# Analytical Studies: Methods and ReferencesHiring and Layoff Rates by Economic Region of Residence: Data Quality, Concepts and MethodsAnalytical Studies: Methods and ReferencesHiring and Layoff Rates by Economic Region of Residence: Data Quality, Concepts and Methods

By René Morissette, Wen Ci, and Grant Schellenberg
Social Analysis and Modelling Division

Release date: June 27, 2016 Correction date: (if required)

Text begins

## Abstract

Every year, thousands of workers lose their jobs as firms reduce the size of their workforce in response to growing competition, technological changes, changing trade patterns and numerous other factors. Thousands of workers also start a job with a new employer as new firms enter a product market and existing firms expand or replace employees who recently left. This worker reallocation process across employers is generally seen as contributing to productivity growth and rising living standards. To measure this labour reallocation process, labour market indicators such as hiring rates and layoff rates are needed. In response to growing demand for subprovincial labour market information and taking advantage of unique administrative datasets, Statistics Canada is producing hiring rates and layoff rates by economic region of residence. This document describes the data sources, conceptual and methodological issues, and other matters pertaining to these two indicators.

## 1. Introduction

Demand for labour market information at subprovincial levels of geography comes from many stakeholders. Information about local labour markets informs discussions about the state of the Canadian economy and the challenges and opportunities faced by firms and individuals in specific areas.

Administrative data files, such as those containing records of the T1 Income Tax Return and the T4 Statement of Remuneration Paid (T4 slip), are valuable sources of information from which small-area labour market information can be derived. Such files contain the large number of observations needed to generate reliable estimates for small areas as well as the postal code information needed to organize these estimates into subprovincial geographic areas.

Information from several administrative data files has been used to create hiring rates and layoff rates for 69 economic regions across Canada.

This document describes the data sources, conceptual and methodological issues, and other matters pertaining to these two indicators.

Because the subprovincial information available in the aforementioned data sets relates to the location of residence, the labour market indicators discussed in this document are defined at the economic region of residence, rather than the economic region of employment. Hence, the indicators will shed light on how residents of a given economic region fare in the Canadian labour market rather than how the economy of their region fares compared to other local labour markets. Readers should keep this distinction in mind throughout the document.

## 2. Data sources

The labour market indicators described in this document are estimated using a subset of linked administrative data files from the Canadian Employer–Employee Dynamics Database (CEEDD). The CEEDD contains information on all firms in Canada that filed a T2 Corporation Income Tax Return, issued a T4 Statement of Remuneration Paid (T4 slip), or remitted a PD7 (statement of account for current source deductions) to the Canada Revenue Agency, as well as information on the paid workers they employ. The administrative data files used to construct labour market indicators of economic regions of residence in Canada are:

• T1 Personal Master File (T1PMF) from the Canada Revenue Agency: Information on the demographic and financial characteristics of individuals is drawn from the T1 tax records.
• T4 records from the Canada Revenue Agency: Job-level information on employment income and the pension adjustment amount is drawn from T4 records.
• Record of Employment (ROE) from Employment and Social Development Canada: Job-level information is drawn on the reason for job termination.
• National Accounts Longitudinal Microdata File (NALMF): The NALMF, constructed and maintained by Statistics Canada, contains employment information on businesses in Canada (both incorporated and unincorporated) that issue a T4 slip to one or more employees for tax purposes. This file is used to identify individual-level transitions between employers.Note 1

Indicators are provided for the period from 2003 to 2011.

### Some concepts: employees, workers, wages and salaries, and earnings

In this document, the terms ‘employees’ and ‘paid workers’ are used interchangeably and refer to individuals who have at least one paid job at some point in year $t$ but have no self-employment income during that year.Note 2 The term ‘workers’ includes both employees and self-employed individuals. Self-employed individuals are defined as individuals who have self-employment income in year $t$, regardless of whether they also have employment income from a paid job. Annual earnings equal annual wages and salaries plus net income from self-employment.

