Some Anglo labour markets

Having suggested yesterday that it might be time to think about cutting the OCR, or at least firmly committing to not raising it unless or until core inflation has already risen close to 2 per cent,  I was reflecting a bit on the handful of countries in which the central bank has raised policy interest rates, in particular Canada, the UK, and the United States.

In the UK case, one could almost discount the single increase, which really only reversed the cut put in place in the climate of heightened uncertainty after the Brexit referendum.   But in Canada and the United States there have been several increases –  in Canada, the policy rate is now 1.25 per cent, up from a low of 0.5 per cent, and in the United States, the Federal funds rate target is 1.25 percentage points off the lows.    In Canada’s case, there has even been signs of a sustained increase in core inflation, although in neither country is core inflation yet at target.

One material difference, if one contrasts New Zealand and Australia on the one hand, and the UK, Canada, and the United States on the other, is spare capacity in the labour market.  Since institutional features (labour regulations, welfare entitlements etc) vary from country to country –  affecting the “natural” rate of unemployment – one can’t take much from simple cross-country comparisons of unemployment rates.   But I’d noticed a headline suggesting Canada’s unemployment rate –  at 5.7 per cent –  was the lowest it had been in decades, and wondered how that comparison looked for the other countries.

Current unemployment rate Minimum since 1986
Australia 5.5 4.1
Canada 5.7 5.7
New Zealand 4.6 3.3
United Kingdom 4.2 4.2
United States 4.1 3.9

Like Canada, the UK also now has an unemployment rate that is the lowest in decades (I started the comparison from 1986 when the New Zealand HLFS started).   The United States unemployment rate is getting close to to the multi-decade low.   But in both Australia and New Zealand, the unemployment rates are well above the 30+ years lows.  Perhaps not very surprisingly, core inflation is weak in both countries  – the December quarter data for Australia are out tomorrow, but in September, the trimmed mean inflation rate was 1.8 per cent, against a target midpoint of 2.5 per cent.

Why these five countries?   Mostly, because all five have (a) data going back thirty years or more, and (b) have had floating exchange rates pretty consistently (the UK had three years in the European Monetary System).   Countries that had fixed exchange rates in the past often had bigger fluctuations in their unemployment rates.

Of course, even this comparison could be overly simplistic.  After all, labour market regulation etc can, and does, change over time, as do things like welfare benefit/work test regimes.  But over 30 years, both the New Zealand and Australian labour markets are generally regarded as having had more policy liberalisation than many other advanced countries.  Our minimum wage policy may be a partial exception, although even there we aren’t alone –  the UK, for example, has moved from having no national minimum wage to an increasingly binding (high) one.

And one area suggesting that our “natural” rate of unemployment (or NAIRU) might have been trending down more than in other countries, is the increased participation in the labour force of people 65 and over.  The OECD data only start in 2000, but here is how things have changed.

Labour force participation rate, age 65+
2000 2016
Australia 6.0 12.6
Canada 6.0 13.7
New Zealand 7.7 23.4
United Kingdom 5.3 10.7
United States 12.9 19.3

New Zealand’s participation rate for old people has increased far more than those of these other Anglo countries. And since the unemployment rate for this age group in New Zealand is a mere 1.2 per cent, almost arithmetically a rising share of the labour force made up of an age group with a very low unemployment rate will tend to lower the average unemployment rate, and the NAIRU.    Our NZS system is structured to provide a near-universal modest welfare benefit, but impose no penalty on those who continue to work.   If an old person loses their job, they face less immediate pressure to find a new one (than, say, a 21 year old), and it isn’t surprising then that the unemployment rate for that age group –  a rising share of the labour force –  is so low.

I wouldn’t want to base any strong conclusions on these simple comparisons, but when you hear talk of some other central banks modestly raising interest rates, remember that conditions aren’t the same from the country to country, and that in New Zealand (and Australia) not only is core inflation persistently low, but there is little sign of any intense pressure on capacity in the labour market.

Workers in a fool’s paradise

A couple of months ago I did a post highlighting some little recognised aspects of the New Zealand data on wages and labour income.   They suggested that, given the underlying relatively poor performance of the economy, workers hadn’t done badly at all. I was curious how the latest national accounts data had changed the picture.

The first chart that attracted my interest was the labour income (“compensation of employees”) share of GDP.   The data are only available annually, but they suggested quite a recovery in the labour share of GDP in the 00s, which had been sustained this decade to date.


That was the picture on the previous iteration of data.     Here is the updated version.

COE jan 17

The picture is subtly different, and if anything the labour income share looks to have been shrinking gradually this decade, even if it is still well above where it was in 2001/2 (the historical low).

But the other chart, which I found more striking, was one in which I compared growth in nominal wage rates against growth in nominal GDP per hour worked.   I used the Statistics New Zealand Analytical Unadjusted Labour Cost Index series.  It isn’t widely referred to, but relative to the headline LCI series it is a pure wages series, not one in which SNZ has already tried to adjust for productivity, and relative to the QES, it is much smoother (the way economists typically think of wage-setting behaviour) and produces more sensible and plausible series (some of the problems with the QES were illustrated in the earlier post).

