Overview
This staff analytical note builds on Ens et al. (2021); Ens et al. (2022); and Ens, See and Luu (2023) to assess the health of the Canadian labour market. These earlier works established a more granular framework for assessing the labour market given its diverse and segmented nature. In this note we do three things:
- We update the range of benchmarks in our dashboard of indicators, to ensure they remain relevant. We add an additional year of data, adjust for population aging and improve our filter estimates. We also estimate a series of Phillips curve equations using each indicator of the dashboard to understand how much inflationary pressure is coming from the labour market.
- We do a deep dive to better understand structural changes in the Canadian labour market that could cloud assessments of labour slack. This includes a closer look at the muted recoveries in self-employment and low-wage jobs.
- And finally, we discuss whether the labour market has moved into balance and how much it is contributing to inflation.
We find that the labour market has likely entered into balance from overheated levels, and we also see some signs of structural changes:
- Overall, the labour market appears to have moved into balance. Notably, we see a large reduction in job openings and other demand indicators but no significant rise in layoffs. Moreover, Phillips curve estimates suggest that the labour market was not a large source of inflationary pressure in the first quarter of 2024.
- Self-employment and low-wage employment appear to be permanently lower than in the past. For low-wage workers, this reflects lower levels of demand coming out of the COVID-19 pandemic, though impacted workers were largely able to reallocate to other sectors given the tight labour market following the reopening of the economy. Workers who are self-employed due to weak job prospects also make up a smaller share of the labour market than in the past, and this appears to be a long-term trend. As a result, weakness in indicators of self-employment and low-wage employment in the dashboard is less likely to represent labour market slack.
Updates to the benchmark ranges
To construct the benchmark ranges, we use estimates that each convey valuable information on the state of the labour market. Each range includes the Hodrick-Prescott (HP) filter (Hodrick and Prescott 1997), a modified Hamilton filter (Hamilton 2018; Quast and Wolters 2022), the most recent period the labour input gap was closed (Ens et al. 2022) and model-based estimates used by Bank of Canada staff for potential output (Devakos et al. 2024). We bound these ranges by the historical minimum and maximum since 2003. In this section, we outline two key updates for 2024.
Vacancy measures
Before presenting the new benchmark ranges, we first re-evaluate two measures related to job vacancies that are included in measures of overall labour market conditions. These are the vacancy–unemployment ratio (tightness) and the ratio of vacant jobs to the sum of vacant and filled jobs (vacancy rate).
Relative to other indicators in the dashboard, both measures suffer from short samples, with the earliest published data on vacancies beginning only in April 2015. At the start of the sample until before the pandemic, both measures were increasing. This was likely a continuation of the recovery from the 2008–09 global financial crisis. After a short but steep decline during the pandemic, both measures once again rose sharply during the economic recovery.
Together with the short sample, the dynamics of vacancies during both recoveries have resulted in a particularly strong trend estimate from the HP filter—implying extremely wide benchmark ranges (where the HP filter constitutes the upper bound). This can be seen in the wide ranges (in green) for both measures in Chart 1. While market tightness may indeed have been on a long-run upward trend, the short sample makes HP filter estimates difficult to take at face value at this juncture. As a result, we temporarily exclude this filter for the benchmark ranges of the vacancy measures.
Chart 1: Benchmark ranges have been updated to reflect new data and re-evaluations of vacancy measures
Additional year of data
We incorporate data released since the last update of benchmarks in the first quarter of 2023. The filters and trend estimates have started to pick up some of the softening of the labour market over 2023, resulting in a slight leftward shift of the range of benchmarks. Chart 1 compares the new 2024 benchmarks (in blue) with their equivalent benchmarks using data as of the last reassessment in Ens, See and Luu (2023) (in green) for the indicators that pertain to overall labour market conditions. Chart A-1 and Chart A-2 in the Appendix display the new benchmarks for indicators that fall under the categories of job characteristics and labour market inclusiveness, respectively.
Are structural changes in the Canadian labour market clouding the assessment of slack?
Self-employment in a structural decline
One indicator that has deviated in the dashboard is self-employment. It has been consistently below its benchmark range even when most other indicators have shown signs of excess demand. While self-employment is often linked to entrepreneurship, innovation and employment creation, it is sometimes an alternative to wage employment when job opportunities are scarce (Poschke 2023).
