Report Shows Job Growth Highest in Low Wage Industries.


Leading pollsters have indicated that the economy has become the number one issue during the election campaign of 2015. It is not surprising, given the economy has recently entered its 2nd recession since 2007. Beyond  the headline numbers, the question arises- who can Canadians trust to build a future economy? The Harper government maintains that they  have been great economic managers and Canadians should trust them (campaigning on this logic during a recession is surely a sign of how bad things have gotten.) The following report finds that contrary to these claims the economy in terms of job growth has stagnated and become recession prone. So much so  that some of the fastest growing industries in Canada are actually rooted in industries that help the growing levels of poverty.  For example the 2nd fastest growing industry in Canada since the recession of 2007 is NAICS444 –“Rooming and boarding houses”. Also making it into  the top 30 high growth industries  are “Used Merchandise stores”, “ Community food and housing, and emergency and other relief services”. Other top growth industries are related to the increasingly privatization of the health care sector and elder care sectors.

Quite ironically  there is not one technology or manufacturing related industry in the top 30 growth industries as measured by job creation rates over this period. However there many technology and manufacturing industries in the top 50 declining industries. The conclusion being that  there is  qualitative side to job creation as the actual nature of job growth and decline under Harper’s government is leading Canada down a pathway that could make the current recession more pronounced. In fact looking at the data- it is hard to fashion any plan or direction in job growth- save for privatization of government services and a declining resource extraction, technology and manufacturing industries.

So what exactly has Mr. Harper’s much advertised “action plan” produced for the Canadian economy. The following report, produces a detailed accounting of what industries produced the highest rates of growth and decline since the great recession measured in terms of jobs. For now these jobs will be treated as “any job”- in future the report with segment these into low wage and high wage jobs. What industries have flourished and which industries have fallen off since the great recession of 2008.
The uniqueness and power of this report rests on the facts  and the detail of where Canada’s economy has shown growth and decline. This based upon a very powerful but underutilized data source from Statistics Canada. The report uses detailed industrial classifications from an administrative Payroll Data file of all business entities that is maintained by Statistics Canada- (also known as the Survey of Employment, Payroll and Hours (SEPH). The data is a very powerful administrative data source- meaning it is not subject to sampling errors, and because the data source rides on the back of the payroll file and is classified using the very large and extensive Business Register at Statistics Canada- it has a great ability to encapsulate Industrial Classification and monitor employment and wages. A very powerful but sadly underused data source. The data source does have some gaps, for example the data does not include the self- employed, and has some large groups of unclassified businesses. (however many of these are undoubtedly own account employers or self employed with intentions of hiring or potentially laid off workers)

Part One: What Industries are Growing- A Qualitative view of the Quantitative

As can be seen in the first chart- job growth has been mixed within several sectors- and job declines have been focused in the manufacturing sector. The box plots show the broader NAICS industry by the percentage of job growth and decline. Each dot within the industry box plot represents the number of jobs gained or lost within that sub-industry so it affords a size dimension for each industries absolute job gains or losses. As can be seen in the manufacturing sector many of the jobs lost in this critical sector have been widespread. This suggests that a macro scale dynamic is at work- namely the appreciation of the dollar over this period has produced a widespread downsizing withing the industry. This across the board disadvantage meaning the affect of a high dollar can impact all companies- even those showing highly innovative capabilities. Possibly verifying what many leading economists have been concluding about economic development strategy- macro policy trumps micro. This has some serious implications for firms struggling to use innovation strategies to survive and prosper and could have a large impact on future productivity development. ChInd2

The graph also shows broader increases in public sevices, educa and health as well as trade. The expansion of public services could lead to some future growth in productive capacity and that is a positive outcome. However- looking at the more detailed job growth and decline at the  4 digit industries, one gains a qualitative perspective.

As can be seen-  in some of the top job creation industries it raises concern and signifies just how bad the economy is doing – that is industries have grown fastest growing industries are not in those one would classify as a healthy sustainable economy on many fronts.  For example some of the leading growth industries are within those that are helping people adapt to economic hardship such as “Rooming and Boarding houses”, “Community Food and Housing and Emergency Services”, and “Used Merchandise Stores”. Several

[embeddoc url=”” download=”all” viewer=”microsoft”]


Part 2: Exploring Industry Growth and Decline by Average Weekly Wage


The analysis next moves onto examine what industries have increased or decreased in size based upon the average weekly wage as measured by the SEPH survey vehicle.  The next part of the analyis rates industries in terms of absolute gains or losses in number of jobs.  A plot was performed on the 4 digit NAICS level on the count of job change over the 2008 to 1014 period by the average weekly wage paid in 2014.  A linear regression algorithm was used to explore whether there was a relationship between size of industry job change and wage level. the regression model used a standard least squares algorithm, and was weighted by the absolute value of the size change in jobs.


