I have come out of retirement to talk minimum wage. Specifically, I want to talk about the Bank of Canada report that everyone is reading and talking about. Just kidding, no one is reading it. Get settled in because this is going to be a long one.
With just that link I have done more to give you the details of the report than the major news agencies. Based on most of the stories I believe they read the summary, reworded it for their “article”, wrote a headline that said the minimum wage increase would lead to the loss of 60,000 jobs, tweeted it out and called it a day.
Did you read anything about it today? Did you read the comments? What percentage of the people read the report? What percentage read beyond the summary? What about the cover page?
From the cover page:
“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.”
Okay fine this is probably just a standard disclaimer but still the BoC was not confident enough in this report to put out a “report” about the topic so they made it an “analyst note”. Also, this analyst note does not appear on their main list of publications when you go to the Bank of Canada website.
This is a very minor point and could be considered nitpicking but now we look at the methods they used to reach the infamous 60,000.
The headline is based on this bullet point from the summary:
Weaker labour demand leads to reduced employment and lower hours worked, although the net impact on labour income is positive. Employment losses amount to about 60,000 workers (hours worked decline by 0.3 per cent), a number that lies in the lower part of the range obtained from a simple accounting exercise (30,000 to 140,000).
This is based on some regression analysis which they elaborate on later in the paper. Now there is controversy in the world of data about “p-values” but for the purposes of this post I will assume that we can accept the values given in the paper as correct.
Crash course. Regressions are a mathematical tool to determine how much of a change can be attributed to specific factors. Example, how much of the rise in house prices can be attributed to either rising wages, foreign investment or new housing build rate. A simpler example would be how much of a workers wage can be determined by their years of education. The basic rule is if a number is “significant at a 0.05 level” then is likely acceptable as a result. 0.01 level is better and 0.10 level is probably pushing your luck.
Why is this important? Well we want to look at the results tables from this report. First they run their regression for the effect of minimum wage increase on levels of employment.
Why three different results? Well first they run it based on what are called SEPH wages which exclude the self-employed and agricultural workers and also exclude salaried workers, Fortin (2010). Then they use SEPH data (no self-employed or agricultural workers) but include salaried workers. Then finally they use LFS data that includes all workers. Ideally you would want to have the triple asterisks beside anything with a meaningful result; double would fine, single is pretty much out and no asterisks is a meaningless number.
What do the numbers mean? For every 1% increase in minimum wage what percentage does employment rate change. Negative means jobs lost, positive means jobs gained. It doesn’t matter what the number if it is not significant at the proper level.
Next step is to try to get some robustness in their number by controlling for sex and education. In this case they tried running it in two separate ways once with all of the age groups separated and then again when all the employment data combined.
See those numbers below “Age 15+”? They have no asterisks. That means when controlling for sex and education and combining all wage data there is no change in the employment rate. That is, 0 jobs lost.
The second bullet point I would like to focus on is something that people against minimum wage can certainly latch onto.
The direct pass-through from a simple reduced-form approach suggests that minimum wages could modestly boost consumer price index (CPI) inflation in 2017, ranging from 0.0 to 0.1 percentage point (pp) and by about 0.1 pp on average in 2018, ranging from 0.0 pp to 0.2 pp. The impact for CPI inflation in 2019 is also likely to be modest, ranging from 0.0 to 0.1 pp.
This is for your friends that say, “higher minimum wage just means higher prices for everyone and it just balances out!” If the National Post was writing the headline on this bullet it would be, “Inflation to Surge 0.3% through 2019 after Minimum Wage Hike”. In reality, the results of the paper are inconclusive on this topic as all the ranges provided include a 0.
So you’ve made it through the regressions, next up, the simulations! Oh the BoC simulations. The mysterious ToTEM III model was just announced in October 2017 and was used in this paper. Now I have not looked closely at ToTEM III but its predecessors have been used for the monetary policy reports in the past. These are the models they use to make the same prediction every few months, that inflation will return to the 2%, and are wrong almost every time.
The main take away from this is that the simulation says that the added consumption from all these higher paid employees (because they do conclude that labour income will increase) will be offset by a rise in interest rates. The bullet points states:
Consumption would be reduced slightly as the higher inflation would elicit a slight interest rate increase, which would more than offset the higher labour income.
Since their predictions about inflation rates based on this model are consistently incorrect you may take this statement as you wish.
The main takeaway from reading through this article is that the results are not conclusive and it appears that the authors feel the same. Look at the passive language of the final summary bullet:
Potential output should remain unchanged in the short run. Longer-term effects are possible through automation, productivity changes or changes in labour force participation. The sign of these longer-term effects is, however, ambiguous.
This paper will be the basis of 1,000 Facebook posts about hating minimum wage. No one will have read it. I am all for debating the topic but nothing in this paper can be used as conclusive evidence for either side of the topic. It makes for great click-generating headlines but the real conclusions of this paper are, however, ambiguous.