Analytical Studies Branch Research Paper Series
Do Layoffs Increase Transitions to Postsecondary Education Among Adults?

by Wen Ci, Marc Frenette, and René Morissette
Social Analysis and Modelling Division
Statistics Canada

Release date: July 19, 2016

Abstract

Faced with job loss, displaced workers may choose to return to school to help them reintegrate into the labour force. Job losses in a given local labour market may also induce workers who have not yet been laid off to pre-emptively enrol in postsecondary (PS) institutions, as a precautionary measure. Combining microdata and grouped data, this study examines these two dimensions of the relationship between layoffs and PS enrolment over the 2001-to-2011 period.

Using individual-level longitudinal microdata and controlling for the unobserved heterogeneity of workers in a flexible way, the study finds that laid-off male and female workers are 2 to 4 percentage points more likely than other men and women to transition to PS education in the year of the layoff or the following year (from a baseline rate of about 3%). For both sexes, full-time PS enrolment accounts for most of the increase in enrolment. Statistically significant correlations between layoffs and full-time PS attendance are detected between two years before job loss and two years after job loss.

The study also takes advantage of the fact that the 2008-2009 recession increased layoff rates in a differentiated way across Canada and, thus, generated exogenous variation in layoff rates at the regional level. Using grouped data models that allow economic regions to display distinct trends in their rates of transition to PS institutions, the study finds that for every additional 100 adult men laid off in an economic region in a given year, there is an additional 2 to 6 men who enrol in PS education institutions on a full-time basis in the following year. The study also detects a positive relationship between regional layoff rates and regional PS transitions for unmarried women. In line with the notion that some non-laid-off workers may pre-emptively enrol in PS institutions, the study finds evidence that movements in regional layoff rates are positively correlated with short-term transitions to PS institutions for adult male workers aged 35 to 44 who have not been laid off yet.

Executive summary

Every year, thousands of workers lose their job in many industrialized countries (Organisation for Economic Co-operation and Development [OECD] 2013). These displaced workers may adjust to job loss by searching for a new job, migrating, temporarily exiting the labour force or retiring. They may also upgrade their skills through government-sponsored training or by enrolling in postsecondary (PS) institutions.

While it is well documented that many displaced workers experience substantial and persistent earnings losses, the extent to which they enrol in PS institutions after job loss remains―to a large extent―unknown. One reason is that while layoffs involve a large number of workers, they remain relatively rare events. For this reason, household surveys usually do not have the sample size required to support credible analyses of the link between job displacement and the enrolment of adults in PS institutions. To analyze this link, large administrative datasets with information on layoffs and PS attendance are required.

This study takes advantage of such a dataset and assesses the relationship between job displacement and the enrolment of adults in PS institutions. Using a unique administrative dataset that links firm-level identifiers to 100% of records from the T1 Income Tax Returns, T4 Statements of Remuneration Paid, and Records of Employment (ROEs) of Canadians, the study estimates the degree to which layoffs are associated with increased transitions to PS education among adult workers.

The study considers the possibility that, at the regional level, increased layoff rates may increase adult education through three distinct channels. First, job losses might induce some laid-off workers to return to school. Second, in distressed firms, workers who have not yet been laid off might anticipate subsequent job losses as they gather information about impending layoffs. Third, workers whose employers sell intermediate goods to distressed firms might also be worried about their job and enrol in PS institutions as a precautionary measure.

Using individual-level longitudinal microdata, the study finds that laid-off adult male and female workers are 2 to 4 percentage points more likely than other adult male and female workers to transition to PS education in the year of the layoff or the following year (from a baseline rate of about 3%). For both sexes, full-time PS enrolment accounts for most of the increase in enrolment. Statistically significant correlations between layoffs and full-time PS attendance are detected between two years before job loss and two years after job loss. This finding suggests that some laid-off workers start enrolling in PS institutions as soon as they gather information about impending layoffs and that, in some cases, their enrolment lasts more than one year.

Using grouped data, the study finds that for every additional 100 adult men laid off in an economic region in a given year, there is an additional 2 to 6 men who enrol in PS education institutions on a full-time basis in the following year. A positive relationship between regional layoff rates and regional PS transitions is also detected for unmarried women. In line with the notion that some non-laid-off workers may pre-emptively enrol in PS institutions, the study finds evidence that movements in regional layoff rates are positively correlated with short-term transitions to PS institutions for adult male workers aged 35 to 44 who have not been laid off yet.

1. Introduction

Every year, thousands of workers lose their job in many industrialized countries (Organisation for Economic Co-operation and Development [OECD] 2013). These displaced workers may adjust to job loss by searching for a new job, migrating, temporarily exiting the labour force or retiring. They may also upgrade their skills through government-sponsored training or by enrolling in postsecondary (PS) institutions.

While it is well documented that many displaced workers experience substantial and persistent earnings losses,Note 1 the extent to which they enrol in PS institutions after job loss remains―to a large extent―unknown. One reason is that while layoffs involve a large number of workers, they remain relatively rare events. For this reason, household surveys usually do not have the sample size required to support credible analyses of the link between job displacement and the enrolment of adults in PS institutions. To analyze this link, large administrative datasets with information on layoffs and PS attendance are required.

This study takes advantage of such a dataset and assesses the relationship between job displacement and the enrolment of adults in PS institutions. Using a unique administrative dataset that links firm-level identifiers to 100% of records from the T1 Income Tax Returns, T4 Statements of Remuneration Paid, and Records of Employment (ROEs) of Canadians, the study estimates the degree to which layoffs increase transitions to PS education among adult workers.

The study contributes to the literature on job displacement and adult education in two ways.

First, using individual-level longitudinal microdata, the study provides recent evidence on a potentially important margin of adjustment to job loss: enrolment in PS institutions. In doing so, the study helps shed light in understanding the determinants of adult education, about which relatively little is currently known.Note 2 The study uses regression models that control for the unobserved heterogeneity of workers―and, thus, potential selectivity―in a flexible way and that allow laid-off workers to adjust their behaviour prior to job loss, as well as after job loss.

Second, the study takes advantage of exogenous spatial and temporal variation in layoff rates induced by the 2008-2009 recession to identify layoff effects on adult education. The study considers the possibility that, at the regional level, increased layoff rates may increase adult education through three distinct channels. First, job losses might induce some laid-off workers to return to school. This is the focus of the aforementioned regression models based on individual-level longitudinal microdata. Second, in distressed firms, workers who have not yet been laid off might anticipate subsequent job losses as they gather information about impending layoffs. Third, workers whose employers sell intermediate goods to distressed firms might also be worried about their jobs and enrol in PS institutions as a precautionary measure.

These mechanisms have testable implications in grouped data. Taken together, the three channels suggest that regions that experience high layoff rates in some years―relative to their own average layoff rate―should, all else being equal, experience relatively high rates of PS attendance during those years. The last two channels suggest that movements in regional layoff rates should be positively correlated with PS attendance for those groups of workers who have not been laid off yet.

The study tests these two hypotheses by grouping microdata at the economic-region level over the 2001-to-2011 period. Because this period includes the 2008-2009 recession, which increased layoff rates in a differentiated way across Canada,Note 3 the study takes advantage of exogenous variation in layoff rates at the regional level when using grouping estimators. Using flexible models that allow economic regions to display distinct trends in their rates of transition to PS institutions, the study quantifies the degree to which economic regions that experienced increases in layoff rates from 2001 to 2011 displayed increases in PS transitions. It also assesses whether groups of workers who had not yet been laid off increased their transitions to PS institutions in the short term as regional layoff rates increased. To the knowledge of the authors, no study has performed this task to date.

Using individual-level longitudinal microdata, the study finds that laid-off adult male and female workers are 2 to 4 percentage points more likely than other adult male and female workers to transition to PS education in the year of the layoff or the following year (from a baseline rate of about 3%). For both sexes, full-time PS enrolment accounts for most of the increase in enrolment. Statistically significant correlations between layoffs and full-time PS attendance are detected between two years before job loss and two years after job loss. This finding suggests that some laid-off workers start enrolling in PS institutions as soon as they gather information about impending layoffs and that, in some cases, their enrolment lasts more than one year.

Using grouped data, the study finds that for every additional 100 adult men laid off in an economic region in a given year, there is an additional 2 to 6 men who enrol in PS education institutions on a full-time basis in the following year. A positive relationship between regional layoff rates and regional PS transitions is also detected for unmarried women. In line with the notion that some non-laid-off workers may pre-emptively enrol in PS institutions, the study finds evidence that movements in regional layoff rates are positively correlated with short-term transitions to PS institutions for adult male workers aged 35 to 44 who have not been laid off yet.

Taken together, these results provide strong evidence that job loss is one determinant of adult education (i.e., it may cause an increase in transitions to PS education among adults).

The paper is organized as follows. Section 2 reviews previous work investigating the link between job loss and adult education. Data and methods are presented in Section 3, and results are shown in Section 4. Section 5 concludes the paper.

