Unionization and Unemployment: My Canadian Ears Are Burning

Three months ago, Anne Layne-Farrar intervened in the US debate about the Employee Free Choice Act (EFCA) with a widely-reported paper and Senate testimony. She used Canadian data to argue that the proposed legislation would eliminate 600,000 American jobs.

As many critics have noted, Layne-Farrar works for a corporate consultancy and business funded this piece of research. Readers can take my comments with a similar serving of salt because I work for a union that strongly supports EFCA south of the border.

The possible relationship between unionization and unemployment has been a hot topic in labour economics for decades. Obviously, it is possible to find many studies pointing in opposite directions. Keynesians would tend to emphasize the role of unions in upholding wages during a downturn, which helps prevent a deflationary spiral and the consequent unemployment.

The mainstream, neoclassical consensus is that unions do not have a statistically significant effect on unemployment one way or the other. Even the Organization for Economic Co-operation and Development, typically a proponent of “flexible” labour markets, takes this view. So, it is notable that Layne-Farrar claims to have demonstrated empirically that higher unionization leads to higher unemployment.

Specifically, her regression analysis suggests that, among Canadian provinces between 1976 and 1997, each 3% increase in the unionization rate caused a 1% increase in the following year’s unemployment rate. Dean Baker points out that, if Canada and the US are so similar (except for unionization) that this relationship is transferable across the border, Canada’s much higher unionization would mean much higher unemployment in Canada than in the US. In fact, the Bureau of Labor Statistics calculates a lower unemployment rate for Canada than for the US (5.3% versus 5.8% in 2008; 7.1% versus 8.9% in April 2009).

But why use historical Canadian provincial data in the first place? Examining unionization and unemployment among American states would seem more immediately relevant to US policy, provide five times as many data points (50 states versus 10 provinces), and take account of wider variations in unionization.

In both her paper and Senate testimony, Layne-Farrar offers a song and dance about how Canada is an ideal “natural experiment”. While US labour law is mostly set at the federal level, certain provincial governments have enacted and repealed card-check certification, a key feature of EFCA.

Despite this seemingly sound rationale for focussing on Canadian history, she does not actually explore the correlation between card-check certification and unemployment. She just looks at the correlation between unionization and unemployment. This comparison is defensible insofar as EFCA’s stated purpose is to increase unionization.

However, since the regression does not consider whether or not card-check certification was in effect, there is no need for the time period to encompass these legislative changes. Indeed, one might as well use the most recent data available. Doing so would likely suggest that higher unionization corresponds to lower unemployment, the opposite of Layne-Farrar’s argument.

Of the five most unionized provinces in 2008 – Newfoundland and Labrador, Quebec, Manitoba, Saskatchewan, and British Columbia – four had unemployment rates among the five lowest in Canada in May 2009. The exception was Newfoundland and Labrador, which has very high unemployment for historic and geographic reasons.

Of the five least unionized provinces, four had unemployment rates among the five highest in Canada. The exception was oil-rich Alberta, which still has a relatively low unemployment rate (albeit a higher one than its more unionized Prairie cousins).

Most Canadians know it would be ridiculous to argue that low unionization “caused” high unemployment in Ontario and the Maritimes, or that higher unionization “caused” lower unemployment in Quebec and western Canada. Similarly, American policymakers could easily think of relevant variables overlooked by a superficial correlation between state unionization and state unemployment.

Therefore, I cannot help but conclude that Layne-Farrar’s main motive for using Canadian data was to formulate an analysis that would be difficult for her American audience to question. It reminds me of the old saw that “trade barriers” between Canadian provinces exceed those between European countries, a claim that Canadians cannot refute without detailed knowledge of the European Union’s economic institutions.


  • A couple of small points.

    I am always amazed at how corporate backed groups have used their position with history to fully envelop the union-unemployment correlation.

    To me, and I have been quantifying labour market issue for much of my career, the relationship between unionization and unemployment is so filled with intervening variables that any findings that suggest their is a correlation in a positive or negative direction are spurious at best.

    It is just too difficult to measure from the level she is attempting.

