.037 0.017** -0.036** -0.006
[0.009] [0.113] [0.009] [0.015] [0.008]
AB -0.011** 0.033 0.005 -0.028*** 0.001
[0.005] [0.058] [0.003] [0.008] [0.004]
37
BC 0.002 0.003 0.017*** -0.028*** 0.007*
[0.004] [0.047] [0.004] [0.007] [0.004]
Language
Majority Lang. (omitted category)
English in Qc 0.012 0.024 -0.001 -0.054*** 0.020**
[0.010] [0.099] [0.005] [0.013] [0.008]
French out. QC. 0.005 -0.045 -0.009 0.054 -0.005
[0.019] [0.177] [0.008] [0.046] [0.010]
Regional labour market
unemployment rate 0.005*** 0.011 -0.001** 0.002 0.002***
[0.001] [0.010] [0.001] [0.002] [0.001]
Union Status
Non-Union (omitted category)
Union -0.058*** 0.198*** -0.030*** -0.093*** -0.016***
[0.003] [0.056] [0.001] [0.006] [0.003]
Year
1997 (omitted category)
1998 -0.027*** -0.008 -0.005** 0.059*** -0.020***
[0.008] [0.042] [0.003] [0.012] [0.004]
1999 -0.021*** 0.004 -0.007** 0.043*** -0.019***
[0.006] [0.051] [0.003] [0.013] [0.004]
2000 -0.018*** 0.032 -0.008** 0.009 -0.014***
[0.006] [0.059] [0.004] [0.013] [0.005]
2001 -0.021*** 0.061 -0.010*** -0.012 -0.017***
[0.005] [0.063] [0.004] [0.013] [0.004]
2002 -0.020*** 0.077 -0.011*** -0.023* -0.023***
[0.006] [0.067] [0.004] [0.013] [0.004]
2003 -0.014** 0.098 -0.011*** -0.047*** -0.025***
[0.007] [0.073] [0.004] [0.011] [0.004]
2004 -0.008* 0.108 -0.013*** -0.066*** -0.021***
[0.005] [0.085] [0.005] [0.011] [0.005]
2005 -0.009* 0.126 -0.014** -0.084*** -0.019***
[0.005] [0.092] [0.005] [0.011] [0.005]
T2 -0.139*** 0.099*** 0.012*** 0.049*** -0.020***
[0.003] [0.023] [0.001] [0.005] [0.002]
T3 -0.136*** 0.072** 0.014*** 0.069*** -0.019***
[0.003] [0.028] [0.002] [0.006] [0.002]
T4 -0.137*** 0.039 0.017*** 0.101*** -0.019***
[0.003] [0.031] [0.002] [0.008] [0.003]
T5 -0.139*** 0.005 0.015*** 0.136*** -0.017***
[0.004] [0.035] [0.002] [0.011] [0.003]
38
Standard errors in brackets
*** p<0.01, ** p<0.05, * p<0.1
39
Table 3
Regression Results for Dependence on Labour Market, Treatment versus Control Groups Logit equation Coefficients Marginal Effects
No Interaction Term
Interaction of Treatment with Age
and Time since layoff No Interaction Term
Interaction of Treatment with Age
and Time since layoff
Male Female Male Female Male Female Male Female family status couple no kids 0.146*** -0.009** 0.145*** -0.009** 0.018*** -0.001** 0.017*** -0.001** [0.004] [0.004] [0.004] [0.004] [0.000] [0.000] [0.000] [0.000] couple with kids 0.549*** 0.234*** 0.549*** 0.234*** 0.059*** 0.022*** 0.057*** 0.020*** [0.006] [0.008] [0.006] [0.008] [0.000] [0.001] [0.000] [0.001] lone parent 0.035 0.161*** 0.037* 0.163*** 0.004 0.015*** 0.005* 0.015*** [0.022] [0.015] [0.022] [0.016] [0.003] [0.001] [0.003] [0.001] age group age4549 2.885*** 2.979*** 2.902*** 2.995*** 0.148*** 0.111*** 0.143*** 