The picture that will define Ontario politics for the next four years

In Ontario, public sector employees earn more than private sector employees. Many workers in the private sector earn the minimum wage, or only slightly above minimum wage. The peak of the public sector earnings distribution is much higher, at twenty-something dollars per hour, and there are a good number of public sector workers earning $40 or $50 an hour. 

There are many things missing from this picture. Most importantly, it excludes highly-paid self-employed professionals, such as doctors, lawyers, and accountants, as well as entrepreneurs and business owners. It also excludes self-employed people in the trades, such as plumbers, electricians and contractors. The numbers are non-trivial: 13 percent of Ontario workers are self-employed. A good chunk of the upper part of the private sector earnings distribution is missing from the picture.

On the other hand, the hourly wage distribution above excludes non-wage benefits that are more common in the public than the private sector, such as employer contributions to health insurance and pension plans. 

Furthermore, the picture does not take into account the differences in the nature of work in the public and private sector. Many public sector jobs, such as nursing, social work, and teaching, require relatively high levels of skill and education. There are private sector jobs that require skill and education as well - but, as noted earlier, many of those jobs are carried out by self-employed professionals, so are not in the graph. 

To control for these confounding factors I ran a regression analysis, and threw in pretty much every determinant of the wage rate that I could think of - age, immigrant status, union status, education, city of residence, marital status, sex, full-time or part-time work, and family type (i.e. with or without kids in the home). Here's what I found:

Doing some quick calculations (described here), this regression result tells us that private sector workers earn 18.5 percent less than public sector workers, even after controlling for differences in education, and so on.

Ontario has had a Liberal government for 15 years, and there is a widespread sense in the province that it is time for a change. If a Progressive Conservative government is elected, it will likely be looking to cut taxes - or at least not increase them - and at the same time deliver more to its base, for example, more roads for suburban commuters, or more health and home care for the elderly. Yet debt finance is not really an option: the province has a debt of about $311 billion, and a provincial debt/GDP ratio of 37%, and there are good reasons to fear an economic downturn.

Cutting the public sector wage bill is a pretty obvious way for a Conservative government to finance its agenda. For the one in five Ontario workers who are public sector employees - at the municipal or provincial level, working in universities, hospitals, or for the federal government -  this could make for interesting times. 

Note: the analysis here was carried out with the September, 2017, public use microfile of the Labour Force Survey. Here are the Stata commands used:

for the kernel density plot: twoway kdensity HRLYEARN if PROV==35 & COWMAIN==1 [aweight = FINALWT] || kdensity HRLYEARN if PROV==35 & COWMAIN==2 [aweight = FINALWT], title("Hourly wage distribution, employees only") subtitle("Ontario, September, 2017") ytitle("Proportion of employees") note("Calculated by F. Woolley using Labour Force Survey PUMF")

for the regression: regress ln_wage i.COWMAIN i.AGE_12 i.EDUC i.IMMIG i.CMA i.SEX#MARSTAT i.FTPTMAIN i.UNION i.EFAMTYPE#SEX [pweight= FINALWT] if PROV==35 Note: restricting the sample to 25-64 year olds reduces the estimated coefficient to 0.198. Occupation not included because of its collinearity with sector  of employment.

Update: here are a few more pictures in response to issues raised in the comments.  Click for full sized images. The difference in kernel densities might appear slightly smaller if one only looks at unionized employees, but it's hard to say. If one restricts the regression analysis above to unionized employees only, the coefficient on "private sector" actually rises in absolute magnitude, to -0.224, which might be because the restriction to unionized workers only drops high earning private sector executives from the sample.

The public/private sector wage differential is much larger for women than for men. Running the regression above for women only produces a coefficient on "private sector" of -0.308; running it for men only produces a coefficient of -0.154. Here are some pictures.

I was asked about Quebec. Here are some pictures with cross-province comparisons. Click on the links for full-sized pictures. BC and Quebec private sector wages seem slightly less concentrated around the minimum wage than Ontario's. Ontario public sector employees seem to do a bit better than those in other provinces, possibly reflecting the presence of the federal government in Ottawa (or U of T?). Interprovincial differences might also different nature of government-provided services in Quebec (more public sector child care centres?) or the ability of the Quebec government to take advantage of worker's relative lack of interprovincial mobility.