When I did the exercise earlier, on the old data, I found that cumulative wage inflation –  particularly that in the private sector –  had run quite a bit ahead of productivity (GDP per hour worked) since around 2002.    Here is the updated version of the chart.

wages and nom GDP phw jan 18

There is a lot of short-term noise in the series –  and wages last year were somewhat “artificially” boosted by the pay equity settlement – but if the extent to which wages have moved ahead of productivity is less than it was in the previous iteration of the data (GDP has been revised up, and wage rate data are unchanged), the trend I highlighted last year is still there.

In my earlier post, I noted that this chart had been done using GDP itslf, and that to be more strictly accurate I should have taken account of, eg, the 2010 change in GST (which boosted GDP but shouldn’t have affected wages).    Data on indirect taxes and subsidies are only available annually, so here is a smoothed (four quarter moving average) version of the chart, this time comparing wages against nominal GDP per hour worked excluding indirect taxes and subsidies.

wages and nom GDP phw ex taxes and subsides jan 18

What has been going on?   One possibility is that the Analytical Unadjusted wages data are just substantially wrong?   But they are series that have now been published by SNZ for more than 20 years, and I don’t have specific things I can point to suggesting that they are wrong.

If the data are picking up something real, what then might be the story?   Here was what I included in the earlier post.

My explanation is pretty simple: the (real) exchange rate, which stepped up sharply about 15 years ago and has never sustainably come down since.    When the exchange rate is high, firms in the tradables sectors make less money than they otherwise would have done.   The usual counter to that is that the terms of trade have risen.  But the increase in the real exchange rate has been considerably more than the higher terms of trade would warrant, and in any case much of the gains in the terms of trade have come in the form of lower real import prices, rather than higher real export prices.

And why has the exchange rate been so high?  Because the economy has been strongly skewed towards the non-tradables sector which –  by definition –  does not face the test of international competition.  Demand for labour in that sector has been strong, on average, over the last 15 years, and it is the non-tradables sector that has, in effect, set the marginal price for labour.  For those firms, in aggregate, the lack of productivity growth doesn’t matter much –  they pass costs on to customers.  But it matters a lot for tradables sector producers, who have to pay the market price for labour, with no ability to pass those costs on (while the exchange rate puts downward pressure on their overall returns).  Another definition of the real exchange rate is the price of non-tradables relative to those of tradables. Consistent with this sort of story, in per capita terms real tradables sector GDP peaked back in 2004 (levels that is, not growth rates).

It isn’t, to repeat, a story in which labour has done well absolutely.  As I illustrated the other day, over the last five years there has been about 1 per cent real productivity growth in total.  For decades, we’ve been slipping backwards relative to other advanced countries.   But given the weak overall performance, labour doesn’t look to have done too badly.   That isn’t a recommendation for the “economic strategy” the last two governments have pursued.  A climate in which firms don’t find investment attractive –  perhaps especially investment in the internationally-competitive tradables sector –  isn’t likely to be one that conduces to generating sustained high performance and strong medium-term income growth.

And here is the proxy for business investment (total investment less housing and government) as a share of GDP

bus inv jan 18

Despite some of the best terms of trade in decades, business investment has been poor this cycle –  following on from several decades when it has typically been well below that of the median OECD country (despite well above median population growth).  The notion that “investment has been weak in lots of countries”, even to the extent true, should be no consolation: we started so far behind there was (and is) plenty of scope for us to have caught up, not being so affected by financial crises, euro-area ructions, zero lower bounds or whatever.

It is a fool’s paradise model: non-tradables focused businesses (of which there are many) do just fine, supported by continuing rapid population growth, but there isn’t much net investment at all outside those sectors as New Zealand proves to be an increasingly unfavourable place to build and base internationally competitive businesses.  Productivity growth remains weak, perhaps even weakens further.   Wages might well outstrip productivity growth, but in the long-run only sustained productivity growth will support high material living standards here.   It isn’t a model that need end in crisis, but rather in mediocrity.  And New Zealanders could do so much better.


Some labour market statistics that really should be looked into

There was a curious line in the Labour-New Zealand First agreement, under “Economy”.

Review the official measures for unemployment to ensure they accurately reflect the workforce of the 21st century.

I wasn’t (and still am not) clear what the two parties had in mind.  It got some people rather hot and bothered, with suggestions of political interference to get numbers that happened to suit the government of the day.  That interpretation seemed pretty far-fetched.  Plenty of people –  politicians included –  have views on what Statistics New Zealand should collect and report data on.  And governments have to decide what to fund Statistics New Zealand for –  regional nominal GDP data got added to the mix a while ago, there are now weird (and intrusive) things like the General Social Survey, and on the other hand we still don’t have monthly CPI data, monthly industrial production data (in both cases, unlike almost every other advanced country) or quarterly income-based measures of GDP.   Rather rashly, governments and SNZ appear on course to degrade our travel and immigration data.