0.104*** [0.006] [0.007] [0.006] [0.007] [0.000] [0.000] [0.000] [0.000] age5054 1.547*** 1.628*** 1.546*** 1.632*** 0.118*** 0.092*** 0.115*** 0.086*** [0.005] [0.006] [0.005] [0.006] [0.000] [0.000] [0.000] [0.000] age5559 0.445*** 0.569*** 0.439*** 0.573*** 0.049*** 0.046*** 0.047*** 0.044*** [0.005] [0.006] [0.005] [0.006] [0.000] [0.000] [0.000] [0.000] dage4549 (interaction) -0.359*** -0.088** -0.050*** -0.009** [0.032] [0.038] [0.005] [0.004] dage5054 (interaction) 0.174*** 0.205*** 0.020*** 0.018*** [0.030] [0.037] [0.003] [0.003] dage5559 (interaction) 0.285*** 0.120*** 0.032*** 0.011*** [0.030] [0.037] [0.003] [0.003] area size of residence 100,000-499,999 -0.289*** -0.370*** -0.289*** -0.370*** -0.040*** -0.042*** -0.039*** -0.040*** [0.004] [0.005] [0.004] [0.005] [0.001] [0.001] [0.001] [0.001] 30,000-99,999 -0.368*** -0.493*** -0.369*** -0.494*** -0.052*** -0.058*** -0.051*** -0.056*** [0.005] [0.006] [0.005] [0.006] [0.001] [0.001] [0.001] [0.001] 15,000-29,999 -0.303*** -0.542*** -0.304*** -0.542*** -0.042*** -0.065*** -0.041*** -0.062*** [0.008] [0.009] [0.008] [0.009] [0.001] [0.001] [0.001] [0.001] 1,000-14,999 -0.349*** -0.503*** -0.350*** -0.504*** -0.049*** -0.060*** -0.048*** -0.057*** [0.005] [0.006] [0.005] [0.006] [0.001] [0.001] [0.001] [0.001] less than 1,000 -0.438*** -0.587*** -0.439*** -0.587*** -0.063*** -0.071*** -0.062*** -0.068*** [0.004] [0.005] [0.004] [0.005] [0.001] [0.001] [0.001] [0.001] Province NF -0.402*** -0.450*** -0.404*** -0.452*** -0.058*** -0.052*** -0.056*** -0.050*** [0.013] [0.016] [0.013] [0.016] [0.002] [0.002] [0.002] [0.002] PE -0.128*** 0.166*** -0.128*** 0.166*** -0.017*** 0.016*** -0.017*** 0.015*** [0.021] [0.024] [0.021] [0.024] [0.003] [0.002] [0.003] [0.002]
40
NS -0.194*** -0.047*** -0.194*** -0.047*** -0.026*** -0.005*** -0.026*** -0.004*** [0.008] [0.010] [0.008] [0.010] [0.001] [0.001] [0.001] [0.001] NB -0.122*** -0.187*** -0.122*** -0.187*** -0.016*** -0.020*** -0.016*** -0.019*** [0.009] [0.011] [0.009] [0.011] [0.001] [0.001] [0.001] [0.001] PQ -0.422*** -0.523*** -0.422*** -0.525*** -0.061*** -0.062*** -0.059*** -0.060*** [0.004] [0.005] [0.004] [0.005] [0.001] [0.001] [0.001] [0.001] MB -0.184*** -0.155*** -0.184*** -0.154*** -0.025*** -0.016*** -0.024*** -0.015*** [0.007] [0.009] [0.007] [0.009] [0.001] [0.001] [0.001] [0.001] SK 0.109*** 0.117*** 0.109*** 0.117*** 0.013*** 0.011*** 0.013*** 0.011*** [0.009] [0.010] [0.009] [0.010] [0.001] [0.001] [0.001] [0.001] AB 0.212*** 0.168*** 0.212*** 0.168*** 0.025*** 0.016*** 0.025*** 0.015*** [0.005] [0.006] [0.005] [0.006] [0.001] [0.001] [0.001] [0.001] BC 0.085*** 0.048*** 0.085*** 0.049*** 0.011*** 0.005*** 0.010*** 0.005*** [0.005] [0.005] [0.005] [0.005] [0.001] [0.000] [0.001] [0.000] Language