2. Background

In response to job loss, workers can choose to upgrade their skills through government-sponsored training or by enrolling in PS institutions.Note 4 To date, the degree to which displaced workers increase their PS enrolment after job loss has been the subject of few empirical analyses.

To the knowledge of the authors, the only study that examines this issue is by Frenette, Upward and Wright (2011). Using Canadian administrative data, they find that job displacement from firm closures and mass layoffs is associated with a 1-percentage-point increase in PS attendance (on a base of 10%). Since job losses from firm closures or mass layoffs account for less than one-third of all layoffs in Canada (Morissette, Zhang and Frenette 2007), their analysis is―contrary to the present study―restricted to a subset of job losses.

A priori, it is unclear whether many laid-off workers will go back to school after job loss. In many industrialized countries, half of displaced workers find a job within one year of being displaced (OECD 2013). For this group, the incentives for making a transition into PS education are fairly low. In contrast, displaced workers who become unemployed after displacement face, at least in the short term, a relatively low opportunity cost―in terms of foregone earnings―of going back to school. In addition, some of them might see their earnings rise as a result of additional education.Note 5 Hence, their likelihood of enrolling in PS institutions might be relatively high.

Yet, other factors might restrict the propensity of displaced workers to go back to school. Parental responsibilities and the non-monetary costs of pursuing more schooling (e.g., the energy and effort required to study and write exams) might limit the degree to which they enrol in PS institutions. In addition, if firms selectively lay off workers who have lower-than-average productivity, and if productivity and learning ability are positively correlated, laid-off workers might have a relatively low ability to learn new concepts. Hence, whether adult displaced workers have a relatively high or a relatively low likelihood of going back to school after job loss is an empirical question.

The goal of this study is to answer this question using a unique administrative dataset that links firm-level identifiers to 100% of records from the T1 Income Tax Returns, T4 Statements of Remuneration Paid, and ROEs of Canadians. Because this dataset tracks workers over an 11-year period, it enables longitudinal microdata analyses that control for the unobserved heterogeneity of individuals in a flexible way. Because this dataset covers virtually all Canadian taxfilers, it allows for the use of grouping estimators that have the power to detect even small layoff effects on PS enrolment.

3. Data and methods

A simple way to model the decision of adults to go back to a postsecondary institution is to consider the following individual-level equation:

Y iart = θ i + θ t + λ i *t+ k=a b L k it * β 1,k + β 2 *URAT E art + X it * β 3 + ε it ,t=2001,...2011, ( 1 ) MathType@MTEF@5@5@+= feaagKart1ev2aaatCvAUfeBSjuyZL2yd9gzLbvyNv2CaerbuLwBLn hiov2DGi1BTfMBaeXatLxBI9gBaerbd9wDYLwzYbItLDharqqtubsr 4rNCHbGeaGqiVu0Je9sqqrpepC0xbbL8F4rqqrFfpeea0xe9Lq=Jc9 vqaqpepm0xbba9pwe9Q8fs0=yqaqpepae9pg0FirpepeKkFr0xfr=x fr=xb9adbaqaaeGaciGaaiaabeqaamaabaabaaGcbaacbiGaa8xwam aaBaaaleaacaWFPbGaa8xyaiaa=jhacaWF0baabeaakiaa=bcacaWF 9aGaa8hiaiabeI7aXnaaBaaaleaacaWGPbaabeaakiaa=bcacaWFRa Gaa8hiaiaa=H7adaWgaaWcbaGaa8hDaiaa=bcaaeqaaOGaa83kaiab eU7aSnaaBaaaleaacaWGPbaabeaakiaacQcacaWG0bGaey4kaSYaaa bCaeaacaWGmbWaaWbaaSqabeaacaWGRbaaaOWaaSbaaSqaaiaadMga caWG0baabeaaaeaacaWGRbGaeyypa0Jaamyyaaqaaiaadkgaa0Gaey yeIuoakiaa=PcacaWFYoWaaSbaaSqaaiaaigdacaGGSaGaam4Aaaqa baGccaWFRaGaeqOSdi2aaSbaaSqaaiaaikdaaeqaaOGaaiOkaiaa=v facaWFsbGaa8xqaiaa=rfacaWFfbWaaSbaaSqaaiaa=fgacaWFYbGa a8hDaaqabaGccaWFRaGaa8hwamaaBaaaleaacaWFPbGaa8hDaaqaba GccaGGQaGaa8NSdmaaBaaaleaacaaIZaaabeaakiabgUcaRiaa=bca caWFGaGaeqyTdu2aaSbaaSqaaiaa=LgacaWF0baabeaakiaacYcaca WG0bGaeyypa0JaaGOmaiaaicdacaaIWaGaaGymaiaacYcacaGGUaGa aiOlaiaac6cacaaIYaGaaGimaiaaigdacaaIXaGaaiilaaaa@7C63@

where Y iart MathType@MTEF@5@5@+= feaagKart1ev2aaatCvAUfeBSjuyZL2yd9gzLbvyNv2CaerbuLwBLn hiov2DGi1BTfMBaeXatLxBI9gBaerbd9wDYLwzYbItLDharqqtubsr 4rNCHbGeaGqiVu0Je9sqqrpepC0xbbL8F4rqqrFfpeea0xe9Lq=Jc9 vqaqpepm0xbba9pwe9Q8fs0=yqaqpepae9pg0FirpepeKkFr0xfr=x fr=xb9adbaqaaeGaciGaaiaabeqaamaabaabaaGcbaacbiGaa8xwam aaBaaaleaacaWFPbGaa8xyaiaa=jhacaWF0baabeaaaaa@3ABB@ is a binary indicator that equals 1 if worker i in age group a and economic region r  attends a PS education institution in year t, and 0 otherwise. L it k MathType@MTEF@5@5@+= feaagKart1ev2aaatCvAUfeBSjuyZL2yd9gzLbvyNv2CaerbuLwBLn hiov2DGi1BTfMBaeXatLxBI9gBaerbd9wDYLwzYbItLDharqqtubsr 4rNCHbGeaGqiVu0Je9sqqrpepC0xbbL8F4rqqrFfpeea0xe9Lq=Jc9 vqaqpepm0xbba9pwe9Q8fs0=yqaqpepae9pg0FirpepeKkFr0xfr=x fr=xb9adbaqaaeGaciGaaiaabeqaamaabaabaaGcbaGaamitamaaDa aaleaacaWGPbGaamiDaaqaaiaadUgaaaaaaa@39CB@ is a vector of binary indicators that equals 1 if worker i is laid off k years prior to year t, and 0 otherwise. The parameters a and b are set to −2 and 5, respectively, thereby allowing job loss to affect transitions to adult postsecondary education between up to two years before layoffs and up to five years after layoffs. URAT E art MathType@MTEF@5@5@+= feaagKart1ev2aaatCvAUfeBSjuyZL2yd9gzLbvyNv2CaerbuLwBLn hiov2DGi1BTfMBaeXatLxBI9gBaerbd9wDYLwzYbItLDharqqtubsr 4rNCHbGeaGqiVu0Je9sqqrpepC0xbbL8F4rqqrFfpeea0xe9Lq=Jc9 vqaqpepm0xbba9pwe9Q8fs0=yqaqpepae9pg0FirpepeKkFr0xfr=x fr=xb9adbaqaaeGaciGaaiaabeqaamaabaabaaGcbaacbiGaa8xvai aa=jfacaWFbbGaa8hvaiaa=veadaWgaaWcbaGaa8xyaiaa=jhacaWF 0baabeaaaaa@3CFD@ is a gender-specific unemployment rate for individuals in age group a in economic region r during year t, X it MathType@MTEF@5@5@+= feaagKart1ev2aaatCvAUfeBSjuyZL2yd9gzLbvyNv2CaerbuLwBLn hiov2DGi1BTfMBaeXatLxBI9gBaerbd9wDYLwzYbItLDharqqtubsr 4rNCHbGeaGqiVu0Je9sqqrpepC0xbbL8F4rqqrFfpeea0xe9Lq=Jc9 vqaqpepm0xbba9pwe9Q8fs0=yqaqpepae9pg0FirpepeKkFr0xfr=x fr=xb9adbaqaaeGaciGaaiaabeqaamaabaabaaGcbaGaamiwamaaBa aaleaacaWGPbGaamiDaaqabaacbiGccaWFGaaaaa@399C@  is a vector of individual-level characteristics observed in year t, θ i MathType@MTEF@5@5@+= feaagKart1ev2aaatCvAUfeBSjuyZL2yd9gzLbvyNv2CaerbuLwBLn hiov2DGi1BTfMBaeXatLxBI9gBaerbd9wDYLwzYbItLDharqqtubsr 4rNCHbGeaGqiVu0Je9sqqrpepC0xbbL8F4rqqrFfpeea0xe9Lq=Jc9 vqaqpepm0xbba9pwe9Q8fs0=yqaqpepae9pg0FirpepeKkFr0xfr=x fr=xb9adbaqaaeGaciGaaiaabeqaamaabaabaaGcbaGaeqiUde3aaS baaSqaaiaadMgaaeqaaGqacOGaa8hiaaaa@397C@ is a vector of individual-level fixed effects, θ t MathType@MTEF@5@5@+= feaagKart1ev2aaatCvAUfeBSjuyZL2yd9gzLbvyNv2CaerbuLwBLn hiov2DGi1BTfMBaeXatLxBI9gBaerbd9wDYLwzYbItLDharqqtubsr 4rNCHbGeaGqiVu0Je9sqqrpepC0xbbL8F4rqqrFfpeea0xe9Lq=Jc9 vqaqpepm0xbba9pwe9Q8fs0=yqaqpepae9pg0FirpepeKkFr0xfr=x fr=xb9adbaqaaeGaciGaaiaabeqaamaabaabaaGcbaacbiGaa8hUdm aaBaaaleaacaWF0bGaa8hiaaqabaaaaa@38FF@  is a vector of year effects, and ε it MathType@MTEF@5@5@+= feaagKart1ev2aaatCvAUfeBSjuyZL2yd9gzLbvyNv2CaerbuLwBLn hiov2DGi1BTfMBaeXatLxBI9gBaerbd9wDYLwzYbItLDharqqtubsr 4rNCHbGeaGqiVu0Je9sqqrpepC0xbbL8F4rqqrFfpeea0xe9Lq=Jc9 vqaqpepm0xbba9pwe9Q8fs0=yqaqpepae9pg0FirpepeKkFr0xfr=x fr=xb9adbaqaaeGaciGaaiaabeqaamaabaabaaGcbaGaeqyTdu2aaS baaSqaaGqaciaa=LgacaWF0baabeaaaaa@39B3@ is a random error term.Note 6