## 3. Economic region of residenceNote 3

An economic region (ER) is a grouping of complete census divisions (CDs) (with one exception in Ontario) created as a standard geographic area for analysis of regional economic activity. Such an area is small enough to permit regional analysis, yet large enough to still be able to release a broad range of statistics after data are screened for confidentiality.

The regions are based upon work by Camu, Weeks and Sametz (1964). At the outset, boundaries of regions were drawn in such a way that similarities of socio-economic features within regions were maximized while those among regions were minimized. Later, the regions were modified to consist of counties which define the zone of influence of a major urban centre or metropolitan area. Finally, the regions were adjusted to accommodate changes in CD boundaries and to satisfy provincial needs.

An ER is a geographic area, smaller than a province, except in the case of Prince Edward Island and the Northwest Territories. The ER is made up by grouping whole CDs, except for one case in Ontario, where the city of Burlington, a component of Halton (CD 35 24), is excluded from the ER of Toronto (ER 35 30) and is included in the Hamilton–Niagara Peninsula ER (ER 35 50), which encompasses the entire census metropolitan area (CMA) of Hamilton.

ERs may be economic, administrative or development regions. Within the province of Quebec, ERs are designated by law (les régions administratives). In all other provinces, ERs are created by agreement between Statistics Canada and the provinces concerned.

The labour market indicators presented in this document are based on individuals’ ER of residence. Individuals’ ER of residence is derived from the postal code information on their T1 tax record. The postal codes from T1 tax records measure individuals’ ER of residence around December of year $t+1$, i.e., at the time the T1PMF is created.Note 4

### Comparing the economic region of residence and the economic region of work

As noted above, the ER of residence refers to the location in which Canadians live, not the location in which they work. Some residents of ER ‘a’ (for example, Laval) may be employed in ER ‘b’ (for example, Montréal) and conversely, some residents of ER ‘b’ may be employed in ER ‘a’. The 2006 Census provides some information on this issue as long-form respondents were asked where they had worked during the census reference week (i.e., the week prior to May 16, 2006) or, if they were not employed that week, where their longest job was located during the previous year.

Table 1 selects individuals aged 18 to 64 who were employed as paid workers during the census reference week and shows what percentage worked in their ER of residence at that point in time. Overall, 91% of these employees worked in their ER of residence.

This average masks important differences across ERs. While 9 Montréal residents out of 10 worked in (the ER of) Montréal, no more than 4 Laval residents out of 10 worked in (the ER of) Laval. Likewise, while 94% of Ottawa residents worked in Ottawa, less than two-thirds of Outaouais residents worked in the Outaouais. In 52 ERs out of 69, 90% or more of employed individuals worked in their ER of residence. These ERs account for 77% of the population of employed individuals. Hence, for the majority of ERs and residents, the concept of ER of residence is closely tied with the concept of ER of employment.

Nevertheless, the fact that in some cases, most residents work outside their ER of residence is important. It highlights the importance of reminding data users that the labour market indicators provided will shed light on how residents of a given ER fare in the Canadian labour market rather than how the economy of their region fares compared to other local labour markets.

## 4. Comparing the Canadian Employer–Employee Dynamics Database with the 2006 Census

Because hiring rates and layoff rates will be computed at the ER of residence level, a key question is whether the CEEDD results are representative of the population of each ER. To gain some insight into this issue, CEEDD results are compared to those from the 2006 Census of Population.Note 5 Specifically, a sample of employees (individuals who were aged 18 to 64 in 2005 and had positive wages and salaries but no self-employment income in that year) is selected from the two data sources. Since the T1PMF used in the CEEDD do not include late tax filers and since late tax filers represent about 5% of all tax filers (Messacar 2014), one would expect estimates from the CEEDD to be about 5% lower than those from the 2006 Census.

### Employment estimates

This is indeed the case. The resulting CEEDD sample contains 13,353,124 individuals, which represents 95% of the corresponding (weighted) estimate obtained from the 2006 Census (14,029,879). Table 2 compares, for each ER, the number of employees aged 18 to 64 in 2005, as measured with the CEEDD, with the corresponding estimate from the 2006 Census. In 40 of the 69 ERs of residence, the CEEDD estimates are within plus or minus 4% of the Census estimates. The CEEDD estimates are within 6% of the 2006 Census estimates in 53 of the 69 ERs, and within 8% of the Census estimates in 62 of the 69 ERs of residence. The CEEDD estimates are less than 90% of the Census estimates in three ERs of residence (Chart 1), with all but one of these in the northern part of their respective provinces.

