By Rodney Dickens*
A review of the predictions for employment growth by the economic forecasters reveals that they are pretty good as long as employment growth is about to head to around average rates; which doesn’t happen very often. In general they have a poor track record of providing advance warning of major upturns and downturns.
Looking at situations where they got it badly wrong helps reveal a fundamental flaw in the approach to forecasting used by the economists.
Largely in response to the large fall in dairy export prices in 2014 and 2015 the Reserve Bank, like the eight forecasters surveyed by NZIER on average, predicted weak employment growth over the next couple of years. The chart below shows the September 2015 Reserve Bank and consensus employment growth forecasts. Instead of being weak, employment growth turned out to be well above average. This example helps understand one of the shortcomings of the traditional approach to economic forecasting.
Why did the economic forecasters get it so wrong? They assumed there was a link between dairy product prices and employment growth that simply didn’t exist as shown in the next chart.
There is no correlation between the two as shown in the previous chart. Tumbles in dairy product prices are just as likely to coincide with strong employment growth as weak growth because other factors are dramatically more important drivers of employment. Equally, there is no correlation between GDP growth and dairy product price inflation as shown in the next chart. The same is true if I allow 12 months for dairy prices to impact.
The economic forecasters were happy to leap to a conclusion without checking whether it was well founded. And they were happy to torture their forecasting models to come up with weak forecasts for employment growth in response to the fall in dairy farm incomes despite there not being any historical relationship or linkage between the two.
The main problem with the traditional approach to economic forecasting is that the forecasters are far too willing to plug in dubious assumptions and to ensure that their predictions match their misguided preconceptions. By contrast, following the fall in dairy farm incomes we focused on the stimulus from falling interest rates and high net migration that fuelled robust growth in residential building activity and GDP (next chart) and employment.
What use are the various employment-related surveys?
On average the seven forecasters surveyed by NZIER in June 2018, like the Reserve Bank (RB) are predicting a large, ongoing slowing in employment growth (next chart). It’s no coincidence that the economic forecasters on average are predicting a similar outlook to the RB as was the case in 2015. This reflects groupthink that is much less useful than quality, independent analysis.
Predictions of weak employment growth are needed to ensure the economists don’t predict the unemployment rate falling to levels that would cause a wage inflation problem. Average hourly earnings inflation has already spiked to 3.6%, but the economists are ignoring this. Too often the forecasters come up with the results they want (e.g. there won’t be a wage inflation problem) and then effectively work backwards from there in coming up with predictions for employment. I saw this firsthand in my time at the Reserve Bank.
In our pay-to-view reports the focus is on the stimulus in the pipeline from the range of government initiatives that should in time underwrite a better outlook for employment growth than suggested by the chart above. However, the focus of this Raving is on how useful the leading indicators of employment are at assessing nearterm prospects. Will employment growth behave in line with the predictions in the chart above at least in the near-term?
The first candidate is the ANZ employment intentions survey that at times in the past has been quite a useful indicator of near-term prospects for employment growth (next chart). At face value this survey suggests the economic forecasters are on the mark. The tumble in employment intentions points to further and significant near-term slowing in employment growth. However, the ANZ survey has been plagued by political gamesmanship since the election (e.g. this link below for the January Raving that pointed out this problem in general).
Comparing one component of the ANZ survey with a similar component from the NAB Australian business survey helps assess whether the ANZ survey has been corrupted by gamesmanship. Despite a lower NZD TWI the ANZ exporters’ survey has crashed while the equivalent NAB Australian survey has spiked (two charts below).
Australian exporters should be facing roughly similar global factors as NZ exporters and like NZ exporters have benefited from a lower exchange rate improving international competitiveness. The NAB survey has behaved roughly as one would expect in light of major relevant developments while the ANZ survey of exporters has gone AWOL bigtime; suggesting that things are almost as bad as they were at the peak of the financial crisis.
Job ads should be quite a useful leading indicator of employment growth. For a number of years ANZ has supplied data on job ads and this is one of the components in the Department of Labour (DOL) online job ads index (next chart). Somewhat strangely, they have behaved differently recently. But how useful are job ads as leading indicators of employment growth?
Annual growth in the ANZ job ads is a moderately useful leading indicator of near-term employment growth in terms of the direction growth is heading but since 2013 it has been a poor indicator of the rate of employment growth (next chart). The peak correlation is with growth in job ads leading employment growth by one quarter with a so-so correlation of 0.57 (1.0 maximum possible). Job ads provide a case for probably expecting a small amount of further slowing in employment growth.
The NZIER survey of employment intentions has been a more accurate leading indicator of employment growth with the peak correlation at a moderately respectful 0.69 being with it leading by two quarters (next chart). It hasn’t been overly accurate in the last several years which isn’t entirely its fault (see the next page) and to the extent it is still useful it suggests little near-term change in employment growth.
The BNZ-Business NZ survey initially offered quite a bit of promise in terms of a weighted average of the manufacturing and service sector employment surveys but since 2013 it hasn’t been very useful (next chart). This partly reflects problems with the official data. The peak correlation of 0.64 is with this combined survey advanced or leading employment growth by two quarters. It suggests there could be a bit more near-term slowing in employment growth.
Westpac McDermott Miller release a quarterly employment conditions index that is made up of components designed to measure more than just employment prospects but the overall index is quite highly correlated with the employment-related components. Like the ANZ job ads this survey can be moderately useful in assessing whether employment growth is in the process of improving or deteriorating but it experienced a major dislocation as a result of the financial crisis that means in level terms it isn’t at all useful at assessing near-term employment growth prospects (next chart). Like the overall employment conditions index, the employment-related components have also improved recently. But this survey didn’t signal the latest slowdown in employment growth making us wary of the near-term improvement it predicts.
From a slightly different perspective, consumer surveys can at times be useful leading indicators of employment growth because developments in the labour market are one important factor shaping the expectations of consumers; most of whom are employees. The peak correlation is with the Westpac McDermott Miller consumer confidence survey leading annual growth in employment by two quarters but the correlation is so-so at 0.56 and there have been significant departures between what the survey has predicted and employment growth (next chart). For what it is worth, this consumer survey suggests employment growth may weaken marginally in the near-term.
There isn’t a shortage of leading indicators of employment growth but none are accurate enough to have a great deal of confidence in what they are predicting for the near-term outlook. I suggest people be wary of the commentary by the people reporting the results of all the surveys in part because none have been particularly accurate in recent years; something the people writing the commentary seldom take the time to point out. In addition, such commentary will suffer from the same problems that undermine the economic forecasts (i.e. preconceptions and wishful thinking will excessively shape the commentary). However, partly in defence of the forecasters and the leading indicators, part of the problem is with the employment data that can have significant shortcomings in part because of questionable decisions made by Statistics NZ as discussed in two Ravings last year (two links below).