The problem with SA's new unemployment data - only 0.1% of SA households were surveyed
- The latest set of unemployment data from Statistics SA reports that 6.5 million people were jobless, with the unemployment rate spiking to a record level.
- But due to the pandemic, only 18,021 households were surveyed to determine these numbers.
- While it's clear that SA has a jobless crisis, Old Mutual's head of economic research believes the past two quarters' data are not reliable.
- For more articles, go to www.BusinessInsider.co.za.
On Thursday, Statistics SA released a set of horrific unemployment numbers for the third quarter. It reported that 6.5 million people were now jobless and that the unemployment rate jump to a record 30.8%. Under the expanded definition of being unemployed (which also would include people who gave up looking for a job, or couldn’t look for a job during lockdown) the jobless rate has now reached 43.1%.
But the numbers for one of the country’s most important economic indicators came with big caveats.
Only 18,021 households were surveyed to determine the employment status for the total adult population. Given that there are between 17 million to 20 million households in the country, this means less than 0.11% of SA’s households were contacted to confirm their employment situation.
Usually, Statistics SA contacts around 30,000 households through in-person interviews. But it suspended face-to-face data collection in March due to the pandemic. “This was to ensure that the field staff and respondents were not exposed to the risk of contracting coronavirus and to contain its spread,” the organisation said in a statement.
It then switched to “computer-assisted telephone interviewing” instead of face-to-face interviews.
Statistics SA tried to call the same households which were surveyed for the first-quarter unemployment data.
But many households didn’t have numbers listed with Statistics SA, and other numbers were found to be invalid, some were not answered and some families moved since the survey at the start of the year. In the end, for the past quarter, Statistics SA only got hold of 18,000 of the original group of 30,000 households.
“Given the change in the survey mode of collection and the fact that Q3: 2020 estimates are not based on a full sample, comparisons with previous quarters should be made with caution,” the organisation warned this week.
As a result, the unemployment data for the past two quarters are not reliable enough to take seriously, says Old Mutual’s head of economic research Johann Els.
Only after collection methods have normalised for some months, will the labour data start to be reliable enough to show the real impact of the pandemic on the labour market. This may only be in a year’s time, Els believes.
SA's unemployment data problem
And even then, there are real problems with the way South Africa collects and reports unemployment data.
“South Africa’s inflation data gathering is excellent, our GDP data is relatively good, but we are very weak when it comes to employment reporting,” says Els.
Statistics SA has been surveying only 30,000 households to determine South Africa’s unemployment data for more than a decade now. This is despite South Africa’s population growing by almost 10 million people in that time – from an estimated 49.99 million to 59.62 million.
This may be due, in part, to the organisation’s severe budget constraints, which it has warned could start to hurt the quality of its statistics. Earlier this year, Statistics SA reported that a fifth of positions at the organisation are now vacant due a lack of money. The South African Statistics Council, which endorses the data released by Stats SA, threatened to resign if the statistical agency did not get the more money. This does not seem to be forthcoming, with Treasury recently approaching Stats SA to cut its budget even further.
Furthermore, collecting employment data in emerging markets like South Africa is notoriously difficulty, given the large informal sector and a lack of infrastructure.
But even in developed countries like the US, very small sample sizes are used, which results in big differences between labour surveys, and huge adjustments made following every release, says Els.
Statistically speaking, a sample size even as low as 16,667 households could still give a 99% confidence level in the employment status of 20 million households in South Africa, according to a PwC analysis. That’s if the households are carefully selected to reflect the broader population.
Els says that while South Africa’s labour data may not be that accurate, it does respond to growth in the economy, which shows that it is reliable to track the broader employment trend. In the middle of the 2000s, for example, the unemployment rate dipped when the economy grew by more than 5% a year.
And whether or not it is accurate to the percentage point, the larger crisis is clear: South Africa has a massive unemployment problem.
Els believes that this will only be solved by a change in government policy, through privatisation and labour deregulation. Due to South Africa's labour regulation, it is very difficult to get rid of employees when a company is struggling. This means that once they have retrenched workers following the pandemic, companies will not be that willing to employ people again – they will opt for machines instead, warns Els.
Receive a daily news update on your cellphone. Or get the best of our site emailed to you.
Go to the Business Insider front page for more stories.