    It is one of those issues that is a dead horse. However, the skeleton is regularly brought out for display and to me every time I read such a paper, I automatically find my self laughing and chucking it into the bin.

    Why do they not turning the tables and start looking into oligopolies and monopolies within the various sectors and have a look at unemployment. Innovation, R & D, organizational behavior, and a whole host of other variables are subjected to disruptive forces a hell of a lot more than some cost (if it exists) of unionization. Especially with the shop floor flexibility the business world has gained through many devices as lean production and a raft of other employer lead and in some cases negotiated changes to the shop floor.

    To me the argument is strictly a qualitative one at the sector level, and this given the nature of the variables under study. Too complicated a for such simplistic measurements, given the layout of the causation involved. (i.e. the cost of doing business)


  • I agree with Paul. The macroeconomic factors far outweigh small changes in union density. The list of things that have a bigger impact on unemployment is quite long: resource prices, housing starts, interest rates, exchange rates, resource endowments, education levels, female participation rate, transit gridlock, access to childcare, technological change…

    Actually if you trash all the research coming from vested interests, the academic literature on this is quite robust in that the impact is either non-existent or rounds to zero after statistical significance is taken into consideration.

    People are always forgetting that after a workplace unionizes and successive collective agreements increase wages, the employer is always able to recruit a better pool of applicants, reduce turnover, derive a higher yield from training funds, save on reduced WCB expenses, and take advantage of the informal operational audit services provided by the union. A large fraction of the higher union wage is made up of higher productivity and value of work.

  • Actually Stuart, I should have made the point that it all depends on what kind of an adding machine one uses to determine the value of a union.

    Health and safety and lives saved from increased safety, increased wages and benefits that add to a person wealth and health that is never directly translated to a balance sheet or income statement and I could go on and on.

    So again I will restate, until the time comes that a proper accounting method is in place, the only practical way of measuring a unions impact on the workplace is the qualitative, and if you are a worker- you know the answer to that question.

    A final note, a properly run company does not require a union, union’s are a result of bad management. It is a natural response by the herd, a collective reaction to those negative actions of one group of individuals against another group.

    If we still lived in the trees, I would bet that the unionized primates would be living a lot higher up picking the dangerous fruit then the ones down below.


  • Very good post Erin. The EFCA is a major issue in the U.S. right now and as Erin mentioned, there has been scaremongering from the right about potential job loss using biased analysis and out-of-date Canadian data and other hyperbole. Home Depot co-founder Bernie Marcus called it the “demise of civilization” and said that retailers who didn’t oppose it should be shot and thrown out of their jobs.

    When CUPE vice-president Barry O’Neill gave a speech to a small business audience in Denver a few weeks ago, he pointed out that two Canadian provinces with card-check and relatively high rates of unionization–Manitoba and Saskatchewan–both had the lowest unemployment rates in Canada and the highest rates of economic growth of all provinces in Canada last year. This may be somewhat incidental, but it helps to put the lie to claims card check will lead to hundreds of thousands of jobs lost (as well as issuing in the general demise of civilization).

    There have been some studies done of the impact of card-check on rates of unionization in Canada. I suspect that there are a lot of other macro factors with more significance that affect this relationship and that card check won’t have a massive impact on rates of unionization in the U.S. The real issue behind this from the corporate lobby in the U.S. seems to be one of trying to inflict a political defeat on Obama.

  • John Raudabaugh

    Disclosure: I am a U.S. citizen, labor lawyer, adjunct professor, and former NLRB Member. Comment: I am of the opinion that “scholars” and “academics” and self-acclaimed “scholars” and “academics” can make a tremendous contribution by having the second footnote (the first footnote citing their affiliation and related information) to all forthcoming articles/”studies” disclose ALL sources of funding for their research and ALL affiliations/memberships/officer positions with ALL groups/entities/NGOs etc. While the above criticism (or observation) of Layne-Farrar for being employed by a non-academic entity is fair, there are many, many articles cranked out by U.S. “scholars” and “academics” that are funded directly by the very interest groups benefitting by the “findings” a/k/a personal beliefs and policy preferences.

Leave a Reply

Your email address will not be published. Required fields are marked *