### Sex and age groups

Table 3 compares the proportion of the samples from the two data sets composed of female employees. In 63 of the 69 ERs, the female share of the two samples is within 1 percentage point. Of the remaining six regions, four are located either in Nunavut or in the northern part of their respective provinces (i.e., Nord-du-Québec, Northern Manitoba and Northern Saskatchewan). Overall, the representation of women is very similar in the two data sets (Chart 2).

Table 4 compares the age distributions obtained with the two data sets for men. With the exception of Yorkton–Melville and Prince Albert (both located in Saskatchewan), the mean absolute deviation between the estimates of the percentage of men in a given age group (18 to 24; 25 to 34; 35 to 44; 45 to 54; 55 to 64) obtained with the CEEDD and with the 2006 Census generally amounts to 2.0 percentage points or less, from baseline proportions that vary between 14 and 25 percentage points at the national level.Note 6 As Table 5 and Charts 3 and 4 show, fairly similar patterns are observed for women.

Taken together, Tables 3 to 5 indicate that the distributions of employees by age and sex, defined at the ER level, are generally very similar in the two data sets.

### Annual wages and salaries

Table 6 compares mean annual wages and salaries, median annual wages and salaries, and the percentage of individuals earning at least $100,000 in wages and salaries across the two data sets.Note 7 At the national level, mean wages and salaries and median wages and salaries in the CEEDD are 1.8% and 4.1%, respectively, lower than those in the 2006 Census data. Average wages and salaries are within plus or minus 4% in about two thirds (48 out of 69) of ERs of residence, and within plus or minus 5% in 56 of the 69 regions. The median wages and salaries estimated using the CEEDD and 2006 Census are within plus or minus 4% in 24 of the 69 ERs of residence and within plus or minus 5% in 37 of the 69 regions. At the national level, the percentage of individuals earning at least$100,000 is about the same in the 2006 Census and CEEDD, at 4.0% and 3.9%, respectively. Within ERs of residence, the shares of individuals earning at least $100,000 are within 0.2 percentage points in 54 of the 69 ERs, with this representing a difference of 10% or less in 50 of them. The differences in median wages and mean wages discussed above are shown graphically in Charts 5 and 6. One discrepancy warrants note. Although the CEEDD and 2006 Census estimates of average wages and salaries differ by about 10% for Nunavut, the CEEDD estimate of median earnings for this territory are about 27% lower than those from the 2006 Census. For the production of labour market indicators at the ER level, a key question is whether cross-regional differences in earnings that are observed in 2006 Census data can also be found in the CEEDD. This is indeed the case. The Pearson correlation coefficient using the two data sources is 0.992 for mean wages and salaries, 0.978 for median wages and salaries, and 0.996 for the percentage of individuals earning at least$100,000 (Table 7). Hence, ERs that display relatively large (median or average) annual wages and salaries in 2006 Census data also exhibit relatively large wages and salaries in CEEDD data (Charts 7 to 9).

Overall, Tables 2 to 7 indicate that the CEEDD yields age–sex distributions and earnings estimates at the ER level that are quite consistent with those obtained from 2006 Census data. This in turn suggests that CEEDD data are well suited for the computation of additional labour market indicators at the ER level.

## 5. Indicators

Although Statistics Canada currently produces several labour market indicators at the ER level (Appendix 1) or at the CMA/CA (census agglomeration) level,Note 8 no subprovincial statistics are produced on two important aspects of the Canadian labour market:

• (a) Hiring rates
• (b) Layoff rates

Hiring rates capture movements of workers into firms. They measure the percentage of employees who start a job with a new employer in a given year and still hold a position with this employer in the following year. They may increase as firms expand, replace a growing number of retirees or employees leaving for other reasons, or start offering a growing number of temporary jobs.

