Politicians have a love hate relationship with the Australian Bureau of Statistics' monthly Labour Force Survey. The monthly measures of the unemployment rate and employment growth are also subject to some of the most error.
For the most part, “error”, does not imply the ABS has made a mistake in adding up their numbers; it just means that they’ve used only a sample of the population in their survey and there is a random element to each estimate.
Unfortunately, all too often this random element is reported as real, with politicians inferring that the other mob is doing a bad job or that they are doing a great job at managing the economy – usually based upon little more than random fluctuations from a sample.
Do politicians ‘create jobs’?
The ABS’ latest update in July 2015 reported that the unemployment rate, in seasonally adjusted terms, increased from 6.1% to 6.3% and that employment grew by 38,500.
Both sides on politics jumped on these numbers immediately. Federal Treasurer Joe Hockey said this week that
In Australia, last month, 38,000 new jobs were created. That’s 10 times the number of jobs in one month than were created in the last month under the Labor Government.
The politicians may be quoting ABS numbers, but there’s something very wrong about the way they’re framing them.
The ABS also publishes its margin of error, or standard errors. For the unemployment rate of 6.3%, the ABS is 95% confident that the true rate is between 5.9% and 6.7%. They are equally confident that employment growth was between -19,500 and 96,500.
So we can’t be entirely sure whether in July employment went up or went down. They’re also uncertain if the unemployment rate went up or down. Statistically, it is most likely that both went up – but we don’t know for sure.
That’s because the labour force report is based on a survey of around 52,000 people. This sample represents around 0.3% of Australian adults in a given month. Had a parallel survey been run in July with a different 52,000 people sampled, the results would have been different.
Results from a sample are only ever an “estimate” and therefore subject to what is known as sampling error. The single estimate the ABS provides is our best estimate, but we know that it is possible the “true” population values for unemployment or employment could be better or worse than the ABS survey estimate.
The Labour Force Survey is large compared to opinion polls, which are often between 1000 and 2000 responses. In spite of this large sample, the survey still results in a large error band for “rare events” such as unemployment or “movement” estimates such as employment growth. The ABS could reduce the error in these estimates by increasing the sample – but this increases the cost of the survey and would require the ABS to make savings and likely reduce quality in other publications, given their budget constraints.
A better way to interpret jobs data
A sensible approach with the labour force survey is to either have one eye on the seasonally adjusted numbers and another eye on the standard error tables or, perhaps more simply use the “trend” numbers which remove much of the month-to-month volatility in the seasonal numbers. These are all published in the same ABS publication and spreadsheet.
The problem for economic journalists is that standard error tables complicate the story. The trend numbers are a much less interesting story. The trend numbers for the unemployment rate show that unemployment has been at 6.1% since March this year and was at 6.2% back to August 2014.
So, a few key points to remember about the labour force survey:
Movements in monthly employment statistics are quite volatile. Focusing on the trend statistics and taking a longer view of seasonally adjusted or original numbers makes more sense.
Employment statistics tend to be a lagging indicator of the labour market and the economy more broadly, since they reflect the hiring intentions of firms in the months prior the release of the numbers.
The strength or otherwise of the labour market is determined by both the demand in the economy and the supply of labour from individuals. Both of these factors are driven by international and domestic economic conditions, confidence in the economy, economic policy from years in the past, structural changes in the economy and a range of other factors. Politicians play a role in these factors, sure, but it’s unlikely they can claim much credit for a strong monthly outcome of the labour force survey – or that they are to blame for a poor outcome.
Employment growth does not equal “jobs creation”. Employment growth is the net outcome of much larger gross flows of employment which includes both job creation and job destruction.
No single economic statistic is perfect. Each has their strengths and weaknesses. A better understanding of the labour market and the broader economy should include taking note of a wide range of economic statistics both within the labour force publication (such as hours worked, underemployment statistics) and beyond the labour force publication such as business confidence, national accounts, job vacancies and population statistics.
Ben Phillips works for the National Centre for Social and Economic Modelling (NATSEM), an independent university research organisation specialising in academic research that undertakes consulting services to a wide range of clients.
Authors: The Conversation