To understand what the drop in self-employment implies about the balance of the labour market, we show the evolution of self-employment according to Statistic Canada’s Labour Force Survey (see Chart 2). The blue line depicts a fall in the total number of self-employed workers as a percentage of employed workers over the past decade and a half. During the post-pandemic economic recovery, the downward trend has become more pronounced, following a brief peak of self-employment and an initial surge in unemployment at the onset of the pandemic.
Chart 2 also shows an important distinction between two types of self-employed workers: those who are entrepreneurs (incorporated), who are more educated and who are more likely to have employees (yellow line) versus those who are not entrepreneurs (unincorporated), who tend to have less education and who are less likely to have employees (green line) (Levine and Rubinstein 2017).1 We refer to the latter as the marginal self-employed.
We can see that both the recent and the long-run drop in self-employment seem driven by the marginal self-employed workers. These are workers who tend to be on the fence between self-employment and other types of employment. Similarly, these marginal self-employed workers seem to explain most of the peak at the onset of the pandemic, as workers may have used self-employment to cope with a loss of other types of employment. The structural decline in self-employment, therefore, may represent a long-term trend of better employment prospects for marginal workers and a compositional shift toward incorporated self-employment.
Chart 2: Structural decline in self-employment has been driven by marginal self-employed workers
Chart 3 drills down to the occupation level. The drop in self-employment seems rather broad-based. For example, it is present in both managerial and non-managerial occupations (panel a), and, since the start of the recovery post-pandemic, in low wage occupations as well (panel b).
Chart 3: The decline in marginal self-employment is broad-based
Chart 3: The decline in marginal self-employment is broad-based
Self-employed workers, as a percentage of employed workers
Sources: Statistics Canada and Bank of Canada calculations
Last observation: February 2024
The structural decline in self-employment is paired with the observation that total employment did not fall at the same time: Chart A-3 in the Appendix shows that the employment rate has been rather stable in the past one and a half decades. This suggests that self-employed workers are likely reallocating to other forms of employment. Hence, the drop in self-employment observed in the dashboard does not mean there is more slack but instead points to a change in composition of employment away from self-employment. The acceleration of this trend in recent years explains why the current observation is outside the benchmark range. The filtering process to obtain this range takes time to incorporate structural changes.
Low-wage occupations have not recovered after the economic shock of COVID-19
Another indicator that has deviated from the rest of the dashboard is employment in low-wage occupations.2 For most of the post-pandemic recovery, the employment rate in low-wage occupations has consistently been at the bottom of its benchmark range. This is in contrast to the path of employment in other occupations (as seen in Chart A-2, panel c, in the Appendix).
To understand what this observation conveys about labour market slack, we use Chart 4 to plot the percent change in employment in low-wage and other occupations relative to the same month in 2019. The blue line shows that overall employment has risen roughly 10% above its 2019 level because of increased population growth and tight labour markets. However, the aggregate measure masks a pattern of reallocation induced by the pandemic—during which low-wage occupations were hit the hardest.
After the significant decline in employment during the COVID-19 recession, employment in low-wage occupations recovered partially but has since remained below pre-pandemic levels. In contrast, employment in other occupations has outpaced average employment growth. This suggests that excess tightness in the labour market during the post-pandemic economic recovery may have enabled workers in low-wage occupations to reallocate to other occupations.
Chart 4: Employment in low-wage occupations is still below pre-pandemic levels
To understand whether the decline in employment in low-wage occupations reflects reallocation of workers to other occupations, we observe the vacancies posted and unemployment rates for both occupation groups. Chart 5 shows that the share of vacancies posted for jobs categorized as low-wage occupations has declined slightly since the beginning of the economic recovery in 2021, in line with lower employment counts in these occupations. This rules out that the change in employment composition has led to excess demand or an outsized increase in unfilled vacancies in these occupations.
Chart 6 shows that the unemployment rates in both occupation groups (low-wage and others) have evolved in a similar fashion, despite the disparities in their employment rate paths. The data suggest that firms have shifted demand away from these occupations at the same time that workers have reallocated to other occupations, such that firm vacancy posting and worker job search behaviour balance out.3
Overall, the weakness in low-wage employment relative to other labour dashboard indicators does not constitute strong evidence of current slack in the labour market. Instead, it is a symptom of structural changes in the occupational distribution that may have been accelerated or induced by the pandemic. As more data arrive, the filters embedded in the benchmarks should eventually account for this.
