Census questions

The 2026 Census was in the media a few weeks back for all the wrong reasons.

The issue is about whether the questions of sexual orientation and gender identity should be included in the 2026 Census.

It doesn't take someone with a PhD to know that the wrong decision has been made to exclude these questions, again. Where this decision sits between the Government and the Australian Bureau of Statistics (ABS) is more complicated.

Based on my cursory reading of how the ABS handled some of the 2021 Census data, their cavalier approach is a bit worrying.

There are various stated reasons for the exclusion of questions on sexual orientation and gender identity, but one not discussed is the length of the Census.

The length of surveys is usually restricted for various reasons. Someone at the ABS has probably defined the overall length of the Census, and therefore, you can't just 'keep adding questions in.'

That length would be governed by the number of pages it takes to print. Adding an extra page would substantially increase the cost of printing. About 10 million dwellings in Australia receive the Census. Adding one extra page requires an extra 10 million pages of paper. This is not a trivial decision.

Constraining the length seems fair and reasonable until you consider the time, cost, and effort required to produce, deliver, and collate the data. I've heard that the Census is the biggest peacetime activity undertaken in Australia.

Pointing fingers at others and telling them what they should have done is easy. I'm not involved in the design of the Census, and I doubt I ever will, as interesting as that would be.

If there's any organisation capable of designing a survey in Australia, it's the ABS.

This is not about the example, they never are, it's about your organisation.

To make my point, I tend to frame questions in a harsh, didactic, and unpleasant way. Being nice doesn't work.

Do you think your organisation can write better surveys than the ABS?

The correct answer is no.

The pertinent question for every organisation in Australia that writes surveys is - what egregious errors have you made that you don't even know about? Hint: they're probably not about gender because you shouldn't be asking about this.

How corrupted and useless is the data you've spent all this time and effort collecting?

Why don't you just write one great survey and use that instead of the pointless crap you send out?

If the model of one great survey is good enough for the ABS, maybe you should try this, too. If your organisation lacks the capability, hire the best person you can.

Why is this hard to implement?

Writing a survey is the litmus test of the analytics profession. It's the equivalent of an omelette or poached egg for a chef. Both the survey and the eggs tell someone a lot about the person's competence in their profession.

There's a popular, widely held and oft-repeated view about 'doing the work'. In a recent Farnam Street blog, the comedian Cameron Esposito is quoted as saying:

“There is no formula for success—you just begin and then you continue. I’m often asked how to have a career in stand-up and the answer is confoundingly simple: Do the work. Over and over again, just do the work. After you build the courage to get onstage that first time, it’s all about repetition.”

Anyone who writes about 'thinking' doesn't understand that statistics doesn't follow the physics model of teaching. The physics model is the idea that you can take something complicated and simplify the concept to be basic without losing the key concept.

The example I use is of a block sliding down a frictionless slope. A young child quickly grasps what's happening without knowing calculus. The steeper the slope, the faster the block moves. You can then build complexity into this over time.

Statistics doesn't follow the physics model; which is why 'smart' people stumble over the concepts of standard deviation and standard error of the mean. Both sound more complicated than they are. Using this correctly is a litmus test of understanding the difference between individual data and mean data.

As is evident by the number of examples I've provided in the past few weeks, incompetence knows no bounds. Irrespective of the company you work for or the title you hold.

You could write 1,000 survey questions and never write a decent one. Just ask Harris Farm.

Speaking, writing, and even managing people can improve by doing the work if you use the feedback.

Without skill, expertise and understanding, you'll never produce a visualisation in Excel that doesn't suck.

If you want to be a comedian - write, practise and perform your show. You'll get better.

If you want to be vaguely useful in analysing data, you'll need to know when to use Fisher's exact test and when to use a paired t-test.

If you want to write a survey that tells you something useful, you need to understand statistics.

Unlike comedy, there is a formula for success with statistics; you just need to understand probability theory. This is much harder than 'smart' people realise.