English in Qc. 0.461*** 0.471*** 0.462*** 0.473*** 0.051*** 0.040*** 0.049*** 0.038*** [0.009] [0.010] [0.009] [0.010] [0.001] [0.001] [0.001] [0.001] French out Qc. -0.334*** -0.463*** -0.334*** -0.465*** -0.047*** -0.054*** -0.046*** -0.052*** [0.014] [0.015] [0.014] [0.015] [0.002] [0.002] [0.002] [0.002] regional UR -0.018*** -0.021*** -0.018*** -0.021*** -0.002*** -0.002*** -0.002*** -0.002*** [0.001] [0.001] [0.001] [0.001] [0.000] [0.000] [0.000] [0.000] union status Union 1.372*** 1.479*** 1.372*** 1.479*** 0.125*** 0.131*** 0.125*** 0.131*** [0.003] [0.004] [0.003] [0.004] [0.000] [0.000] [0.000] [0.000] Year 1998 0.023** -0.088*** 0.020* -0.094*** 0.003** -0.009*** 0.002* -0.009*** [0.011] [0.013] [0.011] [0.013] [0.001] [0.001] [0.001] [0.001] 1999 0.109*** 0.037*** 0.106*** 0.023* 0.013*** 0.004*** 0.013*** 0.002* [0.011] [0.013] [0.011] [0.013] [0.001] [0.001] [0.001] [0.001] 2000 0.203*** 0.163*** 0.199*** 0.145*** 0.024*** 0.015*** 0.023*** 0.013*** [0.011] [0.013] [0.011] [0.013] [0.001] [0.001] [0.001] [0.001] 2001 0.271*** 0.255*** 0.267*** 0.237*** 0.032*** 0.023*** 0.030*** 0.021*** [0.010] [0.012] [0.010] [0.013] [0.001] [0.001] [0.001] [0.001] 2002 0.312*** 0.324*** 0.310*** 0.308*** 0.036*** 0.029*** 0.035*** 0.026*** [0.010] [0.012] [0.010] [0.013] [0.001] [0.001] [0.001] [0.001] 2003 0.351*** 0.372*** 0.349*** 0.358*** 0.040*** 0.033*** 0.039*** 0.030*** [0.010] [0.012] [0.010] [0.013] [0.001] [0.001] [0.001] [0.001] 2004 0.392*** 0.401*** 0.390*** 0.389*** 0.044*** 0.035*** 0.043*** 0.032*** [0.011] [0.013] [0.011] [0.013] [0.001] [0.001] [0.001] [0.001] 2005 0.445*** 0.446*** 0.443*** 0.436*** 0.049*** 0.038*** 0.048*** 0.035*** [0.011] [0.013] [0.011] [0.013] [0.001] [0.001] [0.001] [0.001] event of layoff -0.968*** -1.313*** -2.033*** -2.649*** -0.159*** -0.195*** -0.386*** -0.470*** [0.009] [0.009] [0.029] [0.037] [0.002] [0.002] [0.006] [0.008] years since layoff T2 -0.503*** -0.537*** -0.540*** -0.605*** -0.074*** -0.064*** -0.078*** -0.070***
41
[0.005] [0.006] [0.005] [0.006] [0.001] [0.001] [0.001] [0.001] T3 -0.909*** -0.973*** -0.955*** -1.050*** -0.148*** -0.132*** -0.154*** -0.140*** [0.005] [0.006] [0.005] [0.006] [0.001] [0.001] [0.001] [0.001] T4 -1.267*** -1.358*** -1.317*** -1.440*** -0.221*** -0.204*** -0.228*** -0.212*** [0.005] [0.006] [0.005] [0.006] [0.001] [0.001] [0.001] [0.001] T5 -1.593*** -1.710*** -1.647*** -1.796*** -0.292*** -0.275*** -0.300*** -0.285*** [0.005] [0.006] [0.005] [0.006] [0.001] [0.001] [0.001] [0.001] dT2 (interacted) 0.912*** 1.263*** 0.083*** 0.076*** [0.024] [0.025] [0.002] [0.001] dT3 (interacted) 1.250*** 1.586*** 0.102*** 0.085*** [0.024] [0.025] [0.001] [0.001] dT4 (interacted) 1.436*** 1.763*** 0.110*** 