Layoff rates capture movements of workers out of firms due to a shortage of work or the end of contracts.Note 9 They measure the percentage of employees who are laid-off in a given year and do not return to their original employer during that year or the following year. They may increase as employment in declining industries fall, as firms of a given industry downsize for a variety of reasons, or as contracts signed for a growing number of temporary jobs come to an end.

### Hiring rates

The hiring rates that are produced using CEEDD data are computed initially as follows:

$\begin{array}{cccc}Hiring\text{ }\text{ }\text{ }rate=\frac{number\text{ }\text{ }\text{ }of\text{ }\text{ }\text{ }employees\text{ }\text{ }\text{ }observed\text{ }\text{ }\text{ }in\text{ }\text{ }\text{ }a\text{ }\text{ }\text{ }firm\text{ }\text{ }\text{ }in\text{ }\text{ }\text{ }years\text{ }\text{ }\text{ }t\text{ }\text{ }\text{ }and\text{ }\text{ }\text{ }t+1\text{ }\text{ }\text{ }but\text{ }\text{ }\text{ }not\text{ }\text{ }\text{ }in\text{ }\text{ }\text{ }year\text{ }\text{ }\text{ }t-1}{Labour\text{ }\text{ }\text{ }Force\text{ }\text{ }\text{ }Survey\text{ }\text{ }\text{ }average\text{ }\text{ }\text{ }annual\text{ }\text{ }paid\text{ }\text{ }\text{ }employment\text{ }\text{ }\text{ }in\text{ }\text{ }\text{ }year\text{ }\text{ }\text{ }t\text{ }\text{ }\text{ }and\text{ }\text{ }year\text{ }\text{ }\text{ }t-1}& & & \left(1\right)\end{array}$

This hiring rate concept was selected after considering three questions. First, should hiring rates be computed at the person level or at the job level? Second, should hiring rates include all workers who have been hired in a given year, regardless of their employment status in the following year, or should they restrict attention to those newly hired individuals who are still employed in the following year?  Third, should the denominator used to compute hiring rates measure the number of individuals who have been employed at some point during the year—as measured with administrative data—or should it measure average annual paid employment in that year (and/or the previous year)?

In principle, estimates of hiring can be computed both at the job level and at the person level. These units of analysis measure different concepts. Job-level estimates of hiring capture the number of employer–employee pairings that were newly created in year $t$, while person-level estimates of hiring capture the number of individuals who started at least one job with a new employer in year $t$. Since the same person can be hired several times by various employers in a given year, job-level estimates will be substantially higher than person-level estimates. At the national level, job-level estimates of hiring exceed person-level estimates by a factor of 1.4, on average (Morissette and Qiu 2012).

The hiring rates computed for the 69 ERs of residence using the CEEDD are calculated at the person-level for two reasons. First, doing so allows estimates to be benchmarked, at the provincial and national levels, with the Labour Force Survey (LFS). Second, this approach is consistent with the approach taken by the OECD (2009).

When measuring hiring at the person level, estimates of the number of hires can be computed in three different ways, reflecting different treatments of individuals’ employment in year $t+1$:

• Unconditional hires: the number of hires in year $t$ is estimated as (i) the number of employees aged 18 to 64 who started a job with (at least) one new employer in year $t$, regardless of whether these individuals are employed the following year—that is in year $t+1$;
• Conditional hires: the number of hires in year $t$ is estimated as (i) the number of employees aged 18 to 64 who started a job with (at least) one new employer in year $t$ and (ii) who were still employed with any employer in year $t+1$.
• OECD (2009) hires: in line with OECD (2009), the number of hires in year $t$ is estimated as (i) the number of employees who started a job with (at least) one new employer in year $t$ and (ii) who were still employed with the same employer in year $t+1$.

