Economic Insights
The Effect of Labour Demand on Regional Demographics

View the most recent version.

Archived Content

Information identified as archived is provided for reference, research or recordkeeping purposes. It is not subject to the Government of Canada Web Standards and has not been altered or updated since it was archived. Please "contact us" to request a format other than those available.

Archived

This page has been archived on the Web.

by René Morissette
Social Analysis and Modelling Division

Skip to text

Text begins

Introduction

When the demand for labour falls within a local region, what happens to the size of the working-age population and the demographic dependency ratio in that area? Simple models of labour supply and demand suggest that populations will decline as working-age individuals leave to find employment elsewhere, and that the demographic dependency ratio (the number of youth and seniors divided by the number of individuals aged 15 to 64) will rise. An increase in labour demand is expected to have the opposite effects—drawing in working-age individuals and their families. While this theory provides guidance on the expected direction of change, the magnitude of such a change is not well understood.^{Note 1} If regional labour demand decreases by, for example, 5.0%, how large a decline in the working-age population and how significant an increase in the demographic dependency ratio might be observed? The goal of this paper is to answer this question.

Using administrative data, the study quantifies demographic changes observed in 76 economic regions that were characterized by different trends in labour demand from 2001 to 2015. The study takes advantage of the substantial differences in employment growth across economic regions. For example, from 2001 to 2008, the 8 economic regions in Alberta experienced paid employment growth that averaged about 19.0%—more than three times the amount registered in the economic regions of Quebec and Ontario (Chart 1).^{Note 2} From 2008 to 2015, the 14 economic regions in Newfoundland and Labrador, Nova Scotia and New Brunswick saw paid employment decline by 4.0% or more, while the 24 economic regions in Manitoba, Saskatchewan and Alberta posted gains averaging 2.0% or more. The study uses this spatial variation in paid employment growth to measure the parameters of interest.

Employment growth might be driven not only by changes in regional labour demand, but also by changes in regional labour supply. For example, if worker preferences for amenities change in a way that shifts labour supply towards regions with high job vacancy rates, employment in these regions will increase for reasons unrelated to labour demand. The challenge is to extract changes in labour demand from the observed employment growth.

To do so, this study uses variation in regional labour demand induced by national changes in the distribution of employment by industry. The underlying idea is simple: if, for example, factors external to a given region cause a decline in manufacturing employment nationwide, regions where employment was heavily concentrated in manufacturing at the beginning of the reference period should fare worse than other regions on various indicators. The regional employment growth that would have occurred if industries in a given region had grown at the same pace they grew nationally is used to extract information about regional changes in labour demand. Using this empirical strategy, the study quantifies the impact of changes in local labour demand on the size of the working-age population and the dependency ratio within economic regions (see the “Data and methods” section in the annex).^{Note 3} This informs discussions about labour mobility, population aging and the functioning of local labour markets.

Descriptive evidence

From 2001 to 2015, the population aged 15 to 64 grew at markedly different rates across economic regions. As a result of the oil boom of the 2000s, the number of individuals in this age group increased by 38.0% or more in the economic regions of Calgary, Edmonton, Red Deer and Wood Buffalo–Cold Lake (Table 1-2). In contrast, the population aged 15 to 64 fell by 10.0% or more in the following economic regions: South Coast–Burin Peninsula, West Coast–Northern Peninsula–Labrador, Notre Dame–Central Bonavista Bay, Cape Breton, Southern Nova Scotia, Campbellton–Miramichi, Edmunston–Woodstock, Gaspésie–Îles-de-la-Madeleine, Côte-Nord, Parklands, Cariboo, North Coast and Nechako (Tables 1-1 and 1-2).

The demographic dependency ratio also evolved differently, falling by about 2 to 10 percentage points in all economic regions in Saskatchewan but increasing in virtually all economic regions in Atlantic Canada, Quebec, and British Columbia.

Economic regions with strong growth in paid employment generally experienced higher-than-average population growth (Chart 2). For example, Wood Buffalo–Cold Lake saw its paid employment and its population aged 15 to 64  rise by about 38.0% and 49.0%, respectively, from 2001 to 2015 (Table 1-2), while about  twenty economic regions experienced declines on both measures. Across all 76 economic regions, paid employment and the population aged 15 to 64 grew by 7.9% and 7.3%, respectively, on average.^{Note 4}

Since strong population growth is positively correlated with growth in paid employment (Chart 2) and negatively correlated with changes in the demographic dependency ratio (Chart 3), economic regions that had strong growth in paid employment experienced smaller increases (or larger declines) in their demographic dependency ratio (Chart 4).