Chart 5: The share of vacancies in low-wage occupations has declined slightly
Chart 6: Unemployment rates in both occupation groups have evolved in a similar fashion
Is the labour market adding to inflation pressures?
Excess demand has eased
To help determine whether the labour market is adding to inflationary pressures, we first look for any remaining excess demand in the labour market following the post-pandemic economic recovery (see analysis in Ens, See and Luu 2023). As part of this exercise, we narrow down the list of dashboard indicators to those that most closely resemble pure demand measures (Chart 7). These include measures of unfilled job openings and labour shortages. We also include the involuntary part-time rate, which is a signal that people want to work more hours but there is insufficient demand. Finally, we include the job separation rate, at which employed workers become unemployed. This rate is often associated with involuntary layoffs.
Chart 7: Labour demand indicators have eased
Chart 7 shows that all these indicators have eased to varying degrees, which can be seen in the difference between their recent peaks (in green) and their current levels (in yellow). The declines are most pronounced in job vacancies and survey assessments of shortages. These measures are now all within or below their benchmark ranges. In contrast, the job separation rate eased much more modestly and is sitting just above its benchmark range. The overall picture—one of easing job openings but muted layoffs—is consistent with a gradual rebalancing of labour market conditions.
Wage pressures have moderated, suggesting a balanced labour market
As the “price” of labour, wage measures are one way to assess the balance between supply and demand. While wage measures have lagged behind the easing in other labour market measures over 2023, more recently they have been showing signs of moderation. Differences exist across measures, but for the most part they are now signalling wage growth below 4%. Moreover, around 1 percentage point (pp) of the current strength in wages may reflect past high inflation rather than labour market strength (Bounajm et al. forthcoming). With the marked slowdown in inflation since early 2023, wage measures should see further easing.
Chart 8: Measures of wages have been showing signs of moderation
One indicator that stands out among wage measures is unit labour costs (ULCs). ULCs can be important for inflation as they directly represent the costs that firms face when producing. ULCs are influenced by both productivity—the amount of output produced per unit of labour—and wage growth. Given the easing in wages, ULCs should see some relief. However, weak or negative productivity growth could keep these costs elevated, creating a risk of a more persistent source of pressure on wages.
Labour strength is consistent with its level before the pandemic and close to fundamentals
Despite the fact that excess demand appears to have eased, some uncertainty remains as to whether residual strength in the labour market is still adding to inflationary pressures. One way to assess this is to compare where indicators stand relative to their benchmark ranges in aggregate and over time.
Chart 9 plots the average of the labour market indicator gaps relative to their benchmark midpoint values. In 2022, coming out of the economic shock of the pandemic, the gap was on average more than one standard deviation above its benchmark values. This is considerably above all other time periods during the last 20 years. The easing in labour market conditions since early 2023 is also evident in the data, with the most recent value coming close to pre-pandemic values and in the vicinity of zero. Overall, this picture is consistent with a labour market that has eased but still has some strength.
Chart 9: Labour market conditions have eased
Phillips curve estimates suggest a modest contribution to inflation pressures
To understand whether the remaining strength in the labour market is still generating inflationary pressures, it is important to tease out the impact of different labour market variables, as this is not uniform. One way to do this is by estimating a series of Phillips curve equations using each indicator of the dashboard as the explanatory variable.4 We sort the coefficients of the dashboard indicators on the Phillips curve estimates according to their magnitude (and statistical significance) to learn which factors are more likely to impact inflation.
The results, summarized in Table 1, hint at a clear hierarchy in the strength of the association between different types of indicators and inflation. The strongest link corresponds to unemployment rates, job changing rates and job finding rates. Wages are in the middle of the table with a smaller but significant contribution, in line with the premise that wage growth translates into inflation. Among the variables with the weakest association are participation rates.