The distinction matters empirically. For example, at the national level about 3.95 million individuals aged 18 to 64 started at least one job with a new employer in 2011. Of these, 3.70 million were still employed as paid workers in 2012. A subset of these—2.40 million—were still employed with their new employer in 2012. These differences arise from the fact that while unconditional hires and conditional hires provide fairly exhaustive measures of the number of individuals who start a new job in a given year, they include many individuals who have a marginal labour market attachment. As a result, they tend to overestimate the hiring rates faced by workers who have a stronger labour market attachment.

In line with OECD (2009), the hiring rates computed for the 69 ERs of residence using the CEEDD use as a numerator the third metric; i.e., the number of employees who started a job with (at least) one new employer in year $t$ and who were still employed with the same employer in year $t+1$.Note 10

As mentioned above, at least two options are available regarding the choice of the denominator used to compute hiring rates. The first option uses as a denominator the number of individuals who have been employed at some point during the year, as measured with administrative data. One advantage of this option is its simplicity: it allows one to compute both the numerator and the denominator using the CEEDD. One disadvantage is that this denominator is sensitive to exogenous changes in the number of individual transitions from non-employment to employment and from employment to non-employment that might occur even if the average annual paid employment (or average annual work hours) remains unchanged.Note 11

The second option is to use as a denominator the average annual paid employment, as measured from the LFS. While this denominator requires the use of an additional data set (LFS) for the computation of hiring rates, it is not sensitive to changes in transitions from non-employment to employment and from employment to non-employment that occur at constant employment levels. For this reason, this denominator is used for the computation of hiring rates. Specifically, average annual paid employment in year $t$ and in year $t-1$ is used to compute hiring rates.Note 12Note 13

### Comparing hiring rates from the Canadian Employer–Employee Dynamics Database and the Labour Force Survey

The OECD (2009) definition of hires shown above—requiring that hired individuals be employed by the same firm for two consecutive years—allows comparisons to be drawn between CEEDD- and LFS-based measures of hiring. Such a comparison can be performed as follows.

First, consider paid workers interviewed in the LFS in January of year $t+1$. Workers who report having been employed with their current employer for 12 months or less have, by definition, been hired between January of year $t$ and January of year $t+1$. As such, these workers approximate the number of individuals who were hired at some point in year $t$ and are still employed by the same firm in January of year $t+1$. Now consider the CEEDD. Select workers who: (a) are observed with the same firm in year $t$ and year $t+1$, and (b) were not observed in that firm prior to year $t$. Conditions (a) and (b) imply that these workers were hired at some point in year $t$ and—under the plausible assumption that the majority of employment spells with a firm are uninterrupted—are still with the same employer in January of year $t+1$.

The arguments above suggest that estimates of the number of paid workers with 12 months of seniority or less, obtained from the LFS in January of year $t+1$, should be fairly similar to estimates of the number of paid workers: (a) who are observed with the same firm in year $t$ and year $t+1$, and (b) were not observed in that firm prior to year $t$, when these estimates come from the CEEDD or alternative data sets—such as the Longitudinal Worker File (LWF)—that use input files very similar to those used in the CEEDD.Note 14

Chart 10 confirms this. It shows the hiring rate obtained from the LWF for the period from 1978 to 2010 and the LFS for the period from 1976 to 2011.Note 15Note 16 The LWF-based measure and the LFS yield similar trends and levels over time. Furthermore, the OECD (2009) definition of hiring tracks recessions and expansions quite well over the extended reference period.

Charts 11 to 14 compare the hiring rate derived from the CEEDD and the LFS for individuals aged 18 to 64 in Quebec, Ontario, Alberta, and British Columbia. With the exception of Quebec in 2005/2006, the hiring rates from the two sources display similar temporal movements. As expected, the CEEDD-based hiring rates fell from 2008 to 2009 in each of these provinces, as the Canadian economy entered a recession. The CEEDD-based hiring rate is also higher in Alberta than in the three other provinces, a finding consistent with the relatively strong economic activity in that province.