While Charts 2 to 4 suggest that employment growth tends to increase the population and decrease the dependency ratio of a given region, they are subject to two limitations. First, these charts do not distinguish the degree to which employment growth at the regional level is driven by increases in labour demand, rather than increases in labour supply. Second, they display only bivariate relationships and thus do not control for other confounding factors. To overcome these limitations, multivariate analyses are required.

Multivariate analyses

The extent to which changes in labour demand affected the population aged 15 to 64 in regions from 2001 to 2015 is addressed in Table 2. Results are shown using ordinary least squares (OLS) and instrumental variable (IV) estimators. Unlike the OLS, the IV methods do not confound the effects of labour demand and labour supply when measuring the impact of changes in labour demand on socio-economic outcomes.^{Note 5} For this reason, the preferred multivariate analyses and the ensuing discussion are based on IV methods.

Two sets of regressions are considered regardless of the estimator. In the first set, changes in the logarithmic value of the population aged 15 to 64, measured over two seven-year periods (2001 to 2008 and 2008 to 2015), are regressed on a binary indicator for the 2008-to-2015 period and on changes in the logarithmic value of regional paid employment (as measured by the number of employees aged 15 to 64, estimated from the Canadian Employer–Employee Dynamics Database). Province indicators are added in the second set of regressions.^{Note 6Note 7}

Whether province indicators are included or not, the elasticity of the population aged 15 to 64 with respect to labour demand—based on the IV estimator—varies between 0.9 and 1.2. This suggests that, on average, a 5.0% drop in regional labour demand reduces the population aged 15 to 64 by between 4.5% (5.0% times 0.9) and 6.0% in a given region over a seven-year period.

Since younger workers tend to be more mobile than older workers, they would be expected to leave economically declining regions or migrate to expanding regions at a faster pace than older workers. Table 2 confirms this hypothesis. It shows that for individuals under 35, the elasticity of population with respect to changes in labour demand varies between 1.4 and 1.5. The corresponding elasticity for individuals aged 35 to 64 varies between 0.3 and 1.0. Thus, while a 5.0% decline in labour demand tends to reduce the regional youth population by between 7.0% and 7.5%, the same decline in labour demand will reduce the population of older workers by between 1.5% and 5.0%.^{Note 8}

Declines in regional labour demand reduce the working-age population, suggesting that such declines will also increase the regional demographic dependency ratio.

Table 3 confirms this hypothesis. Results from IV methods indicate that a 5.0% drop in regional labour demand will increase the demographic dependency ratio by between 1.1 percentage points (i.e., 5.0% times -0.22) and 1.5 percentage points (i.e., 5.0% times -0.29) from a baseline rate of 49.0% in 2001.^{Note 9Note 10} A 5.0% increase in regional labour demand is expected to decrease the demographic dependency ratio by the same amount. Table 3 also shows that most of the change in the demographic dependency ratio comes from changes in the ratio of the number of individuals over 64 years of age divided by the population aged 15 to 64, rather than from changes in the ratio of the number of children divided by the population aged 15 to 64.

Conclusion

Although declines in regional labour demand are expected to reduce the working-age population and increase the dependency ratio in a given region, the magnitude of these effects had not yet been estimated in Canada.

Using administrative data and Statistics Canada’s population estimates, this study fills that gap. It shows that, over periods of seven years, a 5.0% decline in regional labour demand reduced the regional population aged 15 to 64 by 4.5% to 6.0%. Because working-age individuals are leaving economically declining regions, a 5.0% decline in labour demand raises the demographic dependency ratio by between 1.1 and 1.5 percentage points, from a baseline rate of roughly 50.0%.^{Note 11} Conversely, increases in labour demand—such as those witnessed in many economic regions of Alberta and Saskatchewan during the 2000s—tend to increase the working-age population and decrease the demographic dependency ratio.

In a context where population aging will pose a number of challenges, these results highlight the key role that employment growth may play to alter the demographics of regions.