Equation (1) controls for the unobserved heterogeneity of workers in a flexible way. The vector θ i MathType@MTEF@5@5@+= feaagKart1ev2aaatCvAUfeBSjuyZL2yd9gzLbvyNv2CaerbuLwBLn hiov2DGi1BTfMBaeXatLxBI9gBaerbd9wDYLwzYbItLDharqqtubsr 4rNCHbGeaGqiVu0Je9sqqrpepC0xbbL8F4rqqrFfpeea0xe9Lq=Jc9 vqaqpepm0xbba9pwe9Q8fs0=yqaqpepae9pg0FirpepeKkFr0xfr=x fr=xb9adbaqaaeGaciGaaiaabeqaamaabaabaaGcbaGaeqiUde3aaS baaSqaaiaadMgaaeqaaGqacOGaa8hiaaaa@397C@  accounts for time-invariant unobserved abilities, as well as time-invariant factors such as the first degree, diploma and field of study of individuals.Note 7 The term λ i *t MathType@MTEF@5@5@+= feaagKart1ev2aaatCvAUfeBSjuyZL2yd9gzLbvyNv2CaerbuLwBLn hiov2DGi1BTfMBaeXatLxBI9gBaerbd9wDYLwzYbItLDharqqtubsr 4rNCHbGeaGqiVu0Je9sqqrpepC0xbbL8F4rqqrFfpeea0xe9Lq=Jc9 vqaqpepm0xbba9pwe9Q8fs0=yqaqpepae9pg0FirpepeKkFr0xfr=x fr=xb9adbaqaaeGaciGaaiaabeqaamaabaabaaGcbaGaeq4UdW2aaS baaSqaaiaadMgaaeqaaOGaaiOkaiaadshaaaa@3A75@  allows individuals to display person-specific trends in their propensity to attend PS institutions. As a result, the likelihood of attending PS institutions may fall or rise over time at a different pace for laid-off workers and other workers. If it is assumed that λ i =0 MathType@MTEF@5@5@+= feaagKart1ev2aqatCvAUfeBSjuyZL2yd9gzLbvyNv2CaerbuLwBLn hiov2DGi1BTfMBaeXatLxBI9gBaerbd9wDYLwzYbItLDharqqtubsr 4rNCHbGeaGqiVu0Je9sqqrpepC0xbbL8F4rqqrFfpeea0xe9Lq=Jc9 vqaqpepm0xbba9pwe9Q8fs0=yqaqpepae9pg0FirpepeKkFr0xfr=x fr=xb9adbaqaaeGaciGaaiaabeqaamaabaabaaGcbaGaeq4UdW2aaS baaSqaaiaadMgaaeqaaOGaeyypa0JaaGimaaaa@3A8F@ , then Equation (1) is a fixed-effects model.Note 8

Equation (1) uses individual-level longitudinal microdata and tracks individuals over the 2001-to-2011 period to answer the following question: among individuals who face similar labour-market conditions (as measured by URAT E art MathType@MTEF@5@5@+= feaagKart1ev2aaatCvAUfeBSjuyZL2yd9gzLbvyNv2CaerbuLwBLn hiov2DGi1BTfMBaeXatLxBI9gBaerbd9wDYLwzYbItLDharqqtubsr 4rNCHbGeaGqiVu0Je9sqqrpepC0xbbL8F4rqqrFfpeea0xe9Lq=Jc9 vqaqpepm0xbba9pwe9Q8fs0=yqaqpepae9pg0FirpepeKkFr0xfr=x fr=xb9adbaqaaeGaciGaaiaabeqaamaabaabaaGcbaacbiGaa8xvai aa=jfacaWFbbGaa8hvaiaa=veadaWgaaWcbaGaa8xyaiaa=jhacaWF 0baabeaaaaa@3CFD@ ), to what extent are those who experience job loss more likely than others to make a transition into PS education?

An alternative question considers the link between regional movements in layoff rates and regional changes in rates of transition to PS education. It acknowledges the possibility that, at the regional level, increased layoff rates may lead to increases in adult education through three distinct channels. First, job losses might induce some laid-off workers to return to a postsecondary institution. This is the focus of Equation (1). Second, in distressed firms, workers who have not yet been laid off might anticipate subsequent job losses as they gather information about impending layoffs. Third, workers whose employers sell intermediate goods to distressed firms might also be worried about their job and enrol in PS institutions as a precautionary measure.Note 9

This alternative question can be answered by estimating the following model using data grouped at the economic-region level:

Y rt = λ r + λ t + γ r *t + γ 1 * L rt + γ 2 *U ' rt + X rt * γ 3 + ε rt ,t=2001,...2011;r=1,...66, ( 2 ) MathType@MTEF@5@5@+= feaagKart1ev2aaatCvAUfeBSjuyZL2yd9gzLbvyNv2CaerbuLwBLn hiov2DGi1BTfMBaeXatLxBI9gBaerbd9wDYLwzYbItLDharqqtubsr 4rNCHbGeaGqiVu0Je9sqqrpepC0xbbL8F4rqqrFfpeea0xe9Lq=Jc9 vqaqpepm0xbba9pwe9Q8fs0=yqaqpepae9pg0FirpepeKkFr0xfr=x fr=xb9adbaqaaeGaciGaaiaabeqaamaabaabaaGcbaGaamywamaaBa aaleaaieGacaWFYbGaa8hDaaqabaGccaWFGaGaa8xpaiaa=bcacqaH 7oaBdaWgaaWcbaGaamOCaaqabaGccaWFGaGaa83kaiaa=bcacqaH7o aBdaWgaaWcbaGaa8hDaiaa=bcaaeqaaOGaa83kaiaa=bcacqaHZoWz daWgaaWcbaGaamOCaaqabaGccaWFQaGaa8hDaiaa=bcacaWFRaGaa8 hiaiabeo7aNnaaBaaaleaaieaacaGFXaaabeaakiaa=PcacaWFmbWa aSbaaSqaaiaa=jhacaWF0baabeaakiaa=bcacaWFRaGaa8hiaiabeo 7aNnaaBaaaleaacaaIYaaabeaakiaa=PcacaWFvbGaa83jamaaBaaa leaacaWFYbGaa8hDaaqabaGccaWFRaGaa8hwamaaBaaaleaacaWFYb Gaa8hDaaqabaGccaWFQaGaeq4SdC2aaSbaaSqaaiaa+ndaaeqaaOGa a83kaiaa=bcacqaH1oqzdaWgaaWcbaGaa8NCaiaa=rhaaeqaaOGaai ilaiaadshacqGH9aqpcaaIYaGaaGimaiaaicdacaaIXaGaaiilaiaa c6cacaGGUaGaaiOlaiaaikdacaaIWaGaaGymaiaaigdacaGG7aGaam OCaiabg2da9iaaigdacaGGSaGaaiOlaiaac6cacaGGUaGaaGOnaiaa iAdacaaMi8UaaGjcVlaacYcaaaa@7CC1@

