Matt Nolan

Are real wages lower than 40 years ago

/in , /by

While rolling around the internet I found the following:

Lets have a look shall we.

Interesting!  But something seems a bit off – surely this can’t be true!?  Let us investigate.

It is from this book.  The data sources can be found here.  Cheers to Shamubeel Eaqub for the heads up.

So we have the following sources tied together:

  • 1957–1978: Real average hourly earnings including overtime, Statistics NZ, Infoshare Table reference: ERN001AA [ERN]
  • 1978–1992: Prevailing wage rate index, deflated by CPI from Graph 2.2, Statistics NZ, Infoshare Table reference: PWI007AA [PWI]
  • 1992–current: Average hourly wage rate, deflated by CPI from Graph 2.2, Statistics NZ, EESQ.SASZ9Z [QEX]

My studies have gone back to 1988 (when the QEX started conveniently) but I have noticed the difference between the PWI and QEX in the past and wanted a correspondence.  Two points come up here – the ERN was published until 1986Q1 (with index figures on the Stats site) and the QES/QEX started in 1988Q1.  Why was the PWI used for such a large section in the middle – also note the published PWI on the site is an index as well.  My biggest query for this period has to be – why use the PWI and not the Employment Survey that QES/QEX replaced?

Looking at the data it is the middle series that looks most out of line with other data – although the 1970s data it quite interesting in of itself, it looks like it might be very methodologically different, being collected by the Department of Labour and not including a sex split until 1973.  So lets look at the data history here on Stats NZ site.  Although the wage data is on the site under Infoshare there is nothing on methodology – so a trip to the National Library is in order.  That can be a future post 😉

Constructing our own wage series

Hey, we can tell there are issues with the data here – but is there some way we can create our own wage series?  Take the assumption that the levels are measured differently – but the “growth rates” would be the same irrespective of the measure we used.  Taking that idea we can tie the data together.

First off I am going to stick to the series tied together here, but instead of using the periods of time mentioned I will kick the series off as soon as it is released.  I am also going to include the Employment Survey (ES) to give us some idea of how much using the PWI “matters” for the result.

So to understand the data I have taken each index, and once the index starts I set it to the value of the prior index (except for the ES and QES which are both set to match the PWI when they start).  What do we get?

Right.  The PWI drives the result.  Not just that, but both the ERN and ES behaved significantly differently – in real level terms – than the PWI.  And in an issue that annoyed a lot of economists, the switch from ES to QEX saw measured average wages change – and there was no overlap between the two series to actually consider what was happening!!

What would a wage index look like if it followed the ERN, then switched to the ES, then switched to the QEX (assuming the real wage was flat rather than collapsing in a single quarter)?  It would look like the following:

Why ignore the PWI?  What was it measuring – from discussions I’ve been told it is a measure of labour cost, like the labour cost index.  It is supposed to adjust for compositional changes in the types of jobs and work, and if it is like the LCI it would have tried to remove productivity improvements.  What were the other surveys measuring – the actual hourly wage paid by firms.

Why would the real PWI have been falling?  If goods and services prices are rising more quickly than the product wage it implies that the labour share is falling.  In the 1970s the labour share of income had shot up, and this declined back from the late-1970s through the 1980s.  As a result, we would expect a labour cost index to rise more slowly than consumer prices during such a period – and if there is any technological or productivity growth (which there was) that can lead to a situation where real wages are rising while the product wage, or wage paid per unit of output produced, is falling!

Real wages – in terms of peoples ability to buy goods and services from the wage they are paid for a job – are higher than they were in the 1970s.

Labour shares, wage rates, and NZ history

This real wage series fits nicely within the existence of labour income share (LIS) information for New Zealand.  Something was fishy about the 1970s data, the labour income share and real wage data rose so quickly in the early 1970s to the point where I am not sure they are even believable!  Even if they were, the chaos in the late 1970s and early 1980s which then led to reforms shows that whatever was happening was not sustainable.

Add to this concerns about the unemployment data (also shown in the slides).  The unemployment data comes from a period where significant groups were not deemed as unemployed (such as a lot of women who would have been willing to work) and government job schemes (not just state sector employment, which was large, but direct job schemes for those who otherwise wouldn’t be working) were in place which may have allowed the government a mechanism to ensure that unemployed people wouldn’t show up in the statistics – prior to the reforms New Zealand wasn’t short on corruption and the manipulation of data.  We may believe such schemes serve a purpose – but it does change your measure of unemployment!

So what is it, bad data or was it truly a mad economic time – would it upset you if I said it was a bit of both?  These are issues I will write about in the future, I just wanted to put the idea in your head as a starting point 😉