With the return of the Long Form Census there has been much discussion about having access to good data in Canada. Yet our world seems to hinge of another data set that frankly isn’t that good and probably won’t get better anytime soon.

This past Friday (November 6th) was “Jobs Friday” the day that Statistics Canada releases updated Labour Force Survey Numbers for across the County including “Windsor”.  At 8:30 AM on Friday morning journalists, economists, bankers, stock traders and nerds like myself eagerly refresh Stats Canada website to see the data can tell us about the Canadian Economy. For Windsor, as I was first to “report” Windsor unemployment rate climbed 0.1% in October while nationally it dropped.

Job’s Friday is also the same day that economists, job market watchers and nerds gather to take part in Labour Force Survey Guesses #LFSGuesses. Under this hashtag, created by Luke Kawa a journalist for Bloomberg TV, individuals guess what the top line jobs numbers are going to be for Canada on a given month with accolades being doled out for the closest guess. (I won in August)!

Part of the reason, that such an important statistical release has devolved into an ironic guessing competition is due to the perceived unreliability of the Labour Force Survey. Economist, Mike Moffatt has written on the issues with the Labour Force Survey in Canadian Business. This friday he tweeted:

after his guess of +38.5k jobs being created in October ended up being the closest the to actual October report. If we have one of the leading economists in Canada calling jobs data basically random, what should the average citizen, the media or businesses take from it?

Statistics Canada pegs the national margin of error at approximately 28,000 jobs each month! Meaning that whatever number is posted nationally for new jobs created you could add or subtract 28,000 jobs from that number and still have a statistically valid measure 19 times out of 20. The Huffington Post reported that many leading economists don’t take the jobs numbers seriously with one saying:

“We should not believe the data from month to month,” {Economist Will Dunning Chief Economist for Canada’s Mortgage Broker L:obby] concluded in an email to HuffPost.

“Windsor” Problems? 

So what does this mean for “Windsor”, the city with the highest unemployment in the Country? Well in my opinion it means a couple of things: first we don’t know if “Windsor” has the highest unemployment in the County. Unfortunately I couldn’t find what the local standard error but given the wide range of error at a national level, and the fact that the sample size shrinks as you get into small geographic spaces there are real problems for “Windsor”.

The LFS surveys approximately 56,000 households across the country each month, now if we assume that this survey is conducted proportionally to the workforce population (everyone over age 15) of 29,350,400 (in Oct 2015) that represents about 0.190% of the population. Windsor’s portion of that sample is in fact is 562 households.

This 562 households is directly tied to the second issue and why I have putting “Windsor” in quotations. Statistics Canada isn’t measuring Windsor Ontario unemployment rate but rather the Windsor CMA’s unemployment rate. The Windsor CMA is made up of Windsor, LaSalle, Lakeshore, Tecumseh and Amherstburg. No one from Windsor-Essex County would call LaSalle, Lakeshore, Tecumseh, or Amherstburg the same as Windsor but Statistics Canada does. These  communities all have different demographic, ethnic, political and socio-economic issues. As I have previously written, the City of Windsor has been a  long term victim of flight of middle/upper class families to the suburban communities that surround it and Windsor is also home much higher concentrations of poverty compared to the other CMA members (see image above). As a result it is quite likely that City of Windsor’s Unemployment rate is actually worse than we expect.

If we were to take 562 households and distribute them by percentage of population of the CMA communities we get the following distribution:

Community	Households Contacted
Windsor	371
Lakeshore	61
LaSalle	50
Tecumseh	42
Amherstburg	38

Now Statistics Canada doesn’t actually sample in this manner but the risk of oversampling from one community over another is present in their design. The 2015 Labour Force Survey Guide outlines their  methods in full but at a high level, Stat Can divides the population into income strata groups. These statas are then allocated at a provincial level based on Census, NHS and Administrative (Income Tax) data. The strats are clustered in groups, mapped and are used as then used as a basis to the individual interviews. The 6 strata’s each take part for 6 month period before they are replaced with a new sample strata, this processes occurs in stages so each month one new sample strata is introduced while another is removed.

The interviews occur the second full week of each month asking a “knowledgeable” member of the household questions about employment activities the entire household.  This means that for sub-groups like Youth, in many cases they are not being directly interviewed, rather a parent is being interviewed on their behalf.  Naturally this method of household questions can result in survey bias as household members are answering questions for other people.  Finally for a typical month, there is approximately a 10% non-response rate which further shrinks the local sampling size.

What to Do?

I am not writing this post to say that unemployment isn’t a problem in Windsor, but the fact of the matter is we don’t know how big of a problem it is.  Media, and politicians react to the numbers like it is the end of the world when in fact we don’t actually know what they mean. The numbers that appear on a month to month basis give us a crude idea of a trend, but nothing more. With the exceptions of the largest swings in the data, we don’t actually know employment trends are changing or if it is just a fluctuation within a statistical margin of error.

In all likelihood, given the concentrations of poverty in parts of Windsor, unemployment in the City  is probably much higher than the 9.8% rate that is currently reported at. Meanwhile, we have no idea what the surrounding communities unemployment rates are. Compounding the issue, is Statistics Canada and the media using “Windsor” interchangeably with the Windsor CMA data that they are reporting on. No one in Essex County considers our region a single unit, yet the data is being reported that way, with it all being attributed (good news or bad) to Windsor.

As a result, media and statistics Canada are in fact not reporting reliable data for our communities. Tecumseh, LaSalle, Lakeshore and Amherstburg politicians are able to dodge answering hard questions about their communities economic situations as the data is only being framed as a Windsor issue. Meanwhile, Windsor doesn’t actually know how big of a challenge it faces or what our actual labour forces looks like as the numbers are muddled with those of the rest of the CMA. Maybe if the data was reported accurately (as a region) we could begin to take steps to tackle the issues that we face. Of course that would require cooperating as a region first.