where Y rt MathType@MTEF@5@5@+= feaagKart1ev2aaatCvAUfeBSjuyZL2yd9gzLbvyNv2CaerbuLwBLn hiov2DGi1BTfMBaeXatLxBI9gBaerbd9wDYLwzYbItLDharqqtubsr 4rNCHbGeaGqiVu0Je9sqqrpepC0xbbL8F4rqqrFfpeea0xe9Lq=Jc9 vqaqpepm0xbba9pwe9Q8fs0=yqaqpepae9pg0FirpepeKkFr0xfr=x fr=xb9adbaqaaeGaciGaaiaabeqaamaabaabaaGcbaGaamywamaaBa aaleaaieGacaWFYbGaa8hDaaqabaaaaa@38F3@  measures the percentage of adult employees in region r who make a transition into PS education from year t  to year t +1. The coefficient γ 1 MathType@MTEF@5@5@+= feaagKart1ev2aaatCvAUfeBSjuyZL2yd9gzLbvyNv2CaerbuLwBLn hiov2DGi1BTfMBaeXatLxBI9gBaerbd9wDYLwzYbItLDharqqtubsr 4rNCHbGeaGqiVu0Je9sqqrpepC0xbbL8F4rqqrFfpeea0xe9Lq=Jc9 vqaqpepm0xbba9pwe9Q8fs0=yqaqpepae9pg0FirpepeKkFr0xfr=x fr=xb9adbaqaaeGaciGaaiaabeqaamaabaabaaGcbaGaeq4SdC2aaS baaSqaaGqaaiaa=fdaaeqaaaaa@3884@  captures the impact that the layoff rate L rt MathType@MTEF@5@5@+= feaagKart1ev2aaatCvAUfeBSjuyZL2yd9gzLbvyNv2CaerbuLwBLn hiov2DGi1BTfMBaeXatLxBI9gBaerbd9wDYLwzYbItLDharqqtubsr 4rNCHbGeaGqiVu0Je9sqqrpepC0xbbL8F4rqqrFfpeea0xe9Lq=Jc9 vqaqpepm0xbba9pwe9Q8fs0=yqaqpepae9pg0FirpepeKkFr0xfr=x fr=xb9adbaqaaeGaciGaaiaabeqaamaabaabaaGcbaacbiGaa8htam aaBaaaleaacaWFYbGaa8hDaaqabaaaaa@38E2@  may have on these transitions through the three aforementioned channels. Y rt MathType@MTEF@5@5@+= feaagKart1ev2aaatCvAUfeBSjuyZL2yd9gzLbvyNv2CaerbuLwBLn hiov2DGi1BTfMBaeXatLxBI9gBaerbd9wDYLwzYbItLDharqqtubsr 4rNCHbGeaGqiVu0Je9sqqrpepC0xbbL8F4rqqrFfpeea0xe9Lq=Jc9 vqaqpepm0xbba9pwe9Q8fs0=yqaqpepae9pg0FirpepeKkFr0xfr=x fr=xb9adbaqaaeGaciGaaiaabeqaamaabaabaaGcbaGaamywamaaBa aaleaaieGacaWFYbGaa8hDaaqabaaaaa@38F3@  can also be measured for the subset of adult employees who have not been laid off in year t. Doing so allows an implication of the last two channels to be tested (i.e., whether movements in regional layoff rates are positively correlated with short-term transitions to PS institutions for adult male workers who have not been laid off yet).

Regional fixed effects ( λ r MathType@MTEF@5@5@+= feaagKart1ev2aaatCvAUfeBSjuyZL2yd9gzLbvyNv2CaerbuLwBLn hiov2DGi1BTfMBaeXatLxBI9gBaerbd9wDYLwzYbItLDharqqtubsr 4rNCHbGeaGqiVu0Je9sqqrpepC0xbbL8F4rqqrFfpeea0xe9Lq=Jc9 vqaqpepm0xbba9pwe9Q8fs0=yqaqpepae9pg0FirpepeKkFr0xfr=x fr=xb9adbaqaaeGaciGaaiaabeqaamaabaabaaGcbaGaeq4UdW2aaS baaSqaaiaadkhaaeqaaaaa@38CD@ ), nationwide year effects ( λ t MathType@MTEF@5@5@+= feaagKart1ev2aaatCvAUfeBSjuyZL2yd9gzLbvyNv2CaerbuLwBLn hiov2DGi1BTfMBaeXatLxBI9gBaerbd9wDYLwzYbItLDharqqtubsr 4rNCHbGeaGqiVu0Je9sqqrpepC0xbbL8F4rqqrFfpeea0xe9Lq=Jc9 vqaqpepm0xbba9pwe9Q8fs0=yqaqpepae9pg0FirpepeKkFr0xfr=x fr=xb9adbaqaaeGaciGaaiaabeqaamaabaabaaGcbaGaeq4UdW2aaS baaSqaaiaadshaaeqaaaaa@38CF@ ), and region-specific linear time trends ( γ r t MathType@MTEF@5@5@+= feaagKart1ev2aqatCvAUfeBSjuyZL2yd9gzLbvyNv2CaerbuLwBLn hiov2DGi1BTfMBaeXatLxBI9gBaerbd9wDYLwzYbItLDharqqtubsr 4rNCHbGeaGqiVu0Je9sqqrpepC0xbbL8F4rqqrFfpeea0xe9Lq=Jc9 vqaqpepm0xbba9pwe9Q8fs0=yqaqpepae9pg0FirpepeKkFr0xfr=x fr=xb9adbaqaaeGaciGaaiaabeqaamaabaabaaGcbaGaeq4SdC2aaS baaSqaaiaadkhaaeqaaOGaey4fIOIaamiDaaaa@3AB3@ ) are included in Equation (2).Note 10 X rt MathType@MTEF@5@5@+= feaagKart1ev2aaatCvAUfeBSjuyZL2yd9gzLbvyNv2CaerbuLwBLn hiov2DGi1BTfMBaeXatLxBI9gBaerbd9wDYLwzYbItLDharqqtubsr 4rNCHbGeaGqiVu0Je9sqqrpepC0xbbL8F4rqqrFfpeea0xe9Lq=Jc9 vqaqpepm0xbba9pwe9Q8fs0=yqaqpepae9pg0FirpepeKkFr0xfr=x fr=xb9adbaqaaeGaciGaaiaabeqaamaabaabaaGcbaGaamiwamaaBa aaleaaieGacaWFYbGaa8hDaaqabaaaaa@38F2@  is a vector of region-specific characteristics observed in year t, and ε rt MathType@MTEF@5@5@+= feaagKart1ev2aaatCvAUfeBSjuyZL2yd9gzLbvyNv2CaerbuLwBLn hiov2DGi1BTfMBaeXatLxBI9gBaerbd9wDYLwzYbItLDharqqtubsr 4rNCHbGeaGqiVu0Je9sqqrpepC0xbbL8F4rqqrFfpeea0xe9Lq=Jc9 vqaqpepm0xbba9pwe9Q8fs0=yqaqpepae9pg0FirpepeKkFr0xfr=x fr=xb9adbaqaaeGaciGaaiaabeqaamaabaabaaGcbaGaeqyTdu2aaS baaSqaaGqaciaa=jhacaWF0baabeaaaaa@39BC@  is an error term. The intuition behind Equation (2) is simple: if layoffs induce some adults to enter PS education through the three aforementioned channels, then, in those years where economic regions experience high layoff rates relative to their own average layoff rate, economic regions should also display relatively high rates of transition to PS education.