As can be seen their is a negative linear association between the two variables- which suggests that as the size of jobs created within an industry grows positive the lower the wage rate becomes  and the reverse for higher wages. Note that the graphical interpretation is a bit hidden because of the weighting structure of the industries. So imagine the larger dots in the plot having much more power in minimizing the variance of the estimate when determining the location of the regression line.

Also note that the large industry in the top left corner- i.e. low wage but large increase in jobs is the Full service Restaurant and semi service restaurant industry- which by far created the most jobs.

























Here is the Regression output- as you can see the p value is significant and the r value is .12 which means the relationship is signficant.

lm(formula = sephm$diff07 ~ sephm$X2014.y, weights = (abs(sephm$diff07)))

Weighted Residuals:
Min 1Q Median 3Q Max
-13652457 -1896031 -1035852 -484681 35907579

Estimate Std. Error t value Pr(>|t|)
(Intercept) 61777.67 7438.46 8.305 1.76e-14 ***
sephm$X2014.y -41.05 7.76 -5.290 3.32e-07 ***

Signif. codes: 0 ‘***’ 0.001 ‘**’ 0.01 ‘*’ 0.05 ‘.’ 0.1 ‘ ’ 1

Residual standard error: 3602000 on 192 degrees of freedom
(49 observations deleted due to missingness)
Multiple R-squared: 0.1272, Adjusted R-squared: 0.1227
F-statistic: 27.98 on 1 and 192 DF, p-value: 3.32e-07


Please note that this is in early stages of research- I will be adding wages to each of the industries and quantifying  industry growth by wage classifications. And finally I will be performing some machine learning on the time series over this period on the monthly data to determine what industries had similarity in terms of growth and decay patterns. The routines will use a k-means clustering algorithm for times series data. I will be adding more analysis as time moves forward.

Update: Work has commenced on expanding this analysis. The goal will be to include an analysis of wage growth within each of these high growth industries- to determine if we can quantity and classify the wage level of jobs being created within these 3 digit NAICS industries.

A second goal will be to explore the nature of job growth between industries that have grown versus those that have declined with a special emphasis on time series analysis. Adding into this analysis will be several additional varablies in which the study will use to expand the understanding of the high growth industries and those in decline.

Lastly the outputs from this study will be to generate some user friendly and accessible outputs to aid and develop a more granular understanding of the actually existing industrial growth and decline since the great recession of 2008.

For questions or comments please use the contact us form to send feedback.




Employment Insurance Levels on the Rise- this is more than a “technical recession”!

caneiThe latest GDP numbers have confirmed that Canada has officially entered into its second recession in less than 10 years. Without much but denial from the federal government- many questions remain on the nature and extent of this new recession. How long will this recession last? How will it impact different sectors and regions? How many workers and families will it affect? As with most economic questions – we must look deeper into the data for clues to make such predictions. One such measure is the Employment Insurance (EI) claims data. The Employment Insurance statistics are an administrative data source.  The level of EI claims are a very sensitive indicator on the health of the labour market. The Conference Board of Canada uses EI claims as one of the single inputs of the labour market into their modelling of a  leading indicator index on the economy. The EI data is far from perfect and excludes many of the unemployed especially those in short term unemployment- but it can provide a strong timely indicator of the functioning of the economy. Given it is administrative data, it goes beyond the data quality of unemployment numbers estimated by the Labour Force Survey which are subject to large sampling errors. The data looks at EI claimants rather than those actually receiving benefits.  The waiting  and processing period of EI can delay the statistical outcome of being counted as a person collecting benefits for up to a four months. Therefore focusing on claimants provides a timelier look at the economy.

(Employment Insurance Claimants are defined as those that have filed a claim and are awaiting a decision to determine their eligibility. The waiting period attached to this process can range from 4-8 weeks. Claimants have a high probability of eventually becoming beneficiaries – for more information on this table see the notes for CANSIM table 276-0004)

Analysis of the Employment Insurance data provides the following summary highlights:

1) The number of EI claims have risen 17% in a year over year change from June 2014 to June 2015, and notably 14% over the first 6 months of this year (latest data available is June 2015). With many new more stringent Employment Insurance eligibility requirements- as compared to previous periods- this has undoubtedly biased the number of claimants downwards. So one must take the 17% as an underestimate when comparing to earlier recessionary periods in our recent economic history.