Comments

[Jim Sentance]

Anything in there that would control for the effects of pay equity?

[Nick Rowe]

How big is the coefficient on union status? I *think* it's true that public sector workers are more unionised than private sector? And I can't articulate this clearly, but: if I heard that the public private differential were explained by e.g. age and education, I might say that's a legitimate differential and not a problem; but if I heard that it were due to a differential in the degree of unionisation, I would say that's a problem, and it's one of the mechanisms behind that public/private differential.

[Frances Woolley]

Nick, if we restrict the sample to unionized workers only, we still get a substantial difference between public and private sector workers - see notes below.

The regression analysis above includes controls for age, education, etc. The public/private wage differential remains (though it gets quite a bit smaller) if one controls for occupation.

Jim - good guess. I ran the analysis and did the kernel densities for men and women separately. The public/private differential is much larger for women than for men.

[Lord]

I suppose there is a tenure effect as well with public workers often being older and more experienced which helps keep income up while private workers often lose income with forced changes.

[Jacques René Giguère]

And what if the problem was that private sector employees were earning too little?

[mpledger]

How did age enter into the regression equation? As a factor (what were the levels?) or as a continuous variable?

[Stephen Gordon]

Were you able to extract levels of government from the LFS PUMF? Although it's probably true that the vast majority of Ontario public sector workers are provincial/municipal.

[Frances Woolley]

mpledger - the Labour Force Survey PUMF has a 12-category age variable, so for the most part we're talking about 5-year age intervals. I just stuck it in as a series of categorical variables - so 11 dummy variables for all of the different age categories. I'm not interested in estimating an age-earnings profile, so a series of categorical variables is just fine for soaking up any age-related variation.

Stephen - no, though one could make some guesses by looking at occupation and CMA. All of the federal workers in Ottawa are included in the public sector numbers here. So perhaps conclusion of the post is somewhat misleading.

Jacques Rene - "And what if the problem was that private sector employees were earning too little?"

I largely agree with you that the problem is that private sector employees are earning too little. The size of the peak around the minimum wage is really striking.

I would love to believe political parties that court and win the low-wage private sector vote will then try to improve conditions for low wage private sector workers. But sadly I think it's much more likely that at least one political party will try play the politics of envy: bringing public sector wages down closer to private sector ones, rather than private up to public.

One possible conclusion from this discussion is that the Ontario Liberal government should have raised minimum wages much earlier in their mandate, and more gradually.

[Tim Smyth]

Can Frances or Jacques provide the same type of data in terms of Quebec private sector workers vs Quebec public sector workers for comparison?

[Frances Woolley]

Tim, your wish is my command! Please see the added pictures at the end of the post. Click for full-sized images

Interesting how the right hand tail of the public sector wage distribution is longer in Ontario. I don't think this can just be the federal government, I'm thinking it has to be universities, especially U of T, and hospitals as well.

[Tim Smyth]

One thing to keep in mind with Quebec private sector wages is that several big industries like Aluminum have no effective competition in the rest of North America. So while they might have a "captive" Francophone workforce they also have a captive customer bases.

Also along these lines in addition to Aluminum in Quebec there are also bunch of large plants in QC that do things like Zinc and Titanium refining that many Canadians have never heard of. One of the things that is starting to come about with the aluminum tariff fight in the US is that Quebec in particular has an almost absolute advantage in Aluminum production and the effect of putting tariffs on Aluminum will only discourage the use of aluminum generally in sectors where it is expanding.

Another thing too I forgot is non wage health benefits in Quebec are considered taxable income for QC Income Tax but are tax free federally and in Ontario provincial PIT.

[Tim Smyth]

Also on a Quebec focus I noticed the new Ontario PC leader criticized Quebec for "bailing out" Bombardier but then said that both the Ontario and Federal Governments should have given more tax breaks to a Proctor and Gamble plant closing in Ontario despite the fact Proctor and Gamble was paying it's workers much lower wages than Bombardier. I remember Jacques once commented years ago that the Ontario business community had chosen a "low tech low wage" industrial strategy while Quebec had choosen a high tech high wage strategy. Perhaps we are starting to see evidence of that in these numbers.

