Update: A revised version of the paper discussed in this blog, with results close to those obtained using the PUMF, is available here. Thanks to Feng Hou for taking my concerns seriously and responding to them promptly.
A recent Statistics Canada study by Wen-Hao Chen and Feng Hou reported a disturbing finding. Many second generation visible minority Canadians - "individuals who were born in Canada to at least one immigrant parent" - experience low employment rates. The relatively poor labour market outcomes of Black and Arab Canadians are disturbing but, given the obstacles racialized minorities face in the labour market, not unexpected.
More surprising were Chen and Hou's reports of low employment rates for second generation Chinese and Korean Canadians. The employment rates for second-generation Chinese Canadian men were, according to Chen and Hou, just 77.6 percent in 2016 - 10 percentage points less than employment rate for comparable Filipino Canadians, and 9 percentage points less than the employment rate for second-generation "white" Canadians. The employment rate for second generation Chinese Canadian women is, according to their estimates, even lower: 73.7 percent.
These numbers seemed odd to me, given that Chinese Canadians typically have favourable labour market characteristics, such as high levels of education and access to strong labour markets. Discrimination might explain part of the gap, as might Chinese Canadians staying in school longer than members of other ethnic groups - but could these two factors add up to a full 10 percentage point difference between Chinese and Filipino employment rates? I didn't think so. I tried, therefore, to replicate Chen and Hou's findings using the 2016 Census Public Use Microfile. I couldn't.
Here are my results:
Some of the numbers are quite close to Chen and Hou's. The largest and most precisely estimated number - the employment rate for whites - is within 2 percentage points of the Chen and Hou figure. However others are very far off.
There are a few possibilities.
One is that I've made a coding error. That is entirely possible. Here is my Stata .do file if anyone would like to check it: Download Chen and Hou replication.
Another is that the 2016 Census PUMF is unrepresentative of the Canadian population. If that is true, it is a matter of serious concern to researchers and others who rely on the PUMF.
A final possibility is that Chen and Hou made a mistake in their analysis.
WCIers, what do you think?
In response to comments/questions on twitter and by email:
- I didn't use weights. The weighted means are pretty much identical to the ones reported in the post.
- using DPGRSUM instead of doing what I did - using VisMin and eliminating people reporting Aboriginal heritage - doesn't make a difference either.
Posted by: Frances Woolley | March 21, 2019 at 05:00 PM
The enormous disconnect between West Asian/Arab number in the two sets of results is also striking.
Posted by: Bob Smith | March 21, 2019 at 05:11 PM
HI Frances,
Interesting puzzle. My instinct is that the Chinese and Korean numbers should be far above the average (as you have them) and not below the average (as the STC paper has them).
I don't know which of your enumerated possibilities is the source of the difference, but I would be very very surprised to see those kinds of employment rates for Chinese and Korean men and women.
Posted by: Kevin | March 22, 2019 at 12:32 AM
Kevin, Bob - Feng Hou has re-run his code and it seems that, yes, there was a mistake in the paper.
Interestingly, even though the employment numbers for Chinese/Korean/white 2nd generation Canadians are roughly the same, the unemployment rates appear to be slightly lower for Chinese and Korean Canadians, because more of them are out of the labour force, presumably in school. There isn't much of statistical (or other) significance in this however.
Also, if you take a narrower age range, i.e. 30 to 39, so minimizing the effects of people still being in school and the non-white population being, on average, a lot younger than the white population, several of the visible minority groups employment rates start edging above the non-visible-minority employment rate, especially for women (approx 88% employment rate for Latin American between 39 and 39, 87% employment rate for Filipino women, 86% for Chinese, v. 82% for white women).
Posted by: Frances Woolley | March 22, 2019 at 07:59 AM
Prof. Woolley you caught a bug, as they say in software development. :)
This post really caught my attention as I'm married to someone of Chinese descent, and have two daughters who are mixed. As a white male who is generally clueless, this has made me more aware of cultural differences and issues of possible discrimination. Pure speculation here, but the cause may also be culturally ingrained; some are no doubt be more comfortable with the act of "selling oneself" in a job interview, while others have been taught to be modest, not brag about yourself, lead by action rather than words, etc. Selling is like any other skill, it requires a template and practice. These factors cannot be measured easily if at all, as there is no single Asian culture, and inter-generational differences are vast.
On a side note, I know several Canadians with Masters Degrees in economics and they are still working in bars, restaurant etc. This may be purely co-incidence, but they are all of happen to be of either Korean or Chinese descent - how frustrating!
Posted by: Peder Jakobsen | March 22, 2019 at 02:43 PM
Peder -
I teach a professional practice course of economics course, and one of my favourite exercises is to get people to write job application letters. Many students find it very difficult, for precisely the reasons you mention. And it's not obvious that bragging about oneself is a good strategy in any event - the job market isn't always kind to people who don't follow social convention.
Arthur Sweetman, Ross Finnie and others have done some interesting work on the probability of going to university as a function of high school grades for Canadians of different ethnic origins - see e.g. Figure 7.5 in this publication https://read.oecd-ilibrary.org/social-issues-migration-health/catching-up-country-studies-on-intergenerational-mobility-and-children-of-immigrants_9789264301030-en#page192. One of the most interesting findings is that very mediocre Chinese Canadian high school students - ones with high school averages of, say, 65% - still end up going to university more often than not.
I suspect a lot of these students would have better life chances if their parents pushed them to work on their non-cognitive skills, and pushed them into less academic subjects. I feel strongly about this because these are precisely the students who often end up taking Econ, because their grades aren't good enough to get into business or the sciences or engineering, and econ at sounds like it should be useful. Plus it doesn't involve a lot of essay writing. But I'm not sure they get a lot out of it.
Posted by: Frances Woolley | March 22, 2019 at 04:32 PM
Good job, nice to see the system working.
I look forward to the revised paper.
Posted by: Bob Smith | March 22, 2019 at 04:42 PM
Prof. Woolley,
A great cover letter for an application is important, but the face to face interview, that's where even the best candidates can fail, while the mediocre can easily oversell themselves by sheer gamesmanship.
I believe the course you are referring to is 3920, the one I'm missing to upgrade my BA to a BA Hon. :) I'm just a hobbyist, will never work as an economist, but it should be interesting regardless.
Posted by: Peder Jakobsen | March 22, 2019 at 05:05 PM
Mobility from a Canadian community that resembles home is probably the main unaccounted variable.
Posted by: lateral Mimas | March 26, 2019 at 10:57 AM