The variable U ' rt MathType@MTEF@5@5@+= feaagKart1ev2aaatCvAUfeBSjuyZL2yd9gzLbvyNv2CaerbuLwBLn hiov2DGi1BTfMBaeXatLxBI9gBaerbd9wDYLwzYbItLDharqqtubsr 4rNCHbGeaGqiVu0Je9sqqrpepC0xbbL8F4rqqrFfpeea0xe9Lq=Jc9 vqaqpepm0xbba9pwe9Q8fs0=yqaqpepae9pg0FirpepeKkFr0xfr=x fr=xb9adbaqaaeGaciGaaiaabeqaamaabaabaaGcbaacbiGaa8xvai aa=DcadaWgaaWcbaGaa8NCaiaa=rhaaeqaaaaa@3993@  measures movements in the unemployment rate of region r in year t that are orthogonal to L rt MathType@MTEF@5@5@+= feaagKart1ev2aaatCvAUfeBSjuyZL2yd9gzLbvyNv2CaerbuLwBLn hiov2DGi1BTfMBaeXatLxBI9gBaerbd9wDYLwzYbItLDharqqtubsr 4rNCHbGeaGqiVu0Je9sqqrpepC0xbbL8F4rqqrFfpeea0xe9Lq=Jc9 vqaqpepm0xbba9pwe9Q8fs0=yqaqpepae9pg0FirpepeKkFr0xfr=x fr=xb9adbaqaaeGaciGaaiaabeqaamaabaabaaGcbaacbiGaa8htam aaBaaaleaacaWFYbGaa8hDaaqabaaaaa@38E2@ . More precisely, U ' rt MathType@MTEF@5@5@+= feaagKart1ev2aaatCvAUfeBSjuyZL2yd9gzLbvyNv2CaerbuLwBLn hiov2DGi1BTfMBaeXatLxBI9gBaerbd9wDYLwzYbItLDharqqtubsr 4rNCHbGeaGqiVu0Je9sqqrpepC0xbbL8F4rqqrFfpeea0xe9Lq=Jc9 vqaqpepm0xbba9pwe9Q8fs0=yqaqpepae9pg0FirpepeKkFr0xfr=x fr=xb9adbaqaaeGaciGaaiaabeqaamaabaabaaGcbaacbiGaa8xvai aa=DcadaWgaaWcbaGaa8NCaiaa=rhaaeqaaaaa@3993@  is equal to the residuals from a regression in which the unemployment rate in region r in year t is regressed on L rt MathType@MTEF@5@5@+= feaagKart1ev2aaatCvAUfeBSjuyZL2yd9gzLbvyNv2CaerbuLwBLn hiov2DGi1BTfMBaeXatLxBI9gBaerbd9wDYLwzYbItLDharqqtubsr 4rNCHbGeaGqiVu0Je9sqqrpepC0xbbL8F4rqqrFfpeea0xe9Lq=Jc9 vqaqpepm0xbba9pwe9Q8fs0=yqaqpepae9pg0FirpepeKkFr0xfr=x fr=xb9adbaqaaeGaciGaaiaabeqaamaabaabaaGcbaacbiGaa8htam aaBaaaleaacaWFYbGaa8hDaaqabaaaaa@38E2@ .Note 11 The rationale underlying the inclusion of U ' rt MathType@MTEF@5@5@+= feaagKart1ev2aaatCvAUfeBSjuyZL2yd9gzLbvyNv2CaerbuLwBLn hiov2DGi1BTfMBaeXatLxBI9gBaerbd9wDYLwzYbItLDharqqtubsr 4rNCHbGeaGqiVu0Je9sqqrpepC0xbbL8F4rqqrFfpeea0xe9Lq=Jc9 vqaqpepm0xbba9pwe9Q8fs0=yqaqpepae9pg0FirpepeKkFr0xfr=x fr=xb9adbaqaaeGaciGaaiaabeqaamaabaabaaGcbaacbiGaa8xvai aa=DcadaWgaaWcbaGaa8NCaiaa=rhaaeqaaaaa@3993@  as a control variable in Equation (2) is the following: economic regions may differ not only in terms of cyclical unemployment (captured by L rt MathType@MTEF@5@5@+= feaagKart1ev2aaatCvAUfeBSjuyZL2yd9gzLbvyNv2CaerbuLwBLn hiov2DGi1BTfMBaeXatLxBI9gBaerbd9wDYLwzYbItLDharqqtubsr 4rNCHbGeaGqiVu0Je9sqqrpepC0xbbL8F4rqqrFfpeea0xe9Lq=Jc9 vqaqpepm0xbba9pwe9Q8fs0=yqaqpepae9pg0FirpepeKkFr0xfr=x fr=xb9adbaqaaeGaciGaaiaabeqaamaabaabaaGcbaacbiGaa8htam aaBaaaleaacaWFYbGaa8hDaaqabaaaaa@38E2@ ), but also in terms of other types of unemployment. For example, some regions might have greater mismatch unemployment than others, because of a lack of concordance between the skills required for vacant positions and the skills of unemployed individuals. Including U ' rt MathType@MTEF@5@5@+= feaagKart1ev2aaatCvAUfeBSjuyZL2yd9gzLbvyNv2CaerbuLwBLn hiov2DGi1BTfMBaeXatLxBI9gBaerbd9wDYLwzYbItLDharqqtubsr 4rNCHbGeaGqiVu0Je9sqqrpepC0xbbL8F4rqqrFfpeea0xe9Lq=Jc9 vqaqpepm0xbba9pwe9Q8fs0=yqaqpepae9pg0FirpepeKkFr0xfr=x fr=xb9adbaqaaeGaciGaaiaabeqaamaabaabaaGcbaacbiGaa8xvai aa=DcadaWgaaWcbaGaa8NCaiaa=rhaaeqaaaaa@3993@  as a control variable in Equation (2) allows the grouped data models to answer the following question: considering regions that are similar in terms of non-cyclical unemployment, to what extent do those experiencing high layoff rates (relative to their own mean) also display relatively high rates of transition to PS education?

Equation (1) is estimated using the Canadian Employer–Employee Dynamics Database (CEEDD), a linked longitudinal administrative dataset that includes data from 2001 to 2011 and that consists of the T1 Personal Master File, the T4 file, the Longitudinal Employment Analysis Program (LEAP) database and the ROE. These files are essentially combined into a 100% longitudinal worker file. This file has been used to track the long-term outcomes of laid-off individuals by Morissette, Zhang and Frenette (2007); Frenette, Upward and Wright (2011); Bonikowska and Morissette (2012); and Morissette, Qiu and Chan (2013).

The sample used for Equation (1) consists of individuals aged 35 to 44 in 2001 who filed a T1 Income Tax Return every year during the 2001-to-2011 period and who, during the 2001-to-2003 period, had positive wages and salaries, no self-employment income and did not experience a permanent layoff. Eight treatment groups are considered. They represent workers whose first permanent layoff after 2003 occurred in year t, where t=2004,2005,...2011 MathType@MTEF@5@5@+= feaagKart1ev2aqatCvAUfeBSjuyZL2yd9gzLbvyNv2CaerbuLwBLn hiov2DGi1BTfMBaeXatLxBI9gBaerbd9wDYLwzYbItLDharqqtubsr 4rNCHbGeaGqiVu0Je9sqqrpepC0xbbL8F4rqqrFfpeea0xe9Lq=Jc9 vqaqpepm0xbba9pwe9Q8fs0=yqaqpepae9pg0FirpepeKkFr0xfr=x fr=xb9adbaqaaeGaciGaaiaabeqaamaabaabaaGcbaGaamiDaiabg2 da9iaaikdacaaIWaGaaGimaiaaisdacaGGSaGaaGOmaiaaicdacaaI WaGaaGynaiaacYcacaaMi8UaaGjcVlaayIW7caaMi8UaaiOlaiaac6 cacaGGUaGaaGjcVlaayIW7caaMi8UaaGjcVlaaikdacaaIWaGaaGym aiaaigdaaaa@50BD@ .Note 12 The control group consists of individuals who were employees throughout the 2001-to-2011 period (i.e., who had positive wages and salaries and no self-employment income during that period) and who were never permanently laid off during that period.

The dependent variable Y iart MathType@MTEF@5@5@+= feaagKart1ev2aaatCvAUfeBSjuyZL2yd9gzLbvyNv2CaerbuLwBLn hiov2DGi1BTfMBaeXatLxBI9gBaerbd9wDYLwzYbItLDharqqtubsr 4rNCHbGeaGqiVu0Je9sqqrpepC0xbbL8F4rqqrFfpeea0xe9Lq=Jc9 vqaqpepm0xbba9pwe9Q8fs0=yqaqpepae9pg0FirpepeKkFr0xfr=x fr=xb9adbaqaaeGaciGaaiaabeqaamaabaabaaGcbaacbiGaa8xwam aaBaaaleaacaWFPbGaa8xyaiaa=jhacaWF0baabeaaaaa@3ABB@  is a binary variable indicating PS attendance (either full-time or part-time). It is computed by looking for the presence of federal non-refundable full-time or part-time education deduction credits on the T1 Personal Master File. Although these may be transferred to another family member for tax purposes, it is possible to identify the student to whom the claims apply for the period examined. These credits and deductions apply to any PS attendance, with only a few exceptions (e.g., university or college preparatory courses). They exclude programs that were paid for by another body (e.g., the government). In other words, only self-financed PS schooling is reported in the tax data.

The layoff indicator L it k MathType@MTEF@5@5@+= feaagKart1ev2aqatCvAUfeBSjuyZL2yd9gzLbvyNv2CaerbuLwBLn hiov2DGi1BTfMBaeXatLxBI9gBaerbd9wDYLwzYbItLDharqqtubsr 4rNCHbGeaGqiVu0Je9sqqrpepC0xbbL8F4rqqrFfpeea0xe9Lq=Jc9 vqaqpepm0xbba9pwe9Q8fs0=yqaqpepae9pg0FirpepeKkFr0xfr=x fr=xb9adbaqaaeGaciGaaiaabeqaamaabaabaaGcbaGaamitamaaDa aaleaacaWGPbGaamiDaaqaaiaadUgaaaaaaa@39CC@  is built using information on permanent layoffs. Permanent layoffs can be identified through the ROE, which contains the reason for job separation. A permanent layoff occurs if the employee separates from the firm because of a shortage of work and does not return to that firm in the year of the layoff or the year after. To determine this, the longitudinal enterprise identifier available in the LEAP dataset is used.