Table 8 shows the hiring rate obtained from the CEEDD and LFS for each province. Table 9 quantifies the degree to which the two series are correlated. Considering all provinces across the nine years of the 2003-to-2011 period, the Pearson correlation coefficient between the two series equals 0.674. Within provinces, temporal variations in hiring rates across the two data sets are more strongly correlated in Ontario and the Western provinces than they are in the Atlantic Provinces. This likely reflects the relatively high sampling variability of LFS estimates of the number of hires in the Atlantic Provinces.Note 17 Surprisingly, the correlation across years observed in Quebec is, at 0.383, relatively low. Within most years, cross-provincial differences in hiring rates from the CEEDD are reasonably correlated with those in the LFS (with a correlation coefficient of 0.550 or more being observed in seven years out of nine), thereby indicating that provinces that display relatively high hiring rates in a given year in one data set tend to display relatively high hiring rates in the alternative data set.

In sum, the CEEDD hiring rates generally display: (a) plausible temporal patterns, being lower in 2008/2009 than during previous years; (b) plausible cross-provincial differences, being higher in Alberta than in the three other large provinces; and (c) reasonable correlations with LFS hiring rates.

### Layoff rates

The layoff rates that are produced using CEEDD data are computed initially as follows:

$\begin{array}{cccc}Layoff\text{ }\text{ }rate=\frac{number\text{ }\text{ }of\text{ }\text{ }employees\text{ }\text{ }laid\text{ }\text{ }off\text{ }from\text{ }\text{ }\text{ }a\text{ }\text{ }firm\text{ }\text{ }\text{ }in\text{ }\text{ }year\text{ }\text{ }t\text{ }\text{ }and\text{ }\text{ }not\text{ }\text{ }returning\text{ }\text{ }\text{ }to\text{ }\text{ }firm\text{ }\text{ }\text{ }in\text{ }\text{ }year\text{ }\text{ }t+1}{Labour\text{ }\text{ }Force\text{ }\text{ }Survey\text{ }\text{ }average\text{ }\text{ }annual\text{ }\text{ }paid\text{ }\text{ }employment\text{ }\text{ }in\text{ }\text{ }year\text{ }\text{ }t\text{ }\text{ }and\text{ }\text{ }year\text{ }\text{ }t-1}& & & \left(2\right)\end{array}$

During recessions as well as expansionary periods, thousands of Canadians lose their job. Information on job losses is thus critical for understanding local labour markets. Because it uses the complete (100%) version of the ROE file, the CEEDD provides an accurate measurement of layoffs experienced by residents of a given ER.

The CEEDD allows the number of layoffs in Canada to be calculated on an annual basis using the ROE, which specifies the reason for the work interruption or separation. Separations due to “shortage of work” (code “A” on the ROE) are identified as layoffs.Note 18

The CEEDD file allows both temporary and permanent layoffs to be identified. A layoff is identified as temporary when the laid-off worker returns to his or her employer during the year of the layoff or in the following year. When such a return does not occur, the layoff is considered permanent.

The layoff rate concept defined above is based on permanent layoffs since job losses experienced by workers are of primary interest.

Before presenting statistics on permanent layoff rates, it is useful to check whether the number of jobs ending with a permanent or temporary layoff divided by the average level of paid employment, obtained from administrative data, displays plausible temporal variation. This is done in Chart 15, where the total layoff rate from the LWF is compared to that derived from the LFS.Note 19

As expected, both series rise sharply with the 1981/1982 recession, the 1990–1992 recession and the onset of the 2008/2009 recession. While the LWF layoff rate is somewhat lower than the LFS layoff rate from 1978 to 1996, both series are very similar afterwards. Thus, Chart 15 indicates that the layoff information contained in the ROE file (which is used to construct the LWF) yields a layoff rate that exhibits plausible temporal variation.

Chart 16 uses data from the LWF and shows that layoff rates based on permanent layoffs also display plausible patterns over the last three decades. Together, Charts 15 and 16 suggest that the ROE file can be used to produce sensible estimates of job losses.