0.089*** [0.025] [0.026] [0.001] [0.001] dT5 (interacted) 1.587*** 1.907*** 0.116*** 0.092*** [0.027] [0.027] [0.001] [0.001] prior earnings earnings 25-40 k -0.079*** -0.434*** -0.080*** -0.435*** -0.010*** -0.050*** -0.010*** -0.048*** [0.010] [0.006] [0.010] [0.006] [0.001] [0.001] [0.001] [0.001] earnings 40-55 k -0.482*** -0.879*** -0.482*** -0.880*** -0.071*** -0.116*** -0.069*** -0.112*** [0.009] [0.006] [0.009] [0.006] [0.002] [0.001] [0.002] [0.001] earnings 55-70 k -0.850*** -1.251*** -0.850*** -1.252*** -0.136*** -0.183*** -0.133*** -0.176*** [0.009] [0.006] [0.009] [0.006] [0.002] [0.001] [0.002] [0.001] earnings 71-85 k -0.942*** -1.511*** -0.942*** -1.513*** -0.154*** -0.234*** -0.151*** -0.226*** [0.010] [0.008] [0.010] [0.008] [0.002] [0.002] [0.002] [0.001] earnings 85-100 k -0.689*** -0.966*** -0.689*** -0.966*** -0.106*** -0.131*** -0.104*** -0.126*** [0.010] [0.012] [0.010] [0.012] [0.002] [0.002] [0.002] [0.002] earnings 100+ k -0.346*** -0.562*** -0.346*** -0.563*** -0.049*** -0.068*** -0.048*** -0.065*** [0.010] [0.013] [0.010] [0.013] [0.002] [0.002] [0.001] [0.002] Constant 1.357*** 1.665*** 1.400*** 1.741*** [0.015] [0.015] [0.015] [0.016] Observations 5983380 4724375 5983380 4724375 5983380 4724375 5983380 4724375 (Pseudo) R-squared 0.221 0.230 0.223 0.232 0.221 0.230 0.223 0.232 Standard errors in brackets *** p<0.01, ** p<0.05, * p<0.1
42
Table 4
Regression Results for Dependence on Labour Market treatment versus control goups
linear probability model
No Interaction Term
Interaction of Treatment, Age, and
Time since layoff Male Female Male Female family status couple no kids 0.016*** -0.001*** 0.016*** -0.001*** [0.000] [0.000] [0.000] [0.000] couple with kids 0.033*** -0.003*** 0.033*** -0.002*** [0.000] [0.000] [0.000] [0.000] lone parent 0.007*** -0.003*** 0.007*** -0.002*** [0.002] [0.001] [0.002] [0.001] age group age4549 0.315*** 0.329*** 0.310*** 0.323*** [0.001] [0.001] [0.001] [0.001] age5054 0.227*** 0.237*** 0.223*** 0.232*** [0.001] [0.001] [0.001] [0.001] age5559 0.059*** 0.080*** 0.055*** 0.077*** [0.001] [0.001] [0.001] [0.001] dage4549 (interaction) 0.200*** 0.255*** [0.006] [0.006] dage5054 (interaction) 0.164*** 0.163*** [0.006] [0.006] dage5559 (interaction) 0.108*** 0.057*** [0.006] [0.007] area size of residence 100,000-499,999 -0.030*** -0.037*** -0.030*** -0.037*** [0.000] [0.000] [0.000] [0.000] 30,000-99,999 -0.039*** -0.050*** -0.039*** -0.050*** [0.000] [0.001] [0.000] [0.001] 15,000-29,999 -0.033*** -0.056*** -0.033*** -0.056*** [0.001] [0.001] [0.001] [0.001] 1,000-14,999 -0.036*** -0.050*** -0.036*** -0.050*** [0.001] [0.001] [0.001] [0.001] les than 1,000 -0.049*** -0.061*** -0.048*** -0.061*** [0.000] [0.001] [0.000] [0.001] province NF -0.050*** -0.047*** -0.049*** -0.047*** [0.001] [0.002] [0.001] [0.002] PE -0.016*** 0.013*** -0.016*** 0.013*** [0.002] [0.002] [0.002] [0.002]