Table 1: Some indicators contribute more to inflationary pressures than others
| Indicator | Total impact | Significance of estimates | Model fit |
|---|---|---|---|
| Unemployment rate, 15–24 males | 0.43 | ||
| Job changing rate | 0.41 | ||
| Job finding rate | 0.40 | ||
| Unemployment rate | 0.39 | ||
| Unemployment rate, <27 weeks | 0.39 | ||
| Involuntary part-time rate | 0.37 | ||
| Unemployment rate, 55+ males | 0.37 | ||
| Unemployment rate, non-university | 0.37 | ||
| Wage growth, SEPH variable weight | 0.36 | ||
| Employment rate, low-wage | 0.36 | ||
| Broad unemployment rate | 0.35 | ||
| Unemployment rate, university-educated | 0.35 | ||
| Unemployment rate, 52+ weeks | 0.35 | ||
| Unemployment rate, 25–54 males | 0.35 | ||
| Unemployment rate, 27+ weeks | 0.35 | ||
| Unemployment rate, 25–54 females | 0.34 | ||
| Participation rate, 15–24 males | 0.34 | ||
| Employment rate, private sector | 0.33 | ||
| Employment rate | 0.32 | ||
| Unemployment rate, 55+ females | 0.31 | ||
| Wage growth, SEPH fixed weight | 0.30 | ||
| BOS labour shortages | 0.30 | ||
| Employment rate, mid-/high-wage | 0.29 | ||
| Unemployment rate, 15–24 females | 0.29 | ||
| Job separation rate | 0.29 | ||
| Wage growth, LFS fixed weight | 0.27 | ||
| Wage growth, LFS variable weight | 0.27 | ||
| Wage-common growth | 0.27 | ||
| Wage growth, NAC | 0.26 | ||
| Labour force participation rate | 0.23 | ||
| Participation rate, 15–24 females | 0.22 | ||
| Unit labour cost growth | 0.22 | ||
| Employment rate, public sector | 0.19 | ||
| Average hours worked | 0.17 | ||
| Participation rate, non-university | 0.16 | ||
| Participation rate, 25–54 males | 0.06 | ||
| Participation rate, 25–54 females | 0.01 | ||
| Participation rate, 55+ females | -0.11 | ||
| Employment rate, self-employed | -0.16 | ||
| Participation rate, university-educated | -0.29 | ||
| Participation rate, 55+ males | -0.32 |
Note: For significance of estimates, a check mark denotes the 5% level, an exclamation mark denotes the 10% level, and a X denotes not statistically significant. For model fit, a full star means an R2 above 0.28, half a star means it is between 0.17 and 0.28, and an empty star means it is below 0.17. SEPH is the Survey of Employment, Payrolls and Hours. BOS is the Business Outlook Survey. LFS is the Labour Force Survey. NAC is the National Income and Expenditure Accounts.
Sources: Statistics Canada, Bank of Canada Business Outlook Survey and calculations.
The relative softness in the indicators that seem to fuel inflation the most (namely, job changing and job finding rates), together with the signs of slack coming from rising unemployment and involuntary part-time rates, indicate a labour market that is not likely exerting inflationary pressures. A similar message can be drawn from looking at multiple measures of wage growth that now sit below or near their benchmark ranges. Also, the estimated impacts of wages do not seem high at present, and most of the signals of labour market indicators are aligned with the aggregate output gap.
Finally, the indicators that most strongly convey remaining strength (e.g., participation and employment rates) seem quite dissociated from inflation according to our Phillips curve estimates. This result is somewhat intuitive given that the indicators partly reflect the available supply of workers.
As a summary measure of these associations, Chart 10 shows the overall contribution of labour market indicators. We sum the indicators’ deviation from the benchmark range midpoint, weighting each of them by their coefficients on the Phillips curve.5 We see that the average contribution has substantially dropped—from a peak of 0.5 pps by mid-2023 to 0.3 pps by the first quarter in 2024.
Chart 10: The contribution of the labour market to above-target inflation was moderate
Appendix
Chart A-1: Measures of job characteristics
Chart A-2: Measures of labour market inclusiveness
Chart A-2: Measures of labour market inclusiveness
Note: This chart presents the current value of labour market indicators when compared with their historical worst and historical best. Benchmarks are composed of the modified Hamilton filter, the Hodrick-Prescott filter, the corresponding value of the indicator during a period when the labour input gap was closed and, for selected indicators, trend estimates produced by the Bank of Canada. Employment levels by wage are not seasonally adjusted.