2) A regional breakdown of the EI Claims shows that several provinces have experienced a rise in the number of claimants. This indicates the recession is digging a wider hole in the economy beyond that of the oil sector and Alberta. Year over year change from June of 2014 to June of 2015 in EI claims have risen in Alberta 42.3%, Saskatchewan 12.6% and Ontario 9.2%. (using data from CANSIM table 276-0004 in which are Statistics Canada seasonally adjusted counts). Given the newer rules of EI eligibility, especially those relating to seasonal workers, it will be difficult to fully assess the regional aspects of EI levels as compared over time, as we know some areas have higher concentrations of seasonal workers such as in Eastern provinces, and Northern areas of the country . However a more complete analysis of the data focusing on the level of seasonality of the data from these provinces could provide some evidence. Such a task is beyond the scope of this short article.


3) The third summary point that arises from the data focuses on the trend in EI claimants as compared to past recessions. As mentioned in the last point, the EI data contains a large seasonal component making it difficult assess the raw data. In making the analysis somewhat clearer, the trend or signal in the monthly EI time series was extracted (blue line in graphs). The data reveals that the relative economic impact compared to previous recessions is beyond a technical recession that pundits have labelled. The evidence is quite clear- that so far in this early stage of the recession, at least according to the growth in EI, this current recession is larger than the 2002 recession that was the result of the meltdown. However it is not as great as that witnessed during the Great recession of 2008. As can be seen in the trend line- the acceleration has not changed from its upward trajectory and therefore we are definitely not at the end of this recession. So it is difficult to compare given this recession has just started. The message is fairly obvious from the trend line- this is much more than a technical recession. (The blue trend line or signal was extracted using the raw non-seasonally adjusted data from CANSIM table 276-0004. The algorithm to extract the trend was the STL with LOESS seasonal decomposition method which used localized polynomial regression combined with a moving average function. This algorithm is similar to the ARIMA method- but is less susceptible to outliers. However it can be more difficult to obtain the greater sensitivity of the ARIMA method. Given the task was to merely create a visual display the STL was chosen.)



4) The coverage rate of Employment Insurance Beneficiaries as a proportion of the unemployed is the last measure that was calculated. This measure is quite important in determining the overall effectiveness of the EI program in reaching its functional goals in providing relief to those experiencing job loss. As can be seen in the last chart, the coverage rate has declined substantively from the past levels and reached a low point of 38% in 2011.(calculated using Employment Insurance Regular Beneficiaries over the total unemployment using data from CANSIM tables 276-0040 and 282-0087) This due to the continued dismantling of the program, where now less than 2 of 5 unemployed workers actually qualify for this job loss insurance- a tragic outcome for workers. These lows in EI benefit payouts occurring during the longest economic stagnation and recession prone times in the history of Canada. The coverage rate has increased a small amount in the past year- but this is mainly due to the uptick in unemployment- and not due to any new more worker friendly policy.



















Summary- Employment Insurance claims have shown a dramatic rise in response to the recessionary period that Canada entered in the first half of the 2015. Given the new more stringent eligibility rules for collecting EI benefits, the rise in the number of claimants is under representing the extent of job loss when compared to previous recessionary periods. This is contrary to what many have concluded- that the Canadian economic recession was merely- a “technical recession”. The question that many ask- why has the unemployment rate not spiked in a traditional manner when entering this recession. The historical linkage and loss of protective power in EI benefits may actually be part of the reason. That is, as benefits have been cut back in terms of benefits paid, as well as the much tighter eligibility rules- the lack of insurance benefits forces many who face job loss- to find some income protection in jobs that are low paying, part-time, self-employment and other necessary non-traditional employment transition and workforce adjustment mechanisms. In the short term- such ad hoc work force adjustment may be less costly in terms of short term outlays- however the longer term inefficiencies and social outcomes measured from a skill development, training, and many related social outcomes to job loss much more costly to the economy and society.  Wider labour market measures- seem to suggest that this is the new trend within the process of workforce adjustment and transition mechanisms. The EI claim data also point to a much wider recession across more sectors and regions of the economy- wider than the oil sector, and more decentralized than Alberta as many economists have suggested. Lastly- the data in terms of trend predicts that the job loss and EI claims associated with it, will remain high for at least the next several months.