43
NS -0.020*** -0.005*** -0.020*** -0.005*** [0.001] [0.001] [0.001] [0.001] NB -0.016*** -0.022*** -0.016*** -0.022*** [0.001] [0.001] [0.001] [0.001] PQ -0.048*** -0.057*** -0.048*** -0.057*** [0.000] [0.000] [0.000] [0.000] MB -0.022*** -0.020*** -0.022*** -0.020*** [0.001] [0.001] [0.001] [0.001] SK 0.010*** 0.009*** 0.010*** 0.009*** [0.001] [0.001] [0.001] [0.001] AB 0.018*** 0.011*** 0.018*** 0.011*** [0.000] [0.001] [0.000] [0.001] BC 0.007*** 0.001*** 0.007*** 0.001*** [0.000] [0.000] [0.000] [0.000] language l English in Qc. 0.053*** 0.053*** 0.052*** 0.052*** [0.001] [0.001] [0.001] [0.001] French out Qc. -0.036*** -0.048*** -0.036*** -0.048*** [0.001] [0.002] [0.001] [0.002] regional UR -0.002*** -0.002*** -0.002*** -0.002*** [0.000] [0.000] [0.000] [0.000] union status union 0.131*** 0.137*** 0.131*** 0.137*** [0.000] [0.000] [0.000] [0.000] year 1998 -0.001 -0.010*** -0.001 -0.010*** [0.001] [0.001] [0.001] [0.001] 1999 0.005*** -0.001 0.005*** -0.002** [0.001] [0.001] [0.001] [0.001] 2000 0.014*** 0.010*** 0.013*** 0.009*** [0.001] [0.001] [0.001] [0.001] 2001 0.020*** 0.017*** 0.019*** 0.016*** [0.001] [0.001] [0.001] [0.001] 2002 0.024*** 0.025*** 0.024*** 0.024*** [0.001] [0.001] [0.001] [0.001] 2003 0.029*** 0.030*** 0.028*** 0.029*** [0.001] [0.001] [0.001] [0.001] 2004 0.034*** 0.035*** 0.034*** 0.034*** [0.001] [0.001] [0.001] [0.001] 2005 0.040*** 0.040*** 0.040*** 0.039*** [0.001] [0.001] [0.001] [0.001] event of layoff -0.132*** -0.192*** -0.382*** -0.491*** [0.001] [0.001] [0.006] [0.006] years since laoff T2 -0.040*** -0.040*** -0.042*** -0.043*** [0.000] [0.000] [0.000] [0.000]
44
T3 -0.079*** -0.081*** -0.082*** -0.084*** [0.000] [0.000] [0.000] [0.000] T4 -0.119*** -0.122*** -0.122*** -0.125*** [0.000] [0.000] [0.000] [0.000] T5 -0.159*** -0.164*** -0.161*** -0.168*** [0.001] [0.001] [0.001] [0.001] dT2 (interacted) 0.106*** 0.159*** [0.004] [0.004] dT3 (interacted) 0.135*** 0.186*** [0.004] [0.004] dT4 (interacted) 0.146*** 0.193*** [0.004] [0.004] dT5 (interacted) 0.154*** 0.199*** [0.005] [0.005] prior earnings earnings 25-40 k -0.017*** -0.047*** -0.017*** -0.047*** [0.001] [0.001] [0.001] [0.001] earnings 40-55 k -0.061*** -0.091*** -0.061*** -0.090*** [0.001] [0.001] [0.001] [0.001] earnings 55-70 k -0.097*** -0.124*** -0.097*** -0.124*** [0.001] [0.001] [0.001] [0.001] earnings 71-85 k -0.106*** -0.150*** -0.105*** -0.150*** [0.001] [0.001] [0.001] [0.001] earnings 85-100 k -0.077*** -0.093*** -0.077*** -0.093*** [0.001] [0.001] [0.001] [0.001] earnings 100+ k -0.040*** -0.052*** -0.040*** -0.051*** [0.001] [0.001] [0.001] [0.001] Constant 0.732*** 0.745*** 0.738*** 0.752*** [0.002] [0.002] [0.002] [0.002] Observations 5983380 4724375 5983380 4724375 R-squared 0.171 0.175 0.172 0.178