Chart 17 compares the permanent layoff rates obtained from the CEEDD with those obtained from the Survey of Labour and Income Dynamics (SLID) when considering all provinces.Note 20 Over the 2003-to-2011 period, the two series track each other fairly well, even though SLID estimates are somewhat higher than those from the CEEDD.Note 21 Table 10 provides the province-specific permanent layoff rates resulting from each data set. Table 11 reports the Pearson correlation coefficients obtained with the two series. Considering all years of the 2003-to-2011 period and all provinces, the two series are highly correlated: they have a correlation coefficient of 0.915. For all years considered, cross-provincial differences in permanent layoff rates are also highly correlated, as the correlation coefficient varies between 0.714 and 0.978. Temporal movements in permanent layoff rates within provinces display smaller correlations. As Charts 18 to 21 show, this is particularly true for Quebec.

## 6. Refining the indicators

Before producing final estimates, the hiring rates and layoff rates defined in Equations (1) and (2) are subject to a few additional adjustments.

First, employees returning to their employer after parental leave are removed from the estimates of new hires. Second, as is the case in the LFS, full-time members of the Armed Forces and individuals on reserves are excluded. Third, a special algorithm is used to determine hires and layoffs among employees working in Education, Health Care and Social Assistance, and Public Administration.  Doing so is necessary to minimize the impact on estimates of hires and layoffs of false changes in the longitudinal employer identifiers that might occur in these sectors. As Table 12 shows, in years during which layoff rates in Public Administration increase substantially, a large proportion of the individuals who (based on Equation [2]) appear to be permanently laid-off from this sector end up being reemployed in the same 3-digit industry in year $t+1$. This pattern suggests that many of these individuals actually remained with the same employer but that the longitudinal firm identifiers erroneously changed from one year to the next.

For this reason, new hires are deemed to occur in these sectors when workers:

• (a) are hired by at least one new employer in these sectors in year $t$;
• (b) did not hold any job that belonged to the same 3-digit industry in year $t-1$;
• (c) still hold at least one job in the same 3-digit industry in year $t+1$.

Likewise, layoffs are deemed to occur in these sectors when workers:

• (a) were laid off from at least one employer in these sectors in year $t$;
• (b) did not work, in year $t+1$, in any job that belonged to the 3-digit industry associated with their layoff.

## 7. Conclusion

In response to strong demand for local labour market information, the Social Analysis and Modelling Division has recently constructed an administrative data set that is a subset of the Canadian Employer–Employee Dynamics Database data, covers virtually all tax filers and allows the computation of several labour market indicators at the level of individuals’ economic region of residence.

Taken together, the evidence presented in this article indicates that these data are well-suited for the computation of hiring rates and layoff rates. In general, these indicators display plausible temporal movements, plausible cross-provincial variation, and reasonable correlations with conceptually comparable indicators from alternative data sources.

## References

Camu, P., E.P. Weeks, and Z.W. Sametz. 1964. Economic Geography of Canada: With an Introduction to a 68-region System. Toronto: MacMillan of Canada.

Messacar, D. 2014. Report on the Comparison of the T1 Personal Master File and Historical Personal Master File. Ottawa: Social Analysis Division, Statistics Canada. Unpublished.

Morissette, R., and T. Qiu. 2012. Worker Flows in the Longitudinal Worker File and Other Canadian Data Sets. Ottawa: Social Analysis Division, Statistics Canada. 10 April. Mimeo.

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. November.

Morissette, R., Y. Lu and T. Qiu. 2013. Worker reallocation in Canada. Analytical Studies Branch Research Paper Series, no. 348. Statistics Canada Catalogue no. 11F0019M. Ottawa: Statistics Canada.

OECD (Organization for Economic Cooperation and Development). 2009. "How Do Industry, Firm and Worker Characteristics Shape Job and Worker Flows?" In OECD Employment Outlook 2009, Chapter 2, p. 117−163. Paris: OECD Publishing.

Rollin, A.M. 2014. “Developing a Longitudinal Structure for the National Accounts Longitudinal Microdata File (NALMF).” In Producing Reliable Estimates from Imperfect Frames: Proceedings: Statistics Canada’s International Methodological Symposium, October 16–18, 2013. p. 306–311. Statistics Canada Catalogue no. 11-522-X. Ottawa: Statistics Canada.

Is something not working? Is there information outdated? Can't find what you're looking for?