So I don’t have a problem if parties to a government want to have a look again at some or other area of our official statistics, and perhaps even get Treasury and MBIE to commission some expert or other to have a fresh look at indicators of unemployment etc.  I’d be even more pleased if such a review led to the allocation of a bit more money to Statistics New Zealand.  But I’m not sure there is much of a problem with the HLFS as it is, even if my confidence in the data have taken a bit of a dip since my household has been in the survey (over the last few quarters).   Oh, and when they made changes to the HLFS last year, and made no attempt to backdate the new employment and hours series, simply leaving a level shift in the official series that was a bit trying too (one always has to remember to make a rough and ready adjustment for the break – I almost forgot to in the charts below).

Is it a bit odd and arbitrary that the headline measure of unemployment doesn’t count you as unemployed if you managed one hour’s paid work in the survey week, even if that was the only hour you managed to get all quarter and you’d really like a 40 hour a week job?    Absolutely it is.    But so long as the headline unemployment measures are used either for cross-country comparisons, or for comparisons within New Zealand over time, precisely where one draws that (inevitably) arbitrary line won’t matter very much.  Other countries also calculate headline unemployment rates that way, and we’ve been using the HLFS since 1986.

It is more of a problem when complacent commentators misuse the measure to go on about how “unemployment” is “only” 4.6 per cent, as if all is rosy.   Of course, even a 5 per cent “true” unemployment rate would mean that over a 40 year working life, the typical person would be unemployed –  on the quite narrow definition –  for two years.  That is a large chunk of time, and (like me) probably few of those commentators ever spent any time unemployed on this measure.

But SNZ does now do quite a reasonable job of providing a richer array of data that enables users –  and media and other commentators –  to get a fuller picture of overall supply/demand imbalances in the labour market.  We have data on the people in part-time work who would like to work more hours.  And data on people who would like a job but have become discouraged by repeated failure, and have given up searching (to the definitions of the HLFS).  Outside the HLFS we have data on those on welfare benefits.  Now there is even an official underutilisation rate, which can also be compared across time and (with more difficulty) across countries.   At 11.8 per cent that is a pretty high number, and probably one that –  were it more widely known –  would trouble many people (as it does me).   These numbers tend not to matter much to macroeconomic commentators, focused mostly on cyclical fluctuations, since the various different series tend to move together and a demand for long-term time series drives people quickly back to the headline measure.  But it doesn’t make the other measures less valuable or important for other purposes.

It is meaningless to say that “the” unemployment rate is 4.6 per cent, but that would have been as true in 1997 as it is 2017.  Then again, it probably isn’t meaningless to say that all the measures of excess labor supply are higher than they were 10 years ago, a period over which demographic trends have probably been working to lower the long-run sustainable rate of unemployment (on whichever measure you choose).

Statistics New Zealand don’t seem any better informed about the review

[Labour market manager] Ramsay said Statistics NZ had no more information about the review apart from what was in the coalition agreement.

“Nothing at this point. No content at all.”

But if there are resources to spend on reviewing and improving labour market statistics, I’d be making a bid for something around wages data.

A repeated theme from the Labour Party during the election campaign was that wage growth has been slow, and that this needed to change.  When the Labour Party leader was, at times, challenged about this claim, her response was that people didn’t “feel” better off.    Now, I’m sure perceptions matter a lot in politics, but ideally perceptions –  and the policies of governments – will be informed and shaped by the data, rather than the other way round.

In a post a few months ago I illustrated, using national accounts data, that the labour share of income has been trending up in New Zealand over the last 15 years or so.  COE

Over that period, on official data, New Zealand’s experience has been quite different from that of the other Anglo countries (and much of the commentary we read is British or American).  Across the OECD as a whole, the labour share in the median country hasn’t changed in the last 15 years, and New Zealand has had one of the larger increases. [UPDATE: An interesting illustration of how different the Australian experience has been.]

One of the problems in making sense of what is going on is that (a) we don’t have a quarterly income-based measure of GDP, so we fall back on the published wages data, and (b) the published wages data are all over the place.

Still most widely quoted is the very-volatile Quarterly Employment Survey measure of average hourly wage rates, a measure that (by construction) is subject to compositional changes  (if, this quarter, lots more low-skilled get jobs, even at good wage rates for those jobs, average hourly wage rates will fall even though no one is earning less per hour than they were).

Then there is the Labour Cost Index (LCI) which doesn’t purport to be a series of wage rates, but rather a proxy for unit labour costs. In other words, it is an attempt to measure wages adjusted for changes in productivity etc.  It is a smooth series, and is given prominence by SNZ, but it tells us nothing at all about the growth in the hourly earnings of the people who are in employment (adjusted for changes in composition).

And then there is the Analytical Unadjusted Index.  Even the name would deter most casual users.  It is found buried among the Labour Cost Index series, and  –  at least on paper –  looks like the best series we have.  It is constructed from the raw wages data SNZ collects to generate the headline LCI series, and is constructed in a stratifed way, to eliminate (or minimise) distortions arising from compositional changes.