[Tim Smyth]

Below is the thread from 2012 where Jacques brought up the idea of a low tech low wage Ontario model.

http://worthwhile.typepad.com/worthwhile_canadian_initi/2012/01/toto-1.html

[Bob Smith]

I'm sort of curious what we think it means when we say that "private sector employees are earning too little", relative to what? Relative to a sector where employees and employer have monopoly power vis-a-vis their respective customers, and the employer has the ability to compel consumers to pay for goods and services (the public sector). Do we think its feasible to replicate the economics of the Ontario public sector in the private sector?

The counterargument, that public sector employees are paid too much, relative to the private sector at least has a credible counter-factual - base private sector compensation on market wages - and practical policy paths to achieve that goal (contracting out, requiring labour arbitrators to respect hard budget constraints, competitive suppliers (school choice), etc.)

Also, I think it's unhelpful to characterize challenging public sector compensation as "the politics of envy", rather I think the right way to think of it is maximizing the bundle of public services we can provide for a given cost - that we treat public services the same way we treat every other good and service we consume. All else being equal, arbitrary increases in public sector compensation (that is to say, compensation above the market rate necessary to attract a given level of talent - perhaps driven by union monopoly power, perhaps driven by the desire to buy off powerful constituencies ahead of an election), either mean you pay more for the same services or, for a given level of tax revenue, that you're going to be getting fewer/worse public services. IF you believe that the role of the public service is to provide services to the public, rather than to provide well-paying jobs to public servants, then that's the right way of thinking about it (to the extent well-paying jobs are necessarily incidental to the provision of public services, great, but they shouldn't be an objective on their own).

[Frances Woolley]

Bob - "relative to what?"one answer: relative to a situation where there is no monopsony power in the labour market.

[Luke I]

I suspect this analysis needs to be done at the occupational level as employers do not pay a spot market wage for generalized labor and education seems far to coarse a control for identifying different skill set (particularly the prized soft skills we hear so much about). Also, if you netted out retail and food and accommodations workers I suspect your leftmost bump in the private sector density would likely look quite different. The groups are not that easily comparable and the interactions between your controls may be quite nonlinear in the public sector (have you tried things like union_status*education_level dummies? - just curious).

[Frances Woolley]

Luke -

I don't know if occupational controls really do it. E.g. the pay of teachers in Ontario reflects the public sector wage structure, it doesn't cause it. Ditto police officers. And if you try to compare police officers and security guards - that's not really a like with like comparison.

I think you're right that retail and food/accommodations drive that spike at the bottom left of the private sector earnings distribution - but that's where the jobs are.

I'm much more concerned about the 13% or so of the workforce that is self-employed - putting in all of those doctors, accountants, etc would change the high end of the private sector distribution quite a bit. But we don't really have accurate income data on the self-employed, especially those who are incorporated, because they can choose to retain earnings in their corporation, write off home office and other expenses (e.g. a fraction of car, phone, etc), convert employment to dividend income, etc.

There aren't any controls here because I'm just looking at a distribution of hourly wages - no regressions - not worth it given how skewed the income distribution is and lack of info on the self-employed.

[Luke I]

Hi Frances, thanks for your response. I agree with you wrt causality. That said, if you compare secondary school teacher wages in private vs. public institutions you would be getting some nice traction, similarly with private sector vs. public sector lawyers (police vs security guards are obviously quite different and a reflection of the roles of private and public sectors). As you mention, the problem becomes the degree of dissimilarity in the skills distributions between sectors. There are not really many public sector jobs in retail and food and accommodations and that is why I brought those two up.

Just a quick aside. I have not used the LFS pumf for at least a decade but I recall a categorical variable for self employment (the statcan folks pulled together something using that five or 6 years ago as well: https://www.statcan.gc.ca/pub/75-001-x/2011004/article/11535-eng.htm). filtering the data using this could get rid of that long tail perhaps?

[Luke I]

Sorry to be a nuisance - I will stop now - the variable I was thinking of in the LFS PUMF was COWCAN.

[Jacques René Giguère]

To add to confusion, the misallocation classification of public vs private sector jobs is sometimes significant.
In QC, cégeps professors are lumped in with high-school teachers even though they must hold much higher qualifications (economists, lawyers, engineers and such). We re told we are well paid for teachers even though the hiring pool is definitely not the same (no wonder we have recruitment problems).

[Frances Woolley]

Jacques Rene " the misallocation classification of public vs private sector jobs is sometimes significant." - very good point. E.g. where does Bank of Canada fit? I have no idea.