45
Table 5
summary of interacted effects - treatment versus control group logit equation (point estimates) linear probability equation Male Female Male Female Male Female Male Female layoff dummy -2.033 -2.649 -0.382 -0.491 age groups no layoff laid off no layoff laid off age4549 2.902 2.995 0.51 0.258 age4549 0.31 0.323 0.128 0.087age5054 1.546 1.632 -0.313 -0.812 age5054 0.223 0.232 0.005 -0.096age5559 0.439 0.573 -1.309 -1.956 age5559 0.055 0.077 -0.219 -0.357age6064 (omitted) -2.033 -2.649 age6064 -0.382 -0.491 interactions dage4549 -0.359 -0.088 dage4549 0.2 0.255 dage5054 0.174 0.205 dage5054 0.164 0.163 dage5559 0.285 0.12 dage5559 0.108 0.057 years since layoff no layoff laid off no layoff laid off t1 (omitted) -2.033 -2.649 -0.382 -0.491t2 -0.54 -0.605 -1.661 -1.991 -0.042 -0.043 -0.318 -0.375t3 -0.955 -1.05 -1.738 -2.113 -0.082 -0.084 -0.329 -0.389t4 -1.317 -1.44 -1.914 -2.326 -0.122 -0.125 -0.358 -0.423t5 -1.647 -1.796 -2.093 -2.538 -0.161 -0.168 -0.389 -0.46 interactions dt2 0.912 1.263 0.106 0.159 dt3 1.25 1.586 0.135 0.186 dt4 1.436 1.763 0.146 0.193 dt5 1.587 1.907 0.154 0.199
46
Appendix Table 1
Sample Exclusions for Construction of Cohorts of Workers at Risk of Layoff
1996 1997 1998 1999 2000 2001 2002 Full LAD 4186120 4255015 4319175 4412095 4483410 4594920 4628290 Filed return 2615045 2664310 2681825 2708715 2769995 2774175 2798765 % of original sample (62.47) (62.62) (62.09) (61.39) (61.78) (60.37) (60.47) Age range 1038425 1082915 1121050 1165545 1221775 1259520 1306645 % of original sample (24.81) (25.45) (25.96) (26.42) (27.25) (27.41) (28.23) Still Living 1033935 1078590 1116765 1161135 1217245 1254930 1301955 % of original sample (24.70) (25.35) (25.86) (26.32) (27.15) (27.31) (28.13) Residing in Canada 1030290 1074800 1112890 1157000 1212680 1250200 1296740 % of original sample (24.61) (25.26) (25.77) (26.22) (27.05) (27.21) (28.02) Exclude Self-Employed 870320 904555 932900 967160 1011575 1042635 1082960 % of original sample (20.79) (21.26) (21.60) (21.92) (22.56) (22.69) (23.40) Earnings>15k 469630 483760 500120 517020 540300 571045 603245 % of original sample (11.22) (11.37) (11.58) (11.72) (12.05) (12.43) (13.03) Exclude receipt of Transfer Income 318485 332280 346465 362900 383465 409510 432170 % of original sample (7.61) (7.81) (8.02) (8.23) (8.55) (8.91) (9.34) Exclude Students 316345 330250 344445 360865 381360 407380 430195 % of original sample (7.56) (7.76) (7.97) (8.18) (8.51) (8.87) (9.29) LAD-EI Check 314685 328415 342525 359275 379810 405165 427970 For regular EI (7.52) (7.72) (7.93) (8.14) (8.47) (8.82) (9.25)
47