Annex: Data and methods

This study combines data from the Canadian Employer–Employee Dynamics Database (CEEDD) and population estimates from the Demography Division to produce the estimates shown in Tables 1 to 3. The three following data sets from CEEDD are used: the T1 personal master file (T1PMF), the T4 Statement of Remuneration Paid (T4) and the Longitudinal Employment Analysis Program (LEAP). The 100% versions of T1PMF and T4 are used.

This paper examines changes in the working-age population and demographic dependency ratios across 76 economic regions (including Yukon, the Northwest Territories and Nunavut) over two periods of seven years (2001 to 2008 and 2008 to 2015).^{Note 12}

Population estimates and demographic dependency ratios are obtained from the Demography Division of Statistics Canada. Three demographic dependency ratios are used. The first ratio is the number of individuals under 15 or over 64 years of age divided by the number of individuals aged 15 to 64. The second is the number of individuals over 64 divided by the number of individuals aged 15 to 64. The third is the number of individuals under 15 divided by the number of individuals aged 15 to 64.

To assess the impact of labour demand on population size, a two-step procedure is used. First, changes in the logarithmic value of log population size are constructed for 20 age–sex cells for each of the 76 economic regions.^{Note 13} Denoting age groups, sex and regions by $a$, $s$ and $r$, respectively, these changes ($\Delta Y_{asr}$ ) are regressed on a full set of age–sex interactions (${\theta }_{as}$), as well as on a vector of regional fixed effects (${\theta }_r$):

Equation (1) is estimated separately for each of the seven-year periods defined above.^{Note 14} The parameter estimates for ${\theta }_r$ measure changes in log population size that are standardized for regional differences in the composition of the population by age and sex. In a second step, these parameter estimates are used to form the dependent variable in the following equation:

where ${\alpha }_p$ is a vector of province and territory indicators,^{Note 15} ${\alpha }_t$ is a binary indicator for the period from 2008 to 2015 (2001 to 2008 is omitted), and $\Delta \ln E_{rt}$ measures changes in regional log paid employment.^{Note 16} Since there are 76 economic regions and two seven-year periods, Equation (2) is initially estimated using the ordinary least squares (OLS) estimator on 152 observations. The parameter $\beta$ measures the elasticity of population size with respect to regional labour demand.

To account for the possibility that $\Delta \ln E_{rt}$ might be correlated with the error term $u_{rt}$, Equation (2) is also estimated using the instrumental variable (IV) estimator. When doing so, variable $IV{1}_{rt}$ is used as an instrumental variable for $\Delta \ln E_{rt}$ :

where $\Delta \ln E_{rt}$ measures employment growth in industry $i$ nationwide during the seven-year period $t$ $(t=2001-2008,2008-2015)$ and $Shar{e}_{ir0}$ measures the share of industry $i$ in total paid employment of region $r$ at the beginning of the seven-year observation period that is considered.  $IV{1}_{rt}$ measures the predicted growth in paid employment that would occur in region $r$ if each industry had grown at the same pace regionally as it had grown nationally.^{Note 17} A second instrumental variable, $IV{2}_{rt}$, is also used. When $IV{2}_{rt}$ is used, employment growth in industry $i$ nationwide excludes employment growth in that industry in the economic region $r$ that is being considered.

In Tables 2 and 3, columns “Model 1” and “Model 2” report instrumental variable results based on $IV{1}_{rt}$ and $IV{2}_{rt}$, respectively.

To analyze regional changes (measured in percentage points) in various demographic dependency ratios, $\Delta D{R}_{rt}$, the following equation is estimated with the OLS estimator and the IV estimator:

where $SHARE\_08\_14_{r0}$ ($SHARE\_58\_64_{r0}$) equals the share of the population in region $r$ aged 8 to 14 or 58 to 64 at the beginning of a seven-year period, ${\theta }_p$ is a vector of province and territory indicators, and ${\theta }_t$ is a binary indicator for the period from 2008 to 2015.

Because economic regions are the unit of analysis in this study and because sampling variability issues do not arise when using the 100% versions of T1PMF and T4, Equations (1) to (4) are unweighted (i.e., estimated without population weights). In all cases, standard errors are clustered by economic region.

References

Bound, J., and H.J. Holzer. 2000. “Demand shifts, population adjustments, and labor market outcomes during the 1980s.” Journal of Labor Economics 18 (1): 20–54.

Marchand, J. 2012. “Local labor market impacts of energy boom-bust-boom in Western Canada.” Journal of Urban Economics 71 (1): 165–174.