The other regressors in Equation (1) include year indicators, the age of workers and their age squared, and a gender-specific economic-region unemployment rate of age group a in year t, obtained from the Labour Force Survey (LFS). Equation (1) is estimated separately for men and women. Since Equation (1) is a worker-level model that uses information from economic regions (their unemployment rate), standard errors are clustered at the economic-region level during estimation.

The sample used for Equation (2) consists of employees who are not students in year t. The main sample focuses on those aged 35 to 44 in year t, but, for robustness, results are also shown for those aged 45 to 54. Microdata are aggregated at the economic-region level. In total, there are 66 economic regions in each year, pooled across 11 years. The three territories (Yukon, the Northwest Territories and Nunavut) are excluded, and some small economic regions are grouped together to obtain larger sample sizes.Note 13

Microdata from the CEEDD and the LFS are used for estimating Equation (2). A transition into PS education is identified through a movement from non-enrolment in one year to enrolment in the next year, using CEEDD data. Aggregating these individual-level transitions across all workers aged 35 to 44 (or 45 to 54) within an economic region yields Y rt MathType@MTEF@5@5@+= feaagKart1ev2aaatCvAUfeBSjuyZL2yd9gzLbvyNv2CaerbuLwBLn hiov2DGi1BTfMBaeXatLxBI9gBaerbd9wDYLwzYbItLDharqqtubsr 4rNCHbGeaGqiVu0Je9sqqrpepC0xbbL8F4rqqrFfpeea0xe9Lq=Jc9 vqaqpepm0xbba9pwe9Q8fs0=yqaqpepae9pg0FirpepeKkFr0xfr=x fr=xb9adbaqaaeGaciGaaiaabeqaamaabaabaaGcbaGaamywamaaBa aaleaaieGacaWFYbGaa8hDaaqabaaaaa@38F3@ , the percentage of adult employees of a given age in region r who make a transition into PS education from year t to year t +1. Likewise, CEEDD microdata on permanent layoffs are aggregated into L rt MathType@MTEF@5@5@+= feaagKart1ev2aaatCvAUfeBSjuyZL2yd9gzLbvyNv2CaerbuLwBLn hiov2DGi1BTfMBaeXatLxBI9gBaerbd9wDYLwzYbItLDharqqtubsr 4rNCHbGeaGqiVu0Je9sqqrpepC0xbbL8F4rqqrFfpeea0xe9Lq=Jc9 vqaqpepm0xbba9pwe9Q8fs0=yqaqpepae9pg0FirpepeKkFr0xfr=x fr=xb9adbaqaaeGaciGaaiaabeqaamaabaabaaGcbaacbiGaa8htam aaBaaaleaacaWFYbGaa8hDaaqabaaaaa@38E2@ , the percentage of workers of a given age in region r who are permanently laid off in year t.

The other regressors used in Equation (2) include average real hourly wages (in 2002 dollars) of employees aged 35 to 44 (or 45 to 54) in a given economic region in a given year, as well as the percentage of employees of the age group who (a) are immigrants, (b) have a bachelor’s degree or more, (c) are employed full time in their main job, (d) are married and have a spouse who is employed full time, and (e) are aged 35 to 39 (for the sample of 35- to 44-year-olds) or 45 to 49 (for the sample of 45- to 54-year-olds). While the percentage of immigrant employees comes from the CEEDD, all other regressors are obtained from the LFS. Like Equation (1), Equation (2) is estimated separately for men and women and uses standard errors that are clustered at the economic-region level.

Even though Equation (2) is estimated by aggregating microdata from the 100% versions of the T1, T4 and ROE files, measurement error might still affect estimates of the key regressor, L rt MathType@MTEF@5@5@+= feaagKart1ev2aaatCvAUfeBSjuyZL2yd9gzLbvyNv2CaerbuLwBLn hiov2DGi1BTfMBaeXatLxBI9gBaerbd9wDYLwzYbItLDharqqtubsr 4rNCHbGeaGqiVu0Je9sqqrpepC0xbbL8F4rqqrFfpeea0xe9Lq=Jc9 vqaqpepm0xbba9pwe9Q8fs0=yqaqpepae9pg0FirpepeKkFr0xfr=x fr=xb9adbaqaaeGaciGaaiaabeqaamaabaabaaGcbaacbiGaa8htam aaBaaaleaacaWFYbGaa8hDaaqabaaaaa@38E2@ . This could happen if some of the longitudinal firm-level identifiers―for example, those in education, health care, social assistance and public administration―were somewhat imprecise, thereby leading some temporary layoffs to be labelled as permanent layoffs and vice versa.Note 14 Since the relative importance of the public sector is generally greater in small economic regions than in larger ones (Chart 1), imprecise firm-level identifiers in education, health care, social assistance and public administration will likely affect L rt MathType@MTEF@5@5@+= feaagKart1ev2aaatCvAUfeBSjuyZL2yd9gzLbvyNv2CaerbuLwBLn hiov2DGi1BTfMBaeXatLxBI9gBaerbd9wDYLwzYbItLDharqqtubsr 4rNCHbGeaGqiVu0Je9sqqrpepC0xbbL8F4rqqrFfpeea0xe9Lq=Jc9 vqaqpepm0xbba9pwe9Q8fs0=yqaqpepae9pg0FirpepeKkFr0xfr=x fr=xb9adbaqaaeGaciGaaiaabeqaamaabaabaaGcbaacbiGaa8htam aaBaaaleaacaWFYbGaa8hDaaqabaaaaa@38E2@  more in small economic regions than in larger ones. Since women are employed in the public sector to a greater extent than men, measurement error issues might be more salient for them than for their male counterparts. If so, one strategy to minimize concerns with measurement error in layoff rates is to estimate versions of Equation (2) in which greater weights are given to larger economic regions.Note 15 For this reason, grouped data results will be shown using both weighted and unweighted versions of Equation (2).

Chart 1 Percentage of jobs in education, health care, social assistance and public administration, by size of economic region, 2001-to-2011 averages

Data table for Chart 1
Percentage of jobs in education, health care, social assistance and public administration, by size of economic region, 2001-to-2011 averages
Table summary
This table displays the results of Percentage of jobs in education. The information is grouped by Economic region number (appearing as row headers), Employees aged 35 to 44 and Public sector jobs, calculated using thousands and percent units of measure (appearing as column headers).
Economic region numberNote 1 Employees aged 35 to 44 Public sector jobs
  thousands percent
1010 25.8 38.4
1021 13.1 31.6
1030 10.3 37.5
1110 13.1 37.9
1210 11.6 36.0
1220 14.8 29.2
1230 11.4 29.5
1240 11.4 28.3
1250 42.6 32.4
1310 14.9 34.5
1320 20.8 28.7
1330 17.4 26.9
1340 13.7 39.2
1350 7.9 30.1
2410 7.4 37.4
2415 16.8 30.5
2420 64.4 34.2
2425 39.2 24.2
2430 27.3 28.7
2433 21.5 20.6
2435 145.2 24.1
2440 178.5 23.6
2445 38.3 24.3
2450 46.7 26.2
2455 54.1 25.7
2460 38.6 49.7
2465 15.4 28.8
2470 23.1 28.9
2475 24.1 29.1
2481 11.9 28.0
3510 132.7 39.8
3515 41.3 33.9
3520 32.2 30.5
3530 604.7 18.8
3540 129.0 24.6
3550 141.6 25.2
3560 65.9 27.4
3570 65.8 24.9
3580 25.9 23.0
3590 54.6 35.2
3595 23.3 36.3
4610 9.3 31.6
4621 7.9 32.8
4630 8.9 40.4
4650 70.2 31.5
4660 8.5 34.4
4671 7.1 41.8
4710 27.3 32.1
4720 7.7 34.0
4730 28.4 34.2
4740 6.3 31.6
4751 16.5 37.2
4810 22.8 27.2
4820 16.9 29.2
4830 134.6 20.1
4841 30.9 22.1
4850 16.7 28.9
4860 113.8 28.3
4880 12.2 21.2
5910 60.3 34.7
5920 250.6 24.9
5930 40.7 30.0
5940 12.1 24.2
5950 15.4 24.4
5961 8.9 28.7
5980 6.2 19.9

4. Results

4.1 Descriptive evidence

In general, relatively few adult workers attend PS institutions. Over the 11-year period during which they were tracked, between 2.4% and 4.1% of the adults who were aged 35 to 44 in 2001 and who were selected in the sample used for Equation (1) attended PS education institutions in a given year (Appendix Table 1). Part-time enrolment among female adults averaged 3.0%, almost twice the rate of 1.6% observed among their male counterparts.Note 16 Full-time enrolment among female adults averaged 1.4%, compared with 1.0% for male adults.