This is what inflation in the Analytical Unadjusted series looks like

analytical unadj nov 17

It is relatively smooth –  conforming to economists’ priors about how labour markets work –  and, of course, (nominal) wage inflation is much lower it was a decade ago.  (Remember that the tick up in the most recent quarter is the impact of the pay-equity settlement.)    Of course, CPI inflation is also a lot lower than it was then.

A couple of months ago, I did a post using the Analytical Unadjusted data, deflating it by core inflation and comparing it with growth in real GDP per hour worked.  Real wage inflation appeared to have been running well ahead of productivity growth (the latter, non-existent, in aggregate, for the last five years).

But in that chart, I didn’t take account of the terms of trade.  A higher terms of trade – and New Zealand’s have done quite well in the last 15 years or so –  lifts the incomes the economy can afford to pay.  A better way to look at things might be to compare nominal wage growth with growth in nominal GDP per hour worked.  There is a lot of short-term variability in nominal GDP growth –  as dairy and oil prices ebb and flow  – but if we look at cumulative growth over fairly long periods we might hope to find something interesting.  Over very long periods of time we might expect hourly wage rates to increase at around the rate of growth in nominal GDP per hour worked.

The Analytical Unadjusted data go back to mid 1990s for the whole economy, and to the late 1990s for the private sector.   Here is what the resulting chart looks like.  Both series –  wage rates and nominal GDP per hour worked – are indexed to 100 when the Analytical Unadjusted data start.  (Recall that we still only have q2 GDP data).   I’m showing the ratio of the two series: when the line is rising, wage rates are rising faster than nominal GDP per hour worked.

wages and nomina GDP phw an unadj.png

For the first seven or eight years, the chart looks much as you’d expect.    There is quarter to quarter volatility in GDP, which is reflected in the ratio, but broadly wages were rising at around the rate of growth of  nominal GDP per hour worked.  Wages outstripped nominal GDP growth in the late boom years –  even as the terms of trade were rising –  and have done so again, in the last five years.   Over the last 15 years, private sector wage rates –  on this measure –  have risen perhaps 12 per cent faster than growth in the value of nominal GDP per hour worked.  (And the tax switch in 2010 will have boosted nominal GDP, without any reason to expect it would change pre-tax wage rates. so the “true” increases in wages relative to underlying GDP is even larger than the chart suggests).

I find this picture plausible, and I think I can tell a sensible story about what might have been going on.  But before I tell that story, here’s an alternative chart.    The QES wages data go back further, to 1989.  And here is what the chart of QES ordinary time wages rates looks like relative to growth in nominal GDP per hour worked back to 1989.

wages and nom GDP QES

It is on exactly the same scale as the previous chart.  But on this measure, private sector wages have barely kept pace with nominal GDP per hour worked growth over almost 30 years now (and have been losing ground since end of the 1990s), while public sector wage rates have outperformed (but almost all the out-performance was in the 1990s, under those spendthrifts, Ruth Richardson and Jenny Shipley.

I just don’t believe that the QES picture is portraying an accurate picture of what has been going on in the labour market.  For a start, it is inconsistent with the national accounts (the labour income share chart, which suggests that something turned in labour’s favour 15 years or so ago).  And the labour income share chart looks more consistent with the stratifed Analytical Unadjusted based measure.

To be clear, I’m not suggesting that labour has done particularly well.  The productivity performance of the New Zealand economy has been pretty lousy –  especially in the last five years –  and the unexpected (and outside our control) improvement in the terms of trade only offsets a bit of that gap.   Absolute levels of nominal GDP per hour worked in New Zealand remain very low by advanced country standards and, thus, so do wage rates.   But given the relatively poor performance of the economy as a whole, labour hasn’t done badly at all.  If people have feelings about these things it doesn’t look as though they should be about evil capitalists (or evil governments) rapaciously transferring money to themselves or their rich mates.  Simply that poorly performing economies –  with little or no productivity growth –  shouldn’t expect much wage inflation.  If there is rage, it should be about successive governments of both parties that have done nothing to redress that failure.

There might still be some serious problems with the statistics.  But if the Analytical Unadjusted series is roughly right (even if not many commentators cite it), how might one explain what it shows?  My explanation is pretty simple: the (real) exchange rate, which stepped up sharply about 15 years ago and has never sustainably come down since.    When the exchange rate is high, firms in the tradables sectors make less money than they otherwise would have done.   The usual counter to that is that the terms of trade have risen.  But the increase in the real exchange rate has been considerably more than the higher terms of trade would warrant, and in any case much of the gains in the terms of trade have come in the form of lower real import prices, rather than higher real export prices.