Sources: Statistics Canada and Bank of Canada calculations
Last observation: February 2024
Chart A-3: Employment rate has been roughly stable in the last two decades
References
Bounajm, F., J. G. Junior Roc and Y. Zhang. Forthcoming. “Sources of Pandemic-Era Inflation in Canada: An Application of Bernanke and Blanchard.” Bank of Canada Staff Analytical Note.
Devakos, T., C. Hajzler, S. Houle, C. Johnston, A. Poulin-Moore, R. Rautu and T. Taskin. “Potential Output in Canada: 2024 Assessment.” Bank of Canada Staff Analytical Note No. 2024-11.
Ens, E., L. Savoie-Chabot, K. See and S. L. Wee. 2021. “Assessing Labour Market Slack for Monetary Policy.” Bank of Canada Staff Discussion Paper No. 2021-15.
Ens, E., C. Luu, K. See and S. L. Wee. 2022. “Benchmarks for Assessing Labour Market Health.” Bank of Canada Staff Analytical Note No. 2022-2.
Ens, E., K. See and C. Luu. 2023. “Benchmarks for Assessing Labour Market Health: 2023 Update.” Bank of Canada Staff Analytical Note No. 2023-7.
Hamilton, J. D. 2018. “Why You Should Never Use the Hodrick-Prescott Filter.” The Review of Economics and Statistics 100 (5): 831–843.
Hodrick, R. J. and E. C. Prescott. 1997. “Postwar US Business Cycles: An Empirical Investigation.” Journal of Money, Credit and Banking 29 (1): 1–16.
Levine, R. and Y. Rubinstein. 2017. “Smart and Illicit: Who Becomes an Entrepreneur and Do They Earn More?” The Quarterly Journal of Economics 132 (2): 963–1018.
Poschke, M. 2023. “Wage Employment, Unemployment and Self-Employment across Countries.” IZA Discussion Paper No. 16271.
Quast, J. and M. H. Wolters. 2022. “Reliable Real-Time Output Gap Estimates Based on a Modified Hamilton Filter.” Journal of Business & Economic Statistics 40 (1): 152–168.
Endnotes
- 1. Self-employed can be solo workers (freelancers, gig workers, etc.) or business owners. In the case of business owners, they may or may not hire employees, depending on the scale of their operations.[←]
- 2. Low-wage occupations are those whose median wage in 2019 was less than $16.03 per hour (two-thirds of the 2019 annual median wage of $24.04 per hour). Occupations are based on two-digit National Occupational Classifications. Other occupations are the rest of the occupations not considered low wage.[←]
- 3. The incomplete recovery in employment is also reflected in the lower output share in the accommodation and food services industry, which is more likely to employ workers in low-wage jobs. See Statistics Canada Table 36-10-0434-01, "Gross domestic product (GDP) at basic prices, by industry, monthly (x 1,000,000)" (2024).[←]
- 4. The dependent variable is the deviation of quarterly annualized growth in the CPI-trim from the 2% target, and we include four lags of each dashboard indicator at a time as the explanatory variable. We include commodity prices and CAD/USD exchange rate changes (four-quarter moving averages) as controls. Explanatory variables are in deviations from the mid-point of their benchmark ranges. We calculate the total impact on inflation of each explanatory variable as the sum of coefficients on all lags, multiplied by its deviations. The equations are estimated over a sample from the first quarter of 1997 to the fourth quarter of 2019.[←]
- 5. Indicators negatively associated with inflation receive a zero weight, as their contribution falls outside the range of meaningful relationships.[←]
Acknowledgements
We thank Mikael Khan, Marc-André Gosselin and Alexander Ueberfeldt for helpful comments on earlier drafts of this paper. We also thank Colette Stoeber and Maren Hansen for their editorial suggestions and Andrée Charbonneau and Marie-Lou Lachance for translating. Finally, we would also like to thank Anna Shatalova and Ethan McTavish for their capable research assistance. Any errors are solely the responsibility of the authors.
Disclaimer
Bank of Canada staff analytical notes are short articles that focus on topical issues relevant to the current economic and financial context, produced independently from the Bank’s Governing Council. This work may support or challenge prevailing policy orthodoxy. Therefore, the views expressed in this note are solely those of the authors and may differ from official Bank of Canada views. No responsibility for them should be attributed to the Bank.
DOI: https://doi.org/10.34989/san-2024-8