For the samples used for Equation (2), between 1.9% and 2.8% of adult employees who were aged 35 to 44 in year t and were not students that year made a transition into PS education the following year (Appendix Table 2). The corresponding percentages vary between 1.0% and 1.6% for their counterparts aged 45 to 54.Note 17

While the average rates of enrolment of adults into PS education institutions are relatively low, the temporal patterns laid-off workers display are markedly different from those of other workers. For instance, as they move from ages 35 to 44 to ages 45 to 54, male adults in the control group used to estimate Equation (1) saw their rates of full-time enrolment drop almost linearly from about 1.4% in 2001 to 0.4% in 2011 (Chart 2). In contrast, their counterparts who were permanently laid off in 2008 experienced a sharp increase in full-time enrolment, from 1.6% in 2007 to almost 4.0% in 2009. The same qualitative patterns are observed for female adults (Chart 3). The substantial increases in full-time enrolment observed among men and women displaced in 2008 suggest that adult workers respond to job loss by enrolling full time in PS institutions. To assess whether this conclusion holds in a multivariate setting, regression results are presented.

Chart 2 Percentage of male adults attending postsecondary institutions full time, 2001 to 2011

Data table for Chart 2
Chart 2 Percentage of male adults attending postsecondary institutions full time, 2001 to 2011
Table summary
This table displays the results of Chart 2 Percentage of male adults attending postsecondary institutions full time. The information is grouped by Year (appearing as row headers), Control group and 2008 treatment group (appearing as column headers).
Year Control group 2008 treatment group
2001 1.44 1.52
2002 1.27 1.50
2003 1.06 1.20
2004 0.93 1.29
2005 0.84 1.11
2006 0.83 1.36
2007 0.70 1.62
2008 0.62 3.14
2009 0.56 3.82
2010 0.49 1.98
2011 0.42 1.19

Chart 3 Percentage of female adults attending postsecondary institutions full time, 2001 to 2011

Data table for Chart 3
Chart 3 Percentage of female adults attending postsecondary institutions full time, 2001 to 2011
Table summary
This table displays the results of Chart 3 Percentage of female adults attending postsecondary institutions full time. The information is grouped by Year (appearing as row headers), Control group and 2008 treatment group (appearing as column headers).
Year Control group 2008 treatment group
2001 2.11 1.74
2002 1.84 1.64
2003 1.62 1.38
2004 1.46 1.49
2005 1.33 1.59
2006 1.30 1.86
2007 1.14 1.99
2008 1.04 3.07
2009 0.90 4.41
2010 0.82 2.56
2011 0.70 1.49

4.2 Regression results

4.2.1 Microdata

Table 1 shows the results of Equation (1) for men. Marginal effects of being laid off are shown from ordinary least squares (OLS) models, fixed-effects models and probit models, for total enrolment, full-time enrolment and part-time enrolment. The probit models are included because they might produce different marginal effects than the OLS models (Wooldridge 2010; Lewbel, Dong and Yang 2012). However, neither the OLS models nor the probit models account for individual-level fixed effects. Thus, the preferred model is the fixed-effects model, and this is the one that will constitute the main focus of the discussion.Note 18

The results generally point to a positive relationship between the occurrence of a layoff and the attendance of PS education institutions by adults. This is particularly the case during the year of the layoff ( t ) or the year after ( t +1 ). For example, the fixed-effects model shows that experiencing a layoff is roughly associated with a 2.5-percentage-point increase in the probability of men transitioning into PS education during the year of the layoff or the year after (Table 1). Most of the increase is driven by full-time PS attendance (1.9 to 2.0 percentage points),Note 19 as opposed to part-time attendance (0.6 to 0.7 percentage points). Both the full-time and part-time enrolment effects are statistically significant at the 0.1% level. Results from the OLS models and probit models are similar.

Male enrolment in PS institutions also tends to increase prior to layoffs, as well as several years after layoffs. For example, men who experience a layoff are 0.8 percentage points more likely to attend PS institutions two years before the layoff than other men (according to the fixed-effects model). Five years after the layoff, they are 0.5 percentage points more likely than other men to attend PS institutions. The positive estimates obtained before layoffs are likely the result of pre-emptive enrolment among men who eventually experienced a layoff, while delayed enrolment may have followed a period of job search for displaced male workers.Note 20

The results for women are similar in most cases. For example, the estimated marginal effect on total PS attendance is 3.1 percentage points in year t and 3.7 percentage points in year t +1, according to the fixed-effects model (Table 2). Once again, most of the increase in attendance is associated with an increase in full-time attendance.

As is the case for men, statistically significant effects are estimated in the years leading up to job loss, as well as for several years after job loss. According to the fixed-effects model, women who experience a layoff are 0.8 percentage points more likely than other women to attend PS institutions two years before the layoff. Five years after the layoff, they are also 0.8 percentage points more likely than other women to attend PS institutions. Regardless of the models used, statistically significant correlations between layoffs and full-time attendance are detected between two years before layoffs and four years after layoffs, for both sexes.Note 21

Whether higher unemployment in a given economic region is associated with higher rates of enrolment in PS education institutions differs for adult men and women. The fixed-effects model suggests that a 1-percentage-point increase in unemployment is associated with a 0.025‑percentage-point increase in the probability of full-time enrolment among men (Table 1), from a baseline rate of 1.0% (Appendix Table 1). Parameter estimates from OLS models and probit models are smaller and statistically significant only at the 10% level (Table 1). In contrast, none of the marginal effects of unemployment shown in Table 2 for women are statistically significant.

Turning to models that include person-specific trends, Table 3 shows that statistically significant correlations between layoffs and full-time attendance are detected between two years before layoffs and two years after layoffs for both sexes. For example, male and female laid-off workers are, one year after job loss, 1.6 percentage points and 2.5 percentage points, respectively, more likely to attend PS institutions than other male and female workers.

4.2.2 Grouped data

Table 4 shows the results from Equation (2) estimated for the entire set of adult employees who are not students in year t. This includes workers who have been laid off during that year, as well as workers who have not been laid off yet.

Weighted regressions that allow for distinct linear trends for each economic region indicate that, for men aged 35 to 44, a 1-percentage-point increase in layoff rates is associated with a 0.048‑percentage-point increase in the proportion of men who make transitions to PS education, from a baseline rate of 1.9% (Appendix Table 2). The corresponding number from unweighted regressions―where each economic region carries the same weight―is, at 0.075 percentage points, somewhat higher. These two estimates imply that, for every additional 100 men aged 35 to 44 laid off in a given economic region, an additional 5 to 8 men aged 35 to 44 enrol in PS education institutions in the following year. Compared with these marginal effects, somewhat smaller marginal effects are found for men aged 45 to 54, although those based on weighted data are not statistically significant.

Increases in full-time enrolment account for all of the positive association between layoff rates and adult education for men. For instance, weighted and unweighted regressions indicate that a 1-percentage-point increase in the layoff rate of men aged 35 to 44 in a given economic region is associated with an increase in the proportion of men aged 35 to 44 who make transitions to full-time enrolment. The increase equals 0.063-percentage-point in weighted regressions and 0.054-percentage-point in unweighted regressions. The corresponding estimates drop to about 0.030 percentage points, but are statistically significant at the 10% level, when a more flexible specification that uses quadratic trends at the economic-region level is used for Equation (2). Layoff effects on full-time enrolment that are statistically significant at the 5% level are also detected for men aged 45 to 54, whether region-specific linear trends or region-specific quadratic trends are used. Overall, the third column of Table 4 indicates that for every additional 100 adult men laid off in an economic region in a given year, there is an additional 2 to 6 men who enrol in PS education institutions on a full-time basis in the following year.

In contrast, there is little evidence in Table 4 of a relationship between layoff rates and PS transitions for women at the economic-region level. This finding may hide differences between married women and unmarried women. Table 5 confirms this hypothesis. Versions of Equation (2) based on weighted data―which minimize concerns with measurement error in layoff rates at the economic-region level―yield a statistically significant relationship between movements in regional layoff rates and transitions to full-time enrolment for unmarried women aged 35 to 44 and for those aged 45 to 54, but not for their married counterparts. Weighted versions of Equation (2) suggest that for every additional 100 unmarried women laid off in an economic region in a given year, there is an additional 4 to 5 unmarried women who enrol in PS education institutions on a full-time basis in the following year. In contrast, results generally do not differ much by marital status for men.

The results shown in Tables 4 and 5 seek to capture the total impact of movements in regional layoff rates on transitions to PS education through the three distinct channels previously identified in this paper. One testable implication of the second and third channels―which refer to non-laid-off workers employed in distressed firms and in other firms―is that movements in regional layoff rates might be positively correlated with short-term transitions to PS institutions for adult male workers who have not been laid off yet.