And why has the exchange rate been so high?  Because the economy has been strongly skewed towards the non-tradables sector which –  by definition –  does not face the test of international competition.  Demand for labour in that sector has been strong, on average, over the last 15 years, and it is the non-tradables sector that has, in effect, set the marginal price for labour.  For those firms, in aggregate, the lack of productivity growth doesn’t matter much –  they pass costs on to customers.  But it matters a lot for tradables sector producers, who have to pay the market price for labour, with no ability to pass those costs on (while the exchange rate puts downward pressure on their overall returns).  Another definition of the real exchange rate is the price of non-tradables relative to those of tradables. Consistent with this sort of story, in per capita terms real tradables sector GDP peaked back in 2004 (levels that is, not growth rates).

Perhaps it isn’t the correct story. Perhaps there is some serious problem with the data.  But if the government is serious about the words in the Speech from the Throne

A shift is required to create a more productive economy

one (small) step towards getting there might be set out to resolve the puzzles, and apparent inconsistencies, in our labour market (wages) data.  At present though, the best-constructed series suggests a badly-unbalanced economy.  Workers haven’t done badly given the poor performance of the overall economy, but the foundations haven’t been laid for durable real income growth –  if anything, they’ve been progressively whittled away as the foreign trade share of the economy has eroded.





On wages: expectations and reality

Last week, when I was tied up with other stuff, I heard a few media reports that a new Westpac survey was showing that public expectations of wage increases were slipping away.  At the time, I didn’t look at the details, but made a note to come back to it.

This was the key chart included in Westpac’s report of the survey results.

wage expectations

Introduced with this text:

Although workers may be feeling more confident about job opportunities, when it comes to the outlook for earnings, sentiment is really in the dumps. Increasing numbers of workers are telling us that they don’t expect any change in their earnings from work over the coming year. In fact, the number of workers who expect to receive a pay increase over the coming year is languishing at the sort of lows we saw during the financial crisis.

Concluding with this

And while nominal wage growth has remained muted, consumer price inflation has picked up. After lingering below 1% for much of the past few years, consumer price
inflation is now running at 1.7% per annum. This means that the limited pay rises many workers have received have only just been keeping pace with changes in the cost of living. And for those workers who didn’t receive a pay rise (and even for some that did), their spending power may be going backwards.

I’m not really convinced.

I’m not doubting that respondents did answer the question the way Westpac reports. I wouldn’t even be surprised if the recent reversal of wage expectations was the real thing: there was all sorts of talk not long away about wage inflation being just about to “take off”, which so far hasn’t come to anything much.   But even with the recent reversal, expectations are still just back to around where they were for a fair part of 2015 and 2016.

My concern is more about how to interpret the longer-run of data in the chart and, in fact, how to make sense of wage data themselves.

For a start, surely respondents to this survey are inclined to bias their answers downwards?  After all, look at the results for the 2005 to 2007 period, when the labour market was unquestionably tight (including the fact that the unemployment rate was below 4 per cent), and general wage inflation –  on any of the measures –  was quite high.  And yet only around 50 per cent of respondents expected a wage increase.  Many more than that must have been achieving a wage increase.  As I’ve noted previously, the labour share of total income has actually been increasing in New Zealand.

Second, it is worth remembering that inflation expectations now are materially lower than they were a decade ago.

household expecs 2017

The numbers bounce around a bit, but at the end of the previous boom the average year-ahead expectation was around 4.5 per cent, whereas now it around 3 per cent.  (One shouldn’t put much weight on the absolute numbers, but the pattern is consistent with others surveys of inflation expectations.)   If inflation expectations have fallen materially, surely it is reasonable that fewer survey respondents will now be expecting nominal wage increases, even if everything else (labour market tightness, productivity growth or whatever) was unchanged?

Westpac also uses as a reference point a 1.7 per cent rise in annual wages.  That number appears to come from the LCI, a series that purports to adjust for what firms’ report were productivity changes.  It is better to use the “analytical unadjusted” measure from the LCI, which is closer to a stratified raw measure of wage increases –  which is, after all, more like what the respondents in the Westpac survey are being asked about.

Many commentator also focus on the even lower wage inflation numbers from the Quarterly Employment Survey (QES) –  a wage measure that is notoriously volatile (and not really representative of how anyone thinks the labour market is actually working).  It is quite prone to compositional changes, and thus doesn’t reflect – or really try to reflect –  an individual’s own experience in the labour market.

I’ve covered this issue in an earlier post.

I’m not sure why people put so much weight on the QES measure of hourly wage inflation.  It has well-known problems (for these purposes) and is hugely volatile.   Here is a chart showing wage inflation for the private sector according to (a) the QES, and (b) the Labour Cost Index, analytical unadjusted series.

wages debate  No economic analyst thinks wage inflation is anything like as volatile as the blue line –  in fact, wage stickiness, and persistence in wage-setting patterns is one of the features of modern market economies.

And here is the chart I ran last week, comparing real private sector wage inflation (the orange line above, adjusted for the sectoral core measure of CPI inflation) with productivity growth.

Real wage inflation now is lower than it was in the pre-2008 boom years, but it is running well ahead of productivity growth (however one lags or transforms it).

As I noted in that earlier post, real wage inflation in New Zealand has been surprisingly strong in recent years, given the complete absence of any (actual or trend) growth in labour productivity (real GDP per hour worked).  Of course, low inflation and low inflation expectations hold down the nominal rates of wage increases (relative to what we were experiencing a decade ago), but the real measures are largely what matter.   Real household purchasing power from labour income in New Zealand has been increasing –  from increased employment, but also from real wage increases that are more than it is likely that the economy can support in the longer-term.

Perhaps then people are right to expect more modest wage increases ahead.  But if so, it will likely be because the non-tradables led pseudo-boom of the last few years comes to an end, and market processes across the economy force an adjustment in wage-setting to something more consistent with our alarmingly poor productivity growth record (in this particular bad phase now five years and counting).


Productivity and employment

With 30 seconds thought it is pretty obvious that if the least productive 10 per cent of our workforce simply dropped out and stayed home, then across the whole economy average GDP per hour worked would increase, all else equal.   All else equal, the productivity of any particular individual still employed wouldn’t change –  in practice it might well, as someone would still have to do the filing or the cleaning –  but the average would.

So far, so uncontroversial.  No one thinks it would be a sensible policy approach to lifting productivity to, say, bar such low productivity people from working.  Doing so would not only be inhumane, but it would make us, on average, poorer (output is still output, even if productivity of the marginal worker is below average).    In practice, of course, high minimum wages (relative to the market median), as in New Zealand, have exactly that effect –  pricing some low-productivity people (who couldn’t, at present, command a wage in the market at least equal to the statutory minimum.

But every so often in the last 20 years, as people have tried to grapple with New Zealand’s continuing poor average levels of GDP per hour worked, and the failure to achieve any convergence to the (now) richer members of the OECD, someone pops up with line “ah, but we are more effective than most in drawing in the low productivity members of our community, which will bias our measured average productivity (and productivity growth) downwards.

The latest example was in the Sunday Star-Times business section yesterday.

New Zealand’s track record on labour productivity may look worse than it is because a growing number of Kiwis are in work, the Productivity Commission says.

In fact, this wasn’t reporting any new Productivity Commission work.  Rather, one of the Productivity Commission’s senior staff had pointed the journalist in the direction of some interesting work done by able researchers at Motu a couple of years ago.  And, despite the implication readers (like me) may have taken from the headlines and the lead sentence (above), the research work related to a period 2000 to 2012, not to the period of nil productivity growth over the last five years.

It suggested annual productivity growth would have been about 70 per cent higher, averaging 0.24 per cent, between 2001 and 2012, instead of 0.14 per cent, were it not for a decline in skills associated with higher employment.   Motu estimated last year that the skill level of the average Kiwi worker fell by 1.8 per cent over the period as more people joined the workforce.

Again, despite the hyped lead-in (“70 per cent higher”) do note that the difference in these two (multi-factor) productivity growth rates cumulates over 11 years to a total difference of around 1.1 per cent.  Welcome, but not exactly game-changing.

Motu provided a nice non-technical summary  (page 3f) on what they’d actually done, using detailed data from the Longitudinal Business Database (LBD).

Productivity estimates are typically based on the quantity of labour used by firms to produce output. However, the characteristics of a firm’s workers also have an important influence on productivity, with different types of labour impacting differently on the technologies that firms adopt and their performance more generally. Because data on individual workers are linked to the data on firms in the LBD, it is possible to construct a measure of the quality of a firm’s labour force and measure the impact of this on productivity.

The measure of worker quality – which is derived from earnings data – reflects the bundle of skills, qualifications and experience of individual workers. As such, it picks up a broader range of worker attributes beyond qualifications.

Based on this measure, the average quality of the New Zealand work force declined slightly by 1.8% from 2001-2012…..

This somewhat surprising decline in the average quality of New Zealand workers reflects the net result of two opposing forces. First, average skills increased due to ageing (ie, greater experience) and rising qualifications. For example, the share of tertiary qualified workers grew from 15% to 25% while the share of workers with no qualifications fell from 19% to 14% between 2001 and 2013. At the same time, full-time equivalent employment increased strongly by around 15% (Figure 1). The large number of new workers who came into the labour market had, on average, lower skills than existing workers. This lead to a dilution in worker quality that more than offset the improvement in qualifications and experience.

They look like nice results.

But since many of the concerns around productivity growth in New Zealand relate to cross-country comparisons –  how have we done relative to the rest of the advanced world, and relative to common underyling global trends –  it might be worth looking at what has happened in other countries.    It would take a pretty big study to replicate the Motu project across, say, the OECD.   But we do have readily accessible data on employment to population ratios across the OECD, and we have that data for a longer period of time than just 2001 to 2012.

Our HLFS goes back to 1986.  Here is how New Zealand’s employment to population ratio has behaved since 1986.

employment to popn 25 Sep

Over the entire 30 year period, our employment to population ratio increased by 2.4 percentage points, which isn’t a lot.  It seems quite plausible that the effect Motu identified was present in the data as the employment to population ratio increases, from the trough in 1992 through to 2007.  But most of that effect will have been reversing the opposite effects resulting from the really sharp fall in the employment to population ratio (disproportionately low productivity workers, almost by construction) from 1986 to 1992.

And what about the international comparison?  Here is the gap between New Zealand’s employment to population rate and that in the median of the 22 OECD countries for which there is data for the whole period (almost all the “old” advanced OECD countries, and not the former Soviet bloc countries).

employment 2

In all but one year, our employment to population ratio has been above that of the median OECD country.    That doesn’t automatically mean we have been employing more low productivity people –  some systems make labour force participation of both parents of small children easier than others, and some systems penalise older people staying in workforce less than others –  but lets grant that some part of the difference may be that we manage to employ more of the less productive groups.   At the margin, that might explain a small part of the levels difference between our average productivity and that of these, mostly richer, OECD countries.

But two things to note:

  • the gap is smaller now than it was thirty years ago.  In other words, even if this “employing the less productive classes” story is some part of the levels explanation, it is almost certainly less of an explanation than it was 30 years ago.  And yet the real puzzle people have been grappling with is why, after all the reforms, we haven’t made any progress in closing the gaps over the last 30 years.   These compositioneffects don’t look as though they can help over the post-1984 period as a whole (useful as they might be for interpreting data for some individual sub-periods).
  • there has been no material change in the gap at all over the last decade, suggesting that this compositional story doesn’t offer any explanation for why from 2008 to 2015 we did no better than middling relative to other OECD countries (not closing the gaps), and since 2012 we’ve been among the very worst productivity performers, with no labour productivity growth at all.

As I’ve pointed out in several posts recently, average real GDP per hour worked in Germany, Netherlands and France is now around 60 per cent higher than that in New Zealand (even though historically all were poorer and less productive than New Zealand).  In 2016, employment to population ratios in New Zealand and Germany were identical (while those in Netherlands and France were lower).  But here is the chart showing New Zealand’s employment to population ratio less the average of the ratios of each of those three countries.

employment 3.png

Over the period for which observers have been struggling for an explanation of our poor productivity growth, our employment to population ratios have been falling relative to those in several of the leading, and most productive, European economies.

Compositional effects (around the skill levels of the labour force) just don’t look like a credible part of an explanation for why the level of productivity here is now so much below that in the leading OECD economies, or why no progress has been made in closing the gap, over the last 30 years or the last five.


Employment growth: simply not that spectacular

There was another post on Kiwiblog this morning, attempting to cast New Zealand’s recent economic performance in a particularly good light.   Here was the bit that really caught my eye:

this is not just exceptional job growth locally, but internationally. Here’s the percentage increase in in major OECD countries in 2016:

  1. NZ 5.7%
  2. Germany 2.9%
  3. Ireland 2.9%
  4. US 1.8%
  5. OECD 1.6%
  6. Australia 1.6%
  7. Sweden 1.5%
  8. UK 1.4%
  9. Canada 0.7%
  10. France 0.6%
  11. Finland 0.5%

Now there are at least three problems with this comparison:

  • it makes no allowance for the much more rapid rate of population growth in New Zealand than in almost any other OECD country,
  • it cherry-picks the OECD countries it compares us with (I’m not sure when Ireland and Finland became “major” OECD countries), and
  • it ignores the break in the HLFS hours worked and employment series in 2016q2.  In fairness, the author might not have been aware of the break, but serious economic analysts (including the Treasury) are.

I illustrated the break in the series in a post several months ago.

What about the rate of job growth.  Fortunately, we have two measures: the (currently hard-to-read) HLFS household survey measure of numbers of people employed, and the QES (partial) survey of employers asking how many jobs are filled.   Unsurprisingly, the trend in the two series are usually pretty similar, even if there is a fair bit of quarter to quarter volatility.


Since we know there are problems in the HLFS, and the QES doesn’t look to be doing something odd, perhaps we are safest in assuming that the number of jobs has been growing at an annual rate of around 2.5 to 3 per cent.   That isn’t bad at all. But SNZ also estimates that the working age population has been growing at around 2.7 per cent per annum.  No wonder the unemployment rate is only inching down.

Now that we have 2017q2 data, so a full year on the new HLFS questions, the annual percentage growth rates of the two employment series have indeed converged again.

hlfs and qes E

In other words, one can’t take as meaningful any annual percentage growth in the HLFS employment (or hours) numbers for calendar 2016.

A better way to deal with all three issues is to look at the percentage point change in the employment to population ratio for the whole OECD group.   The most recent period for which we have full data for all countries is 2017q1.  For New Zealand, using growth in employment over the year to 2017q1 would still be distorted by the break in the series, so for New Zealand only I’ve shown the change in the employment to population ratio from 2016q2 to 2017q2.

E to popn last year

And on this – much more useful – comparison, New Zealand ends up as a middling performer, the median country.   There is no stellar New Zealand “job creating machine”, just a huge increase in working age population.     Job growth isn’t to be gainsaid, but it is productivity growth (or the absence of it) that is the key determinant of gains in medium-term living standards.  And did I mention that there had been no productivity growth, at all, for the last five years now?

(To be clear, I would not put much –  if any –  weight on a single year comparison.  After all, all labour force surveys have some sampling error.  But if people want to make sense of employment growth, in international comparison, over just the most recent year, this is really the only sensible way to do it.  As it happens, over that year, our change in the employment to population ratio was the same as that for the OECD as a whole.  It was just a bit less than that for the EU as a whole and for the euro-area –  who, of course, generally had a deeper unemployment hole to climb out of.)

Labour share of income

The other day I ran this chart showing how the labour share of income (“compensation of employees” in national-accounts-speak) had changed in New Zealand over recent decades.    COE

It isn’t data I usually pay any attention to, and I was somewhat surprised by the trend increase since around 2002.

And then I was reading a Financial Times article about last weekend’s Jackson Hole retreat for central bankers (perhaps including Graeme Wheeler) and assorted other eminent people.   The journalist mentioned that one prominent Asian central banker had warned that a declining labour share of income around the world could make problems for central bankers (the idea being that workers  –  especially low income ones – tend to spend most of their income, and demand shortfalls are a potentially serious issue, especially when the next recession happens).    And that left me wondering just how unusual New Zealand’s experience –  a rising labour share –  had been.

So I downloaded the OECD data back to 1970.    They have data for 25 advanced countries for the entire period (the exceptions are mostly the former eastern bloc countries).  Here is the share of GDP accounted for by compensation of employees for the median of those countries.


Slightly higher at the end of the period than at the start, and not very much change overall for the last thirty years.

And here is how the labour share has changed in the individual countries since 1970.

lab share since 1970

The median change is basically zero, but what is striking is how diverse the experiences of these advanced countries have been.  There are easy explanations for some of them –  Ireland’s change, for example, will reflect the company tax structure, which has encouraged (a) a lot of foreign investment, but (b) a lot of effort by multinationals to book profits in Ireland.  For Ireland, the labour share of GNI would be more enlightenning.  But for most of these countries the GNI/GDP gap is small, and yet there are still huge differences in the experience.

New Zealand –  like all the Anglo countries –  is towards the left of that chart.  But what about the experience since 2001, when the labour share of income troughed in New Zealand?  For that period, there is data for almost all OECD countries, not just the 25 in the earlier charts.

lab share since 2001

Over this period, not only is New Zealand near the right of the chart, but our experience has been quite different from that of the other Anglo countries.

This just isn’t my field, and I’m not pushing any particular interpretation of these data.  I simply found them interesting, and a little surprising.  But if they data are broadly correct, they do suggest that whether over 45 years or over the last 15, the overall labour share of income in advanced countries hasn’t changed much.  Of course, in most countries, productivity growth has been a lot slower than it was in the glory days of the post WW2 decades, and thus real wage growth will have been slower.  But the overall labour share hasn’t changed much –  and the differences across countries are much larger than the differences across (this period of) time for the advanced world as a whole.

These data also don’t shed any light on the inequality narrative.   Labour as a whole may have held its share of overall income, and yet differences in pre-tax market labour incomes may have become more pronounced (eg increases in chief executive salaries in the US or UK relative to median wages, or the rise of an extremely highly-remunerated subset of financial markets employees).     But if there are trends there, they haven’t been mirrored in a shrinking share of the cake going to labour as a whole.  Indeed, in New Zealand the labour share of income has increased quite a bit in the last 15 years or so (concentrated in the boom years of the 2000s, but not reversed since then).

And finally, another curiosity I stumbled on.   There is quite a sense that New Zealand’s labour market functions reasonably well and  –  in conjunction with counter-cyclical macro policy – delivers us unemployment rates that have been relatively low by OECD standards.    And I think that is probably not a totally inaccurate story –  who’d trade our labour market for that of Spain or Italy?

But here is a comparison of unemployment rates of Mexico and New Zealand

mexico U

Over the 25 years for which there is data for both countries, in only one –  at the height of the Mexican financial crisis –  did Mexico have an unemployment rate even slightly higher than that of New Zealand.   And if people suspect (as I do) that our long-run sustainable unemployment rate is getting down to around 4 per cent now, experience suggests that in Mexico it has been that low for decades.

Why mention Mexico?  Mostly because, despite its advantages –  oil, proximity to the United States, coasts on two oceans –  Mexico is a  relatively poor and (absolutely) not very productive OECD country.  Data are a bit patchy, but best indications are that Mexico’s productivity performance over the last 50 years has been even worse than that of New Zealand.    And yet, whatever the reasons, they’ve managed a system with consistently less unemployment than New Zealand has had (and, actually, even their prime age male participation rate is higher than that in New Zealand –  as perhaps one might expect in a materially poorer country).