Table 6 tests this hypothesis by computing Y rt MathType@MTEF@5@5@+= feaagKart1ev2aaatCvAUfeBSjuyZL2yd9gzLbvyNv2CaerbuLwBLn hiov2DGi1BTfMBaeXatLxBI9gBaerbd9wDYLwzYbItLDharqqtubsr 4rNCHbGeaGqiVu0Je9sqqrpepC0xbbL8F4rqqrFfpeea0xe9Lq=Jc9 vqaqpepm0xbba9pwe9Q8fs0=yqaqpepae9pg0FirpepeKkFr0xfr=x fr=xb9adbaqaaeGaciGaaiaabeqaamaabaabaaGcbaGaamywamaaBa aaleaaieGacaWFYbGaa8hDaaqabaaaaa@38F3@  for the subset of adult employees who have not been laid off in year t and regressing Y rt MathType@MTEF@5@5@+= feaagKart1ev2aaatCvAUfeBSjuyZL2yd9gzLbvyNv2CaerbuLwBLn hiov2DGi1BTfMBaeXatLxBI9gBaerbd9wDYLwzYbItLDharqqtubsr 4rNCHbGeaGqiVu0Je9sqqrpepC0xbbL8F4rqqrFfpeea0xe9Lq=Jc9 vqaqpepm0xbba9pwe9Q8fs0=yqaqpepae9pg0FirpepeKkFr0xfr=x fr=xb9adbaqaaeGaciGaaiaabeqaamaabaabaaGcbaGaamywamaaBa aaleaaieGacaWFYbGaa8hDaaqabaaaaa@38F3@  on year effects, region-specific fixed effects, region-specific linear or quadratic trends, L rt MathType@MTEF@5@5@+= feaagKart1ev2aaatCvAUfeBSjuyZL2yd9gzLbvyNv2CaerbuLwBLn hiov2DGi1BTfMBaeXatLxBI9gBaerbd9wDYLwzYbItLDharqqtubsr 4rNCHbGeaGqiVu0Je9sqqrpepC0xbbL8F4rqqrFfpeea0xe9Lq=Jc9 vqaqpepm0xbba9pwe9Q8fs0=yqaqpepae9pg0FirpepeKkFr0xfr=x fr=xb9adbaqaaeGaciGaaiaabeqaamaabaabaaGcbaacbiGaa8htam aaBaaaleaacaWFYbGaa8hDaaqabaaaaa@38E2@  and U ' rt MathType@MTEF@5@5@+= feaagKart1ev2aaatCvAUfeBSjuyZL2yd9gzLbvyNv2CaerbuLwBLn hiov2DGi1BTfMBaeXatLxBI9gBaerbd9wDYLwzYbItLDharqqtubsr 4rNCHbGeaGqiVu0Je9sqqrpepC0xbbL8F4rqqrFfpeea0xe9Lq=Jc9 vqaqpepm0xbba9pwe9Q8fs0=yqaqpepae9pg0FirpepeKkFr0xfr=x fr=xb9adbaqaaeGaciGaaiaabeqaamaabaabaaGcbaacbiGaa8xvai aa=DcadaWgaaWcbaGaa8NCaiaa=rhaaeqaaaaa@3993@ .Note 22 There is evidence that adult male workers who have not yet been laid off respond to movements in regional layoff rates. Using models that include region-specific linear trends, a 1-percentage-point increase in layoff rates in a given economic region is, for male employees aged 35 to 44 who have not been laid off in year t, associated with an increase in transitions to full-time enrolment that varies between 0.029 percentage points and 0.035 percentage points.

5. Conclusion

Understanding the relationship between job loss and postsecondary (PS) attendance is important, given the well-established literature indicating a negative relationship between job displacement and subsequent earnings. Programs already exist to assist displaced workers by offering government-sponsored training. This study provides complementary information by assessing the degree to which Canadian workers adjust to job loss or rising layoff rates in their region through self-financed PS attendance.

Using longitudinal microdata, the study shows that, regardless of their gender and marital status, laid-off employees are more likely than other employees to attend PS institutions in the year of the layoff or the following year. The study also finds that laid-off employees appear to respond to job loss a few years before layoffs occur and that their response―in terms of increased schooling―spans several years.

Using grouped data, the study finds that for every additional 100 adult men laid off in an economic region in a given year, there is an additional 2 to 6 men who enrol in PS education institutions on a full-time basis in the following year. A positive relationship between regional layoff rates and regional PS transitions is also observed for unmarried women. In line with the notion that some non-laid-off workers may pre-emptively enrol in PS education institutions, the study finds evidence that movements in regional layoff rates are positively correlated with short-term transitions to PS institutions for adult male workers aged 35 to 44 who have not been laid off yet.

To date, a massive literature has assessed the causal impact of numerous factors―for example, classroom size, teacher quality and peer effects―on the school achievement of children. Yet, relatively little is known about the determinants of adult education. Taken together, the results from individual-level analyses and group-level analyses provide compelling evidence that job loss is one determinant of adult education.

References

Bonikowska, A., and R. Morissette. 2012. Earnings Losses of Displaced Workers with Stable Labour Market Attachment: Recent Evidence from Canada. Analytical Studies Branch Research Paper Series, no. 346. Statistics Canada Catalogue no. 11F0019M. Ottawa: Statistics Canada.

Dar, A., and I. Gill. 1998. “Evaluating retraining programs in OECD countries: Lessons learned.” World Bank Research Observer 13 (1): 79–101.

Decker, P., and W. Corson. 1995. “International trade and worker displacement: Evaluation of the Trade Adjustment Assistance Program.” Industrial and Labor Relations Review 48 (4): 758–774.

Eliason, M., and D. Storrie. 2006. “Lasting or latent scars? Swedish evidence on the long-term effects of job displacement.” Journal of Labor Economics 24 (4): 831–856.

Frenette, M. 2003. Access to College and University: Does Distance to School Matter? Analytical Studies Branch Research Paper Series, no. 201. Statistics Canada Catalogue no. 11F0019M. Ottawa: Statistics Canada.

Frenette, M., R. Upward, and P.W. Wright. 2011. The Long-term Earnings Impact of Postsecondary-education Following Job Loss.  Analytical Studies Branch Research Paper Series, no. 334. Statistics Canada Catalogue no. 11F0019M. Ottawa: Statistics Canada.

Heckman, J.J., R.J. Lalonde, and J.A. Smith. 1999. “The economics and econometrics of active labor market programs.” In Handbook of Labor Economics, ed. O. Ashenfelter and D. Card, vol. 3. Chapter 31, p. 1865–­­2097. Amsterdam: North Holland.

Heinrich, C.J., P.R. Mueser, and K.R. Troske. 2008. Workforce Investment Act Non-experimental Net Impact Evaluation. Final Report. Columbia, Maryland: IMPAQ International.

Hijzen, A., R. Upward, and P.W. Wright. 2010. “The income losses of displaced workers.” Journal of Human Resources 45 (1): 243–269.

Huttunen, K., J. Møen, and K. Salvanes. 2006. How Destructive is Creative Destruction? Investigating Long-term Effects of Worker Displacement. IZA Discussion Paper 2316. Bonn, Germany: Institute for the Study of Labor.

Jacobson, L., R.J. Lalonde, and D.G. Sullivan. 1993. “Earnings losses of displaced workers.” American Economic Review 83 (4): 685–709.

Jacobson, L., R.J. Lalonde, and D.G. Sullivan. 2005. “Estimating the returns to community college schooling for displaced workers.” Journal of Econometrics 125 (1–2): 271–304.

Leigh, D. 1994. Retraining Displaced Workers: The US Experience. Training Policy Study 1. Geneva, Switzerland: International Labour Office.

Lewbel, A., Y. Dong, and T.T. Yang. 2012. “Comparing features of convenient estimators for binary choice models with endogenous regressors.” Canadian Journal of Economics 45 (3): 809–829.

Morissette, R., W. Ci, and G. Schellenberg. 2016. Hiring and Layoff Rates by Economic Region of Residence: Data Quality, Concepts and Methods. Analytical Studies: Methods and References, no. 1. Statistics Canada Catalogue no. 11-633-X. Ottawa: Statistics Canada.

Morissette, R., H. Qiu, and P.C.W. Chan. 2013. “The risk and cost of job loss in Canada, 1978–2008.” Canadian Journal of Economics 46 (4): 1480–1509.

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. Analytical Studies Branch Research Paper Series, no. 291. Statistics Canada Catalogue no. 11F0019M. Ottawa: Statistics Canada.

OECD (Organisation for Economic Co-operation and Development). 2013. Back to Work: Re‑employment, Earnings, and Skill Use after Job Displacement. Employment Analysis and Policy Division, Directorate for Employment, Labour and Social Affairs. Paris, France: OECD.

Wooldridge, J.M. 2010. Econometric Analysis of Cross Section and Panel Data. 2nd ed. Cambridge: MIT Press.

Date modified: