Unemployment: ethnic differences

Having written last Friday on some of the differences in unemployment rates by age cohort, I got curious about the HLFS data broken out by ethnicity.  I’ve never paid much attention to it –  much of the data hasn’t been published for long, and my main interest has always been macroeconomics (the whole economy, rather than specific outcomes for particular subgroups).

But many of the ethnic differences are stark.   Here, I will mostly focus on those between those identifying as European and those identifying as Maori.

Take the headline unemployment rates for example (the longest run of data I could find on Infoshare).

U by eth

or the underutilisation rates

underutil by eth

The European numbers are bad enough –  10.5 per cent of the labour force underutilised –  but the Maori numbers are astonishing.  22.4 per cent of the Maori labour force underutilised is a sad reflection of (probably) a whole series of failures.    Perhaps because the numbers are so large, the gap between the Maori underutilisation rate now and that prior to the recession is visibly stark.

As I illustrated last week, labour market characteristics differ quite a lot by age cohort, and the Maori population is, on average, quite a bit younger than the the European population (both because of higher birth rates and because of lower life expectancy).

But even when one looks by age cohort, the differences between Maori and European outcomes are stark.   Overall labour force participation rate of Maori and Europeans aren’t so very different –  on average over the last three years, 67.2 per cent of working age Maori were in the labour force, and 70.2 per cent of Europeans.   But here are participation rates by age cohort for the two ethnic groups.

partic by age eth

In every cohort, except the 65 plus group, the European labour force participation rate is materially above that for Maori –  on average by almost 10 percentage points.   I’m not sure what to make of the 65 plus group, where Maori and European participation rates are almost equal.   It may be, at least in part, a reflection of greater Maori relative poverty (less success in building up wealth over earlier years).

The picture is even more stark if one looks just at employment rates (given that Maori unemployment rates are higher).

empl rates by age eth

Focusing in on young people, here is what the data show people aged 15 to 24 are doing (again averaged over the last three years).

15 to 24

Interestingly, a slightly larger proportion of the the Maori young population are in education and not simultaneously in the labour force than is the case for young Europeans.     Perhaps the most visible difference on that chart is the first set of bars –  the much larger proportion of European young people simultaneously working and studying than for Maori.      Here is a chart (over the same three years) of that one component for each of the four ethnic groups SNZ reports the results for.

15 to 24 employed

I did notice the difference in the “not in education, not in the labour force –  caregiving” proportions.  Here is that chart for each of the four ethnic groups.

15-24 caregiving

SNZ reports no (statistically significant) numbers of young men in this category.  In other words,  around 11 per cent of young Maori women are engaged in full-time caregiving (presumably mostly for young children), doing no study or even an hour’s paid work (the HLFS criterion) a week.   By contrast no (statistically significant numbers of) young Asian women are.

I’m not going to attempt to hypothesise about why the differences in these various charts (or various others in the data) exist.  But none of the gaps strike me as things simply to be relaxed about.   Says he who is –  contentedly – not in the labour force, not in education, and is caregiving children.

Unemployment: age matters

Despite the reaction of the foreign exchange market, there didn’t seem to be much new in the suite of labour market data the other day.  Sure, employment was down a touch, and the participation rate fell back.  Then again, it had been hard to take entirely seriously the reported strength of the participation rate over the last year or so.  And the unemployment rate did keep edging downwards (although at this rate it will take another two years until the unemployment rate is back to where it was when the current government took office) and hours worked, in both the HLFS and QES, grew quite strongly.   And there wasn’t much sign of any pick-up in wage inflation.  So unless you had been determined to believe the data was  just about to confirm an overheating economy, probably not too many surprises.  We’ll see next week what the Governor of the Reserve Bank makes of it.

But, largely prompted by a question in Parliament yesterday, I downloaded the age breakdowns in the HLFS.  Unemployment rates are very very different by age cohort.

Here are official unemployment rates for the latest year, by age.

U rates by age

People aged 15-24 and people aged 65 and over make up roughly the same share of the total working age population (about 18 per cent each).    But people in the young cohort make up 45 per cent of the total number of people unemployed, while over 65s make up 1.5 per cent of the unemployed.

And that difference isn’t something new.  Here are the average shares of the total number of people unemployed, by age cohort, for the entire history of the HLFS.

U by age avg

The age patterns shouldn’t be terribly surprising.    At one end of the spectrum,  young people are often in part-time work, dropping in and out of jobs (voluntarily or not), in transition between school, work, and further study, even in the course of a single year.  And, of course, they are just starting out –  finding out what they might be good at, or might enjoy, and proving themselves (or not) to employers.   High minimum wages (relative to median wages) hit younger people harder than other age groups.

Every OECD country has an unemployment rate for 15-24 year olds materially higher than the overall unemployment rate –   although as I was confirming that I was a little surprised to discover that although our overall unemployment rate has been consistently below that of the median OECD country, our unemployment rate for young people is now as high as that in the median OECD country.

U 15 to 24

At the other end of the age spectrum, by contrast, work is typically much more of a choice.  Almost everyone living here aged 65 or over is entitled to NZS, the state pension.  If old people are working, they probably aren’t changing jobs very often –  their circumstances are rarely changing as rapidly –  and if for some reason they lose their job, they may not be particularly aggressive in looking for a new job. Recall that the official definition of unemployment involves actively searching for work and being available to start right away.  With a stable income buffer, that search of active search is likely to be less imperative for most than it will be for young people just starting out.

It is relatively easy to understand why there are quite high unemployment rates for young people and quite low ones for old people.  But these differences can matter for how we think about the overall unemployment rate if the importance of different age cohorts in the labour market has been changing over time.  It has.

age shares of lab force In the early 90s 5 per cent of old people were in the labour force (working or actively seeking), and now that figure is almost 25 per cent.  By contrast, the labour force participation rate for 15-24 year olds is now only around 63 per cent (it was 74 per cent when the HLFS began).    (Frankly, the drop in youth participation rates surprises me a bit given that (eg) university fees are much much higher than they used to be, and that one has to work only an hour a week to be counted as employed so, for example, after school jobs should be captured.)

The change in the age structure of the work force does then affect how we think about any particular rate of unemployment.   The natural rate of unemployment for young people –  normal frictional stuff –  is materially higher than that for old people, so that a 5 per cent aggregate unemployment rate means something different than a 5 per cent unemployment rate does today.

One way of illustrating the point is shown in this chart.  It shows the actual reported unemployment rate, and also an artificial unemployment on the assumption that each age cohort had the same unemployment rate is actually had, but that the relative size of the various age cohorts was the same as it was in 1987 (the first year for which there is data).

U rate age adj

For the last year, the actual unemployment rate was 5 per cent.  With constant age chort shares, it would have been 6.2 per cent.    It isn’t an effect that makes much difference from year to year, but over time it can materially affect how we look at any particular unemployment rate.    In effect, and all else, the change in composition of the labour force –  more old people, fewer young people, –  appears to have lowered the NAIRU.

As one final chart, here is the change in the unemployment rate (in percentage points) for each age cohort, from the year to June 2008 (the previous cyclical low point) to now (year to June 2017).

U chg by age

The overall unemployment rate is still 1.4 percentage points higher than it was then.   There has been almost no change in the unemployment rate for the over 65s –  which isn’t surprising as, for reasons outlined above, there is very little cyclicality in that series. On the other hand, it is quite sobering how large an increase in the unemployment rate for 15 to 24 year olds there has still been.   Some of that is cyclical, and some will likely reflects the effects of higher minimum wages, but whatever the cause it should be cause for disquiet, given how important it is to get a start in the labour market and to stay connected.



A radical alternative to macro policy?

Last Friday, an outfit called Strategy2040 New Zealand, together with Victoria University’s School of Government, hosted a lunchtime address by an Australian academic, Professor Bill Mitchell of the University of Newcastle.   He is a proponent of something calling itself Modern Monetary Theory, but which is perhaps better thought of as old-school fiscal practice, with rhetoric and work schemes thrown into the mix.

Mitchell attracted some interest on his trip to New Zealand.  He apparently did two substantial interviews on Radio New Zealand and attracted perhaps 150 people to the lunchtime address –  a pretty left-liberal crowd mostly, to judge from the murmurs of approval each time he inveighed against the “neo-liberals”.    In fact, the presence of former Prime Minister Jim Bolger was noted –  he who, without apparently recanting any specific reforms his government had put in place, now believes that “neo-liberalism has failed New Zealand”.     Following the open lecture, 20 or so invitees (academics, journalists and economists –  mostly of a fairly leftish persuasion) joined Mitchell for a roundtable discussion of his ideas.   Perhaps a little surprisingly, I didn’t recognise anyone from The Treasury or the Reserve Bank at either event.

Mitchell has it in for mainstream academic economics.   Quite probably there is something in what he says about that.  Between the sort of internal incentives (“groupthink”) that shape any discipline, and the inevitable simplifications that teaching and textbooks require, it seems highly likely there is room for improvement.   If textbooks are, for example, really still teaching the money multiplier as the dominant approach to money, so much the worse for them.   But as I pointed out to him, that was his problem (as an academic working among academics): I wasn’t aware of any floating exchange rate central banks that worked on any basis other than that, for the banking system as a whole, credit and deposits are created simultaneously.  He quoted the Bank of England to that effect: I matched him with the Reserve Bank of New Zealand.    And if very few people correctly diagnosed what was going on just prior to the financial crises in some countries in 2008/09, that should be a little troubling.  But it doesn’t shed much (any, I would argue) light on the best regular approach to macroeconomic management and cyclical stabilisation.  Perhaps especially so as (to us) he was talking about policy in Australia and New Zealand, and neither country had a US-style financial crisis.

He seemed to regard his key insight as being that in an economy with a fiat currency, there is no technical limit to how much governments can spend.  They can simply print (or –  since he doesn’t like that word – create) the money, by spending funded from Reserve Bank credit.     But he isn’t as crazy as that might sound. He isn’t, for example, a Social Crediter.    First, he is obviously technically correct –  it is simply the flipside of the line you hear all the time from conventional economists, that a government with a fiat currency need never default on its domestic currency debt.     And he isn’t arguing for a world of no taxes and all money-creating spending.  In fact, with his political cards on the table, I’m pretty sure he’d be arguing for higher taxes than New Zealand or Australia currently have (but quite a lot more spending).  Taxes make space for the spending priorities (claims over real resources) of politicians.  And he isn ‘t even arguing for a much higher inflation rate –  although I doubt he ever have signed up for a 2 per cent inflation target in the first place.

In listening to him, and challenging him in the course of the roundtable discussion, it seemed that what his argument boiled down to was two things:

  • monetary policy isn’t a very effective tool, and fiscal policy should be favoured as a stabilisation policy lever,
  • that involuntary unemployment (or indeed underemployment) is a societal scandal, that can quite readily be fixed through some combination of the general (increased aggregate demand), and the specific (a government job guarantee programme).

Views about monetary policy come and go.   As he notes, in much academic thinking for much of the post-war period, a big role was seen for fiscal policy in cyclical stabilisation.  It was never anywhere near that dominant in practice –  check out the use of credit restrictions or (in New Zealand) playing around with exchange controls or import licenses –  but in the literature it was once very important, and then passed almost completely out of fashion.  For the last 30+ years, monetary policy has been seen as most appropriate, and effective, cyclical stabilisation tool.  And one could, and did, note that in the Great Depression it was monetary action –  devaluing or going off gold, often rather belatedly – that was critical to various countries’ economic revivals.

In many countries, the 2008/09 recession challenged the exclusive assignment of stabilisation responsibilities to monetary policy.  It did so for a simple reason –  conventional monetary policy largely ran out of room in most countries when policy interest rates got to around zero.   Some see a big role for quantitative easing in such a world.  Like Mitchell – although for different reasons –  I doubt that.    Standard theory allows for a possible, perhaps quite large, role for stimulatory fiscal policy when interest rates can’t be cut any further.

But, of course, in neither New Zealand nor Australia did interest rates get anywhere near zero in the 2008/09 period, and they haven’t done so since.    Monetary policy could have been  –  could be –  used more aggressively, but wasn’t.

As exhibit A in his argument for a much more aggresive use of fiscal policy was the Kevin Rudd stimulus packages put in place in Australia in 2008/09.   According to Mitchell, this was why New Zealand had a nasty damaging recession and Australia didn’t.  Perhaps he just didn’t have time to elaborate, but citing the Australian Treasury as evidence of the vital importance of fiscal policy –  when they were the key advocates of the policy –  isn’t very convincing.   And I’ve illustrated previously how, by chance more than anything else, New Zealand and Australian fiscal policies were reamrkably similar during that period.   And although unemployment is one of his key concerns –  in many respects rightly I think –  he never mentioned that Australia’s unemployment rate rose quite considerably during the 2008/09 episode (in which Australian national income fell quite considerably, even if the volume of stuff produced –  GDP –  didn’t).

On the basis of what he presented on Friday, it is difficult to tell how different macro policy would look in either country if he was given charge.   He didn’t say so, but the logic of what he said would be to remove operational autonomy from the Reserve Bank, and have macroeconomic stabilisation policy conducted by the Minister of Finance, using whichever tools looked best at the time.  As a model it isn’t without precedent –  it is more or less how New Zealand, Australia, the UK (and various other countries) operated in the 1950s and 1960s.  It isn’t necessarily disastrous either.  But in many ways, it also isn’t terribly radical either.

Mitchell claimed to be committed to keeping inflation in check, and only wanting to use fiscal policy to boost demand where there are underemployed resources.    And he was quite explicit that the full employment he was talking about wasn’t necessarily a world of zero (private) unemployment  –  he said it might be 2 per cent unemployment, or even 4 per cent unemployment.     He sees a tight nexus between unemployment and inflation, at least under the current system  (at one point he argued that monetary policy had played little or no role in getting inflation down in the 1980s and 1990s, it was all down the unemployment.  I bit my tongue and forebore from asking “and who do you think it was that generated the unemployment?” –  sure some of it was about microeconomic resource reallocation and restructuring, but much it was about monetary policy).   But as I noted, in the both the 1990s growth phase and the 2000s growth phase, inflation had begun to pick up quite a bit, and by late in the 2000s boom, fiscal policy was being run in a quite expansionary way.

I came away from his presentation with a sense that he has a burning passion for people to have jobs when they want them, and a recognition that involuntary unemployment can be a searing and soul-destroying experience (as well as corroding human capital).  And, as he sees things, all too many of the political and elites don’t share  that view –  perhaps don’t even care much.

In that respect, I largely share his view.

Nonetheless, it was all a bit puzzling.  On the one hand, he stressed how important it was that people have the dignity of work, and that children grow up seeing parents getting up and going out to work.   But then, when he talked about New Zealand and Australia, he talked about labour underutilisation rates (unemployment rate plus people wanting more work, or people wanting a job but not quite meeting the narrow definition of actively seeking and available now to start work).   That rate for New Zealand at present is apparently 12.7 per cent –  Australia’s is higher again.     Those should be, constantly, sobering numbers: one in eight people.      But some of them are people who are already working –  part-time –  but would like more hours.  That isn’t a great situation, but it is very different from having no role, no job, at all.  And many of the unemployed haven’t been unemployed for very long.  As even Mitchell noted, in a market economy, some people will always be between jobs, and not too bothered by the fact.  Others will have been out of work for months, or even years.   But in New Zealand those numbers are relatively small: only around a quarter of the people captured as unemployed in the HLFS have been out of work for more than six months (that is around 1.5 per cent of the labour force).       We should never trivialise the difficulties of someone on a modest income being out of work for even a few months, but it is a very different thing from someone who has simply never had paid employment.  In our sort of country, if that was one’s worry one might look first to problems with the design of the welfare system.

Mitchell’s solution seemed to have two (related) strands:

  • more real purchases of good and services by government, increasing demand more generally.  He argues that fiscal policy offers a much more certain demand effect than monetary policy, and to the extent that is true it applies only when the government is purchasing directly (the effects of transfers or tax changes are no more certain than the effects of changing interest rates), and
  • a job guarantee.    Under the job guarantee, every working age adult would be entitled to full-time work, at a minimum wage (or sometimes, a living wage) doing “work of public benefit”.     I want to focus on this aspect of what he is talking about.

It might sound good, but the more one thinks about it the more deeply wrongheaded it seems.

One senior official present in the discussions attempted to argue that New Zealand was so close to full employment that there would be almost no takers for such an offer.   That seems simply seriously wrong.    Not only do we have 5 per cent of the labour force officially unemployed, but we have many others in the “underutilisation category”, all of whom would presumably welcome more money.     Perhaps there are a few malingerers among them, but the minimum wage –  let alone “the living wage” – is well above standard welfare benefit rates.   There would be plenty of takers.   (In fact, under some conceptions of the job guarantee, the guaranteed work would apparently replace income support from the current welfare system.)

But what was a bit puzzling was the nature of this work of public benefit.    It all risked sounding dangerously like the New Zealand approach to unemployment in the 1930s, in which support was available for people, but only if they would take up public works jobs.  Or the PEP schemes of the late 1970s.   Mitchell responded that it couldn’t just be “digging holes and filling them in again”.  But if it is to be “meaningful” work, it presumably also won’t all be able to involve picking up litter, or carving out roadways with nothing more advanced than shovels.  Modern jobs typically involve capital (machines, buildings, computers etc) –  it accompanies labour to enable us to earn reasonable incomes –  and putting in place the capital for all these workers will relatively quickly put pressure on real resources (ie boosting inflation).   If the work isn’t “meaningful”, where is the alleged “dignity of work”  –  people know artificial job creation schemes when they see them –  and if the work is meaningful, why would people want to come off these government jobs to take existing low wage jobs in the prviate market?

The motivation seems good, perhaps even noble.  I find quite deeply troubling the apparent indifference of policymakers to the inability of too many people to get work.   The idea of the dignity of work is real, and so too is the way in which people use starting jobs to establish a track record in the labour market, enabling them to move onto better jobs.

But do we really need all the infrastructure of a job guarantee scheme?  In countries where interest rates are still well above zero, give monetary policy more of a chance, and use it more aggressively.   For all his scepticism about monetary policy, it was noticeable that in Mitchell’s talks he gave very little (or no) weight to the expansionary possibilities of exchange rate.    But in a small open economy, a lower exchange rate is, over time, a significant source of boost to demand, activity, and employment.    And winding back high minimum wage rates for people starting out might also be a step in the right direction.

And curiously, when he was pushed Mitchell talked in terms of fiscal deficits averaging around 2 per cent of GDP.  I don’t see the case in New Zealand –  where monetary policy still has capacity –  but equally I couldn’t get too excited about average deficits at that level (in an economy with nominal GDP growth averaging perhaps 4 per cent).  Then again, it simply can’t be the answer either.    Most OECD countries –  including the UK, US and Australia –  have been running deficits at least that large for some time.

It is interesting to ponder why there has been such reluctance to use fiscal policy more aggressively in countries near the zero bound.   Some of it probably is the point Mitchell touches on –  a false belief that somehow countries were near to exhausting technical limits of what they could spend/borrow.      But much of it was probably also some mix of bad forecasts –  advisers who kept believing demand would rebound more strongly than it would –  and questionable assertions from central bankers about eg the potency of QE.

But I suspect it is rather more than that –  issues that Mitchell simply didn’t grapple with.  For example, even if there is a place for more government spending on goods and services in some severe recessions, how do we (citizens) rein in that enthusiasm once the tough times pass?  And perhaps I might support the government spending on my projects, but not on yours.  And perhaps confidence in Western governments has drifted so low that big fiscal programmes are just seen to open up avenues for corruption and incompetent execution, corporate welfare and more opportunities for politicians once they leave public life.  Perhaps too, publics just don’t believe the story, and would (a) vote to reverse such policies, and (b) would save themselves, in a way that might largely offset the effects of increased spending.      They are all real world considerations that reform advocates need to grapple with –  it isn’t enough to simply assert (correctly) that a government with its own currency can never run out of money.

I don’t have much doubt that in the right circumstances expansionary fiscal policy can make a real difference: see, for example, the experience of countries like ours during World War Two.    A shared enemy, a fight for survival, and a willingness to subsume differences for a time makes a great deal of difference –  even if, in many respects, it comes at longer term costs.

But unlike Mitchell, I still think monetary policy is, and should be, better placed to do the cyclical stabilisation role.    That makes it vital that policymakers finally take steps to deal with the near-zero lower bound soon, or we will be left in the next recession with (a) no real options but fiscal policy, and (b) lots of real world constraints on the use of fiscal policy.  Like Mitchell, I think involuntary unemployment (or underemployment for that matter) is something that gets too little attention –  commands too little empathy –  from those holding the commanding heights of our system.  But I suspect that some mix of a more aggressive use of monetary policy, and welfare and labour market reforms that make it easier for people to get into work in the private economy,  are the rather better way to start tackling the issue.   How we can, or why we would, be content with one in twenty of our fellow citizens being unable to get work, despite actively looking –  or why we are relaxed that so many more, not meeting those narrow definitions, can’t get the volume of work they’d like  –  is beyond me.   Work is the path to a whole bunch of better family and social outcomes –  one reason I’m so opposed to UBI schemes –  and against that backdrop the indifference to the plight of the unemployed (or underemployed), largely across the political spectrum, is pretty deeply troubling.

But, whatever the rightness of his passion, I’m pretty sure Mitchell’s prescription isn’t the answer.




A questionable indicator of the labour market (geeky)

The Reserve Bank has long been averse to too much focus on the unemployment rate.  Some of that was political.  Opposition MPs back in the 1990s would try gotcha games, trying to extract estimates of a non-accelerating -inflation rate of unemployment (NAIRU) with the aim of then being able to tar us with lines like “Reserve Bank insists full employment is x per cent unemployment” [at the time x might have been 7 per cent or more] or “Reserve Bank insists on keeping x per cent of New Zealanders unemployed”.   So there was an aversion to using the concept, and most certainly to writing it down.  If you don’t write things down, it is hard to have them OIAed.

But for decades there has been something a lot like a NAIRU embedded in successive versions of the models the Bank uses as a key input to the forecasting process.   But there was still an unease.  Some of it was those old political concerns.  Some of it was the aversion to being pressured into any sort of dual objective model –  even though discretionary monetary policy was only ever introduced to allow short-term macro stabilisation together with medium price stability.

In the last decade or so, it seemed to be some mix of things.   In 2007, the unemployment rate was down at around 3.5 per cent, but the official view at the time was that the overall economy was more or less in balance (as I noted yesterday the output gap estimates then were positive but quite small).  So, the unemployment rate tended to be discounted.   To the extent there were NAIRU estimates implicit in the thinking, they had been trending down for 15 years.  Perhaps, some felt, 3.5 per cent unemployment wasn’t much below a NAIRU at all.

In the years after the 2008/09 recession, there seemed to be two problems.  The first was a quite genuine one.  The HLFS unemployment rate numbers were quite volatile for a while, and while it was clear that the unemployment rate was still quite high, it was hard to have much confidence in each new quarterly observation.    The rise in the unemployment rate in 2012 only further undermined confidence among some of my then colleagues.  The economy seemed to be recovering.

U 2006 to 13

So there had always been a tendency to discount the unemployment rate.  Odd short term developments in the series itself reinforced that, and then as the recovery began to develop the Bank kept convincing itself that the output gap had all but closed.  If the output gap  –  which tended to be central focus for much of the analysis, even though it was something of a black-box, and prone to significant revisions –  had really closed, then surely the labour market must also be more or less in balance?  If the unemployment rate appeared to suggest otherwise, so much the worse for that particular indicator, which many at the Bank had never much liked.    When the Bank started the ill-fated tightening cycle in March 2014, they thought then that excess capacity had already been used up for a couple of years.  This is the output gap chart they ran then.

output gap mar 14 mps

At the time, the unemployment rate was still 5.6 per cent

(Contrast that output gap chart with the one from yesterday’s MPS, which I included in yesterday’s post.  At first glance, they look quite similar, but whereas in 2014 they thought excess capacity had been exhausted in 2011/12, now they think it was only exhausted in 2013/14.)

Various people had various ideas for how best to look at labour market data.  But mostly the unemployment rate was just ignored.

A year or two back, some US researchers had done an interesting exercise, trying to combine formally the information in a whole variety of labour market indicators, to distill an overall picture.  And somewhere along the line, someone got the idea of doing something similar for New Zealand.  And thus was born the Reserve Bank’s Labour Utilisation Composite Index, with the pleasing acronym of LUCI.

One day last April, Reserve Bank Deputy Governor Geoff Bascand gave a major address on monetary policy matters, Inflation pressures through the lens of the labour marketAs I noted at the time, it was little odd for him to be giving the speech –  as his day job was mostly responsibility for notes and coins, and the Bank’s corporate functions.  Then again, his ambition for higher office was pretty apparent, and earlier in his career he had led the Department of Labour’s Labour Market Policy Group.   Geoff on labour market matters should have been worth listening to.

The broad thrust of the speech was that high immigration wasn’t going to put much pressure on demand (contrary to the usual experience, and past Bank research), and that the labour market was pretty much operating at full capacity.  In support of these propositions, the Bank simultaneously released not just the speech, but three new Analytical Notes.    Since I had been harrying the Bank about its change of view of immigration, I focused then on the two immigration research papers, and identified a number of issues with them.     At the end of a long post I noted

There is a more material on other topics in Bascand’s speech, and another whole Analytical Note on other labour market issues which I haven’t read yet. I might come back to them next week.

But I never did.  Other things presumably distracted me, and thus I never got round to reading the Analytical Note introducing LUCI.   I didn’t until yesterday afternoon – I’d heard the Governor, or the Chief Economist, mention it at the press conference, and it was a wet afternoon, so I went and downloaded the paper.

The Reserve Bank has given quite a lot of attention to this brand new indicator.  In the Deputy Governor’s speech, it gets three whole paragraphs

Over the business cycle, a key driver of wage growth is the balance of supply and demand, or labour market ‘slack’. However, the unemployment rate is an inadequate indicator of labour market slack, particularly when the participation rate fluctuates. Researchers at the Bank have recently constructed a labour utilisation composite index, or LUCI, to help address this problem. Such indices combine the information in a large number of labour market variables into a single series of labour market tightness, and are used internationally to help gauge labour market pressures. The New Zealand index uses official statistics such as the HLFS and survey measures of the difficulty of finding labour, such as the QSBO.

By construction, the LUCI has an average value of zero. A LUCI value above zero indicates greater labour market tightness than usual – a value below zero indicates greater labour market slack than usual. Our research shows that, historically, a higher LUCI has been associated with stronger wage growth.

The LUCI suggests there was a large degree of slack in the labour market at the trough of the 2008-09 recession. The LUCI then gradually returned to zero, and has been around that level since early 2014 (figure 7). This movement is consistent with the range of the Bank’s suite of output gap indicators.

This wasn’t just some speculative new tool dreamed up down in the engine room by smart researchers, but still needing road-testing.  The Deputy Governor (presumably with the approval of the Governor and the Chief Economist) seemed sure it was action-ready.

Consistent with that approach, a chart of LUCI has appeared in all but one of the five MPSs since that speech.    Here was LUCI as she appeared yesterday

LUCI may 2017

On the Deputy Governor’s telling

A LUCI value above zero indicates greater labour market tightness than usual – a value below zero indicates greater labour market slack than usual.

In other words, on that particular interpretation, labour market tightness was now more severe than it had been in 2007 (just prior to the recession), and almost as severe as at the peak of the series in around 2004.

And, of course, Geoff Bascand didn’t just make up that interpretation.  It was what the researchers had said in their Analytical Note, almost word for word.   It was also what the footnotes on the LUCI charts in the MPSs said.  At least until yesterday.     In yesterday’s MPS, slipped in so quietly, the interpretation of LUCI had changed rather materially.   The chart was still there, it still looked the same as ever.  But below it, it has these words written

Note: a positive value indicates a tightening in labour market conditions

Lay readers may not immediately notice the difference, but it is substantial.  When the indicator was launched, with some fanfare (there are many clever new indicators, and not many get three paragraphs in a Deputy Governor speech), a positive value was interpreted as the labour market being tighter than usual (thus the pesky single indicator, the unemployment rate could keep on being largely ignored).   But now a positive value simply means that the labour market is a bit tighter than it was last quarter.    If there was lots of slack last quarter, it simply doesn’t matter to the indicator –  all that matters is the direction of change.

The Bank didn’t draw this to readers’ attention at all.  They didn’t change the heading on the chart either.  It was a bit naughty really.       What it means is that whereas we once had an indicator –  using 16 different variables –  that might have suggested the unemployment was completely misleading as a measure of slack, what we actually have now in LUCI saying similar things to the unemployment rate.  The unemployment has been edging down, and LUCI has been positive.   But that LUCI number is now felt to tell us nothing at all about the absolute level of labour market slack (or even the level relative to a long run of history).

I don’t quite know how the Bank fell into this –  except, perhaps, for the wish (for a labour market measure suggesting the market was at full capacity –  in line with the output gap estimates) getting the better of hard-headed challenge and scrutiny.

The approach they used to construct LUCI was very much the same used to construct the sectoral factor model of core inflation.  In essence (there), get the rates of price increases for all the CPI components, and look for a common theme (or “factor”) that runs through them all.    It seems to work quite well there.  All the variables are expressed the same way (percentage changes).      But LUCI is a bit different.     Here is the list of all the variables used.

LUCI components

They take each variable and standardise it relative to its own mean.

But it is an odd mix of variables.   There are annual percentage changes mixed in with levels.  And some of those levels measures (eg the QSBO ones) are the outcome of questions that are actually technically expressed in change terms.  It is also a mix of quantity measures and price (wage) measures.   When I looked through the list, the first real oddity that struck me was that unemployment itself is not expressed as a percentage rate.  Rather, they take the number of people unemployed and calculate the percentage change.   If, then, every single person was fully employed –  so the labour market was very tight –  this component  (used in calculating LUCI)  would be showing a zero percentage change.  They’ve done the same thing to all the quantity variables, so it should have been obvious from the start that what they were doing (at least for that half of the inputs) was looking at variables that would produce a change indicator, not a levels one.

I understand why they did it.  The unemployment rate trended down for 15 years or so.   The gap between the actual unemployment rate and the average unemployment rate wouldn’t be a very meaningful indicator.   But it meant, almost inevitably, that the new indicator couldn’t be an indicator of absolute labour market tightness, only (at best) an indicator of changes in tightness.

But there are other problems.  Think about what would happen if (to be deliberately extreme) the population doubled, and yet there was no change in “true” labour market tightness.    That isn’t far from the Reserve Bank’s story about the recent immigration shock –  they told us again yesterday that, contrary to past common experience, they think the demand and supply effects have more or less offset each other.

But if the population doubled, and there was no change in “true” labour market slack,  we would still expect to see employment, numbers unemployed, the number of short-term unemployed, numbers underemployed, the number of filled jobs, hours worked, the working age population, the number of job ads, and the number of registered job seekers all increase (that is roughly half the variables in LUCI).   The unemployment rate (%), for example, might stay the same, but both the numerator and the denominator would increase a lot.  Since LUCI uses annual percentage changes, it seems highly likely that such a shock would show up as a big increase in LUCI in the year the population shock happened  (in my extreme examples, the APCs would go to 100 per cent that year), even if there had been little or no actual change in labour market tightness.

It seems astonishing that, for a variable launched by a Deputy Governor in a speech playing down the net demand effects on a large immigration shock, this issue never seems to have occurred to them.

There are other, probably more minor and mundane, problems too.  These ones  probably weren’t foreseeable in April last year, but probably still should have been highlighted as they became apparent.  The revised HLFS was introduced in June last year.  It is pretty clear that the modifications to the questions have led to material step changes in the HLFS measure of hours worked and of employment (you can see the anomalies in the charts in this post from earlier in the week).   (For some reason, they don’t even use the QES measure of hours in LUCI).     Perhaps a distortion of this sort to only two of the component variables won’t have affected the common factor that the model is trying to identify.  But we don’t know.  I hope the Reserve Bank does.   This particular problem will wash out of the data in the next HLFS release (since a year will have passed, and LUCI uses annual percentage changes), but for now it is another reason not to have much confidence in LUCI as any sort of indicator of labour market tightness (level or changes).

I did put some of these points to the Reserve Bank yesterday, mostly just to check that my understanding of the technical points was correct. I had a helpful response this afternoon which essentially establishes that they are (obviously I’m not associating the Bank with the interpretations I’ve put on those technical points).

Overall, it isn’t a case of the Reserve Bank at its best.  I have no problem at all with them doing the research in the first place.  We should always be looking for new or better ways to understand what is going on, and how best to combine sometimes conflicting bits of data.      But it doesn’t have the feel of something that was at all well road-tested before being launched as a major indicator variable.  And then, when they did finally realise that it was much more like an indicator of changes in labour market pressure rather than the level of pressure, that recognition was sneaked through without even an explanatory note, leaving anyone who had taken the earlier use of LUCI at face value none the wiser unless they were a particularly assiduous reader (and I’m usually not, when it comes to changes in footnotes on charts).

We now have a recognition that LUCI isn’t a measure of overall pressure on the labour market.  It may be, loosely speaking, an indicator of changes in labour market pressure, but even then that reading is made more difficult when you get large population shocks (of the sort that NZ is prone to, with quite variable large immigration flows, in a way that many other countries are not), and when the Reserve Bank repeatedly assures us that its overall interpretation is that this particular population shock isn’t putting additional net pressure on demand, and may even be easing capacity pressure in the labour market.

We really should expect better from our central bank.  Speeches like Bascand’s, and documents like the MPS, will have had heavy involvement from all the members of the Governing Committee –  the Governor himself, the incoming acting Governor, the incoming Deputy Chief Executive, and the long-serving Chief Economist.   One can’t help thinking that they’d have been better served taking the unemployment rate –  actually designed directly as a measure of excesss capacity –  and wage developments rather more at face value.   And of recognising that, contrary to LUCI, no serious observer thought that the labour market was at its tightest in 2004.   It isn’t much harder than that.

Reflecting on the macro data

The Reserve Bank’s Monetary Policy Statement (Graeme Wheeler’s second to last) will be out on Thursday.  I’m not in the market economists’ game of trying to tell you what the Bank will do and say (although no one expects they will do anything concrete with the OCR this time).  I’m more interested in questions around what they should do.  In time, what they should do, they usually will do.  But sometimes not until they’ve tried the alternatives.

I wrote about last month’s CPI data a few weeks ago, concluding that there had been a welcome, and expected, increase in core inflation (it is what typically happens if inflation is below target and the OCR is cut fairly substantially) but that

With the unemployment rate still above estimates of the NAIRU, and most indicators of inflation suggesting that core is probably (a) still below target, and (b) not picking up very rapidly, it certainly isn’t time for hawkish talk about near-term OCR increases.

Not everyone agrees of course.  I noticed the BNZ’s economic commentary yesterday which opened with this confident assertion

There is no excuse for the cash rate to be just 1.75% in New Zealand.

I don’t think I’m unduly caricaturing their record to say that, for at least the last decade, the BNZ economics team has never seen an OCR increase they didn’t like, even –  or perhaps especially –  those which had to be quickly reversed.  But mindful that in the story of the boy who cried wolf, the wolf eventually did come, I thought it was worth having a look at the latest wave of data.  Last week, we got the full quarterly set of labour market data (HLFS, QES, and LCI), and the Reserve Bank’s quarterly expectations survey.  To cut a long story short, it doesn’t alter my view.

Take the expectations survey first.   The headline story was one in which the two year ahead expectations of the inflation rate (of a sample of moderately informed observers –  including me) rose quite materially, and now stand at 2.17 per cent (up from around 1.65 per cent in each quarter last year).

infl and expecs

This measure of expectations isn’t typically very volatile, but it is typically somewhat responsive to changes in headline CPI inflation.  We’ve just had quite a large change in headline inflation, so some increase in the expectations measure shouldn’t be surprising. It certainly shouldn’t be concerning.  After all, ideally, the Reserve Bank wants people to believe, and act as if they believe, that on average over time CPI inflation will average around 2 per cent –  the mid-point of the target range, and the explicit focus of the current (but about to expire) PTA.

In fact, no one really knows whether this survey measure captures how people actually think and behave in real transactions in the goods, labour and financial markets.   It might be as good a proxy as we have, but (a) we don’t know, and (b) it still might not be good at all.  Glancing at the time series, there is a tendency for falls and rise to be at least partly reversed quite quickly.

But if inflation expectations are really in some sort of 2 to 2.2 per cent range, I’d welcome that.  With repeated increase in tobacco excises –  not some underlying economic process –  there is a reasonable case, in terms of the PTA, that headline inflation should average a little higher than the mid-point, and than “true” core inflation.  Only if inflation expectations were to rise further from here might I start to get a little disquieted.

In trying to make sense of the inflation expectations numbers, one thing I haven’t seen mentioned is the Labour Party’s monetary policy release.   There was a quite a bit of focus last month on their pledge to add some sort of employment objective to the Reserve Bank Act, and concerned expressed in some quarters that that could lead to higher inflation over time.   If it was a factor, you’d presumably have to take the probability of Labour leading a new government (call it a coin toss at present?) and multiply that by the probability that the change in regime (and perhaps the sort of people a new government might appoint) would make a material difference over time.  I have no evidence one way or the other, but it wouldn’t surprise me if there was a small effect of this sort.   (My own two year ahead expectation in the survey was 1.5 per cent –  around the current rate of inflation in the Bank’s preferred sectoral factor model).

Not many commentators seem to pay much attention to the rest of the expectations survey, even though its strength is partly the range of macro questions that are asked (although I’ve suggested some modifications to the Bank in their review of the survey).

Take GDP for example. There is no sign of respondents expecting real growth to accelerate.  Two years out they expect annual real GDP growth of 2.6 per cent – down on the previous quarter, but not far from the average response over the last couple of years.    But the survey also asks for quarterly GDP predictions for the next couple of quarters, and year-ahead predictions.   That enables one to derive an implied six monthly growth rate for the second half of the coming year.  Here is the gap between the expected growth rates for the first six months and the second six months, going back to just prior to the 2008/09 recession.

expec GDP growthAs we headed into the recession there was a lot of expectation of a strong rebound.  Even up to around 2012, respondents expected growth to accelerate.   For the last few years they haven’t expected any acceleration, and now the expect it to slow.  To be specific, respondents expect 1.6 per cent total growth in the first half of this year, slowing to 1.2 per cent in the second half of this year.     We don’t know quite why –  perhaps they expect immigration numbers to slow –  but it doesn’t speak of a sense that things are getting away on the Reserve Bank.   Similarly, two years out respondents expected that the unemployment rate would still be 4.9 per cent.

Perhaps these respondents will be proved wrong –  they often are, forecasting is like that –  but at the moment it doesn’t look like an imminent risk of core inflation rising much further, or to levels that might prove problematic for a flexible inflation targeter focused on medium-term inflation outcomes around 2 per cent.

What of the actual labour market data?   We have some problems at present because of the breaks in various HLFS series that occurred when the revised survey questions were put in place last year.  I’m still staggered they could have made these changes without running the two sets of questions in parallel for perhaps a year, to allow robust adjustments to be made for the discontinuities.   HLFS hours worked measures, employment measures, and probably participation rate measures all seem to have been affected to some extent.   We are pretty safe in saying that the number of people employed in New Zealand did not grow by 5.7 per cent last year (as the HLFS suggests).

What of the simplest headline number, the unemployment rate?   There isn’t much doubt that the unemployment rate has been falling over the last few years.  It is what one should expect after a serious recession, and with the stimulus to demand provided by low interest rates and large migration inflows (given that immigration typically adds more to demand in the short-term than it does to supply, thus tending to lower unemployment and use up spare resources in the whole economy).

But what should be somewhat disconcerting is that the unemployment rate has (a) gone largely nowhere in the last year, and (b) is still well above pre-recession levels (unlike the situation in many other advanced countries with their own monetary policies).   In the prevous boom, the unemployment rate got down to around 4.9 per cent as early as the start of 2003.     The picture isn’t much different if one looks at the broader (not seasonally adjusted) SNZ underutilisation measure.

U and under U

There still appears to be some progress in using up spare capacity in the labour market, but not very much at all.

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.

One can do a similar picture for the annual growth rates in the two (HLFS and QES) hours worked series.

hours qes and hlfs

It was pretty clear that there was around a 2 per cent lift in HLFS hours worked from last June, just on account of the new survey questions.  It seems safer to assume that total hours worked across the economy might have grown by around 3 per cent in the last year.   That is faster than the growth in the working age population, pointing to some increase in effective utilisation, but not a dramatic one.  For what it is worth, in the latest releases, the two hours measures were both quite weak in the March quarter.

(And remember that nothing in the expectations survey data suggested pressures were likely to intensify from here.)

And what of wages?    There is a variety of measures.  The QES measure is quite volatile –  there are issues of changing composition –  and I don’t put much weight on it.  But for what it is worth, average hourly earnings rose 1.6 per cent in the last year on this measure, around the lowest rate of increase seen for decades.    The Labour Cost Index measures should get more focus (but have some challenges of their own).

lci inflation 2Perhaps there is some sign of a possible pick-up in the analytical unadjusted series (which doesn’t try to correct – inadequately –  for productivity changes) but it is a moderately volatile series, and the most recent rate of increase is still below the peak in the last little apparent pick-up a year or two back.

A common response is “ah, but what about the lags?”.  But as we’ve shown, there is little sign of any material tightening occurring in the overall labour market, no sign of expectations that that is about to change, and so little reason to expect much different wage inflation outcomes over the next couple of years from what we’ve seen in the last couple.  At best, there might be some slight pick-up in wage inflation (especially if the increase in inflation expectations is real), but any pick-up is going to be from rates of increase that have, over the last couple of years, been consistent with disconcertingly low rates of core inflation.

So where does it all leave me?  Mostly content that an OCR around 1.75 per cent now is broadly consistent with core inflation not falling further, and perhaps continuing to settle back where it should be –  around 2 per cent.   Of course, there is a huge range of imponderables, domestic and foreign, so no one should be very confident of anything much beyond that.   But it is worth bearing in mind that the unexpectedly strong net migration over the last few years has been a significant source of stimulus to overall domestic demand (including demand for labour).  In the face of typically too-tight monetary policy, it is part of why the unemployment rate finally started gradually coming down again after 2012.

Whatever happens to the cyclical state of the Australian economy, the National government is already putting in place immigration policy changes that should be expected to lead to some reduction in the net inflow of non-citizens, and two of the main opposition parties are campaigning on promises of much sharper reductions than that.   If such policy changes come to pass then, all else equal, the OCR will need to be set lower than otherwise.  It isn’t something that Graeme Wheeler can or should actively factor into this week’s OCR decision, but it may well be something the acting Governor needs to think hard about (if any decisions he makes are in fact lawful) after the election.

The underperforming New Zealand economy

Today’s labour market data seem to point again to the underperformance of the New Zealand economy.  Oh, the headline rates of GDP growth haven’t looked too bad –  although they are quite modest in comparison with previous New Zealand growth cycles –  and employment growth has been strong.  But to what end?  Labour productivity looks still to be shockingly weak, yet another year ends with the unemployment rate well above Treasury estimates of NAIRU, and even as core inflation has picked up somewhat (yesterday’s post) wage inflation seems to be about as subdued as ever.  There seems to be something quite wrong with the economic strategy that presides over such outcomes –  and no sign from the major opposition parties that they have anything materially better or different to offer.

Hours worked, as captured in the HLFS, have increased strongly in the last five quarters, up by 6 per cent (adjusting for the break in the series, because of new methodology in the June quarter last year).  There have only been a couple of periods in the 30 year history of the series that have seen growth in hours worked that rapid.

We don’t have GDP data for the December quarter yet, and of course earlier quarters are always subject to revision.  But for the four quarters we do have, real GDP (averaging expenditure and production measures) rose by 4.0 per cent.   In other words, unless quarterly GDP growth for the December quarter is at least 1.9 per cent, we’ll again have had no productivity growth at all during that five quarters of rapid growth in hours worked.  Few commentators I’ve seen think GDP growth was anything like that strong  –  something a bit over 1 per cent seems closer to expectations.  If so, we’ll have had really rapid increases in hours worked and employment, but the economy will have got less productive at the same time.  (And recall that we’ve now had five years of no productivity growth).

In the past, periods when growth in hours worked have been very strong haven’t always seen rapid productivity growth.  There can be good reasons for that, if (on average) lower productivity workers are being reabsorbed into employment for example.  In the early-mid 1990s we had a couple of years of very rapid growth in hours worked, and over that period productivity growth, although positive, was pretty weak.     But over that couple of years the unemployment rate fell from around 10 per cent to around 6 per cent, and the employment rate also rose by around 4 percentage points.

Here is the unemployment rate (four quarter moving average to smooth through some of the quarterly noise, down and up)

u-rate-dec-16  The unemployment is still slowly trending downwards, but the pace is quite excruciatingly slow.  Over the five years in which there has been no productivity growth, the average unemployment rate has fallen from around 6 per cent to around 5 per cent, and over that period Treasury estimates that the natural rate of unemployment (determined by things like demographics, welfare provisions and labour market regulation) has been falling –  and is now around 4 per cent.

So we’ve had:

  • no productivity growth (perhaps even a contraction over the last year)
  • high and only slowly falling unemployment (and for those inclined to glibly respond that 5 per cent unemployment isn’t high, recall that that numbers mean that any one time one in 20 of those people available for wanting, wanting to work and making active efforts to find work can’t find a job).

And what of wage increases?  Unsurprisingly perhaps, there has been little sign of any recovery nominal wage inflation.    A standard response is that wages will inevitably lag improvements in the labour market, but….the unemployment rate has now been falling slowly for five years or so.

There is a variety of different wage inflation measures.   Here are two from the Labour Cost Index –  both the headline published series, which tries to adjust for productivity growth, and the Analytical Unadjusted index which is more like a raw measure of wage inflation.   In both cases, I’ve shown the data for the private sector.


Of course, if one believes this data (in particular the red line) there must have been some continuing productivity growth in New Zealand, even if at a slower rate than previously.  Quite why SNZ finds (implied) productivity growth here, and not in national accounts (real GDP per hour worked) is a bit of a mystery.

The other measure of wage increases if from the QES. In this case, the annual rate of increase in private sector ordinary time hourly wages.


There is some volatility in this series, and I’m not sure I’d want to put much at all on the reported sharp fall-off in hourly wage inflation over the last year, but…….there is certainly no sign of an increase in wage inflation.

It is always easy to look around and find countries that have done worse than New Zealand –  several of the euro area countries spring readily to mind.    But our performance, and the gains for our people, are nothing much to celebrate.  And while, for example, there has been a global slowdown in productivity growth since the mid 2000s, New Zealand’s productivity levels are so far below those of the more strongly performing OECD countries, that there was no necessary reason why we needed to share in the slowdown.  It should, if anything, have been an opportunity for some convergence.  But there has been no sign at all of that.

I don’t find that particularly surprising –  an economic strategy that appears to involve attracting ever more people to one of the most isolated corners on earth, in an era in which connections, contacts, and proximity seem to matter more than ever, all while producing a very high real exchange rate (again resurgent in recent weeks/months), and the highest real interest rates in the advanced world, is simply a recipe for continued long-term underperformance.  One would like to think that the government –  and the Opposition which seems to support very similar policies –  has been surprised. They can’t, surely, have planned on such a bad performance.  But persistent bad outcomes, of the sort New Zealand continues to see, should be prompting some serious policy rethinks, not just more PR about how rapidly employment numbers are growing.


Weak productivity growth: can composition effects explain it?

One of the charts I’ve run a few times in the last few months has had a bit of extra coverage in the last few days.


It is a pretty straightforward chart (although it would be a little easier if SNZ followed the practice of the ABS and reported the series routinely, rather than leaving it for people to calculate).  I simply averaged the expenditure and production measures of real GDP, and divided the results by the total number of hours worked (from the HLFS).  And real GDP per hour worked itself is a pretty standard measure of labour productivity.

The interest, of course, has been in the now four years or so of no growth in labour productivity.  On the face of it, it is a pretty poor performance and tends to act as something of a counterpoint to a focus by the government (and its business and media cheerleaders) on headline GDP numbers –  which are high largely because the population has been growing so rapidly, rather than because resources are being used more productively.  Productivity is, in the long run, almost everything when it comes to improving material living standards.

I would add a few caveats around the chart, some of which I’ve made here before.  The first is that the very final observation should be heavily discounted or ignored.  SNZ introduced a revised HLFS methodology in June, which has resulted in a step up in the number of hours recorded (perhaps by around 1 to 1.5 percentage points).  At some point that might be reflected in a slightly higher level of GDP, but for the moment there is just an inconsistency.  (And, of course, there is always some quarter to quarter volatility in the data too.)

The second caveat is the old warning that when a number looks particularly interesting it might well be wrong.  Four years of no productivity growth at all is not unprecedented here or abroad (on current data, for example, GDP per hour worked in the UK now is only around 2007 levels) but….these series are prone to revision, and while they could be revised either up or down, it shouldn’t greatly surprise us if the picture for 2012 to 2016 looks a bit different when we review the data a few years hence.

The big revisions tend to happen as a result of the annual national accounts.  Statistics New Zealand gets a lot more detailed data, produces full annual data once a year (including revisions to earlier years), and then updates the quarterly series that have already been published.  The annual data for the year to March 2016 are out later this month, and the revised quarterlies will presumably be available with the September quarter GDP release next month.  Expect changes (including in the chart above).

But for now, the data is as it is.  Bernard Hickey gave the chart some prominence, with the editorial comment “We’re just pumping more low wage workers in the economy and working more hours”, and observing “Jobs soaked up by net migration & more >65 yrs working”

That prompted Eric Crampton of the New Zealand Initiative, writing on his Offsetting Behaviour blog,  to produce a post asking whether compositional changes in the labour force might account for some or all of the weak productivity growth in recent years.  As he quite rightly notes, if a lot of very unskilled people started working lots more hours (in total), while higher skilled people worked the same number of hours, at the same real output, average real GDP per hour worked would fall even though no one individually was less productive.

First, Eric noted that the number of people on welfare benefits has fallen quite a bit over the last few years.  If –  as seems reasonable –  those people had been of below average productivity, that might tend to lower overall productivity somewhat.

But here is my problem with that story.


Working age beneficiary numbers have certainly fallen over the last few years, but they rose a lot during the recession.  There is seasonality in the data so I’ve only shown one observation per annum (June), but in June this year the share of the population of working age on welfare benefits was almost exactly equal to the share as the recession was getting underway in 2008.  People moving on and off benefits might affect average labour productivity to some extent, but absent any sign of an upward surge in productivity over, say, 2008 to 2010 it is difficult to believe this effect explains much of the recent absence of productivity growth.  (And, of course, the decline in beneficiary numbers doesn’t appear to have been in the faster than in the five years leading up to 2008).

Eric also includes a graph showing changing employment rates for different age cohorts, observing

The youngest workers are least productive. They hugely dropped out of the labour market with the changes to the youth minimum wage, but that decline’s since reversed a bit. There’s been a long trend growth in hours worked among older workers, but typical wage patterns over the lifecycle have wages flattening out from the early 50s or thereabouts. Big increases in employment rates among cohorts with lower than average productivity, or at points in the life cycle where wage profiles (and presumably productivity) flatten out, will both flatten or worsen GDP per hour worked.

What to make of that?  Here is a chart of the changes in employment rates for each age group, both since 2012 (when productivity seems to have gone sideways) and since 2007, just before the last recession –  and it isn’t a misprint/error; we’ve had no change in the employment rate over the full period from 2007 to 2016.


Over the last four years, the least productive age group (15 to 24) had the largest increase in employment rates,  and the 65+ employment rate has kept on growing quite a bit.  But….employment rates for 25 to 44 years olds increased quite a lot too (more than the over 65s).    And if we take the full period (Sept 07 to Sept 16), we’ve had a big drop in the employment rate for the lowest productivity age group.  That fall was, of course, concentrated in the first half of the period, but there was no obvious corresponding surge in average productivity at that time (granting that one never knows the c0unterfactual).

And by New Zealand standards, there is nothing very obviously unusual going in the 65+ employment rates.  Between 2007 and 2016 the 65+ employment rate rose by 8.7 percentage points. In the previous nine years, it has risen by 8.1 percentage points.

Perhaps one could dig deeper (if the data existed) and the impression might change, but it isn’t obvious that the changing age composition of the workforce can explain four years of no labour productivity growth.

Sometimes people suggest that perhaps our labour market is performing so much better than those of other advanced countries which might in turn explain the poor productivity growth.   But here is a chart showing employment rates for New Zealand, Australia, and the median OECD country.


There might be something in the story relative to Australia over the last few years.  But comparing New Zealand with the OECD median, our employment rate fell about as much as that median did during the recession, and has rebounded only slightly more since.  Compare New Zealand and the typical employment rate just prior to the recession and almost half of OECD countries have had more of an increase than (the slight rise) New Zealand has had.

Eric also suggests that we need to think about the role of immigration

And, obviously, net migration’s increased over the last few years. New workers getting settled in New Zealand might take a bit to find their feet as well, while still being better off than they were before.

Just two thoughts.  First, around half the huge swing upwards in net inward migration has been the result of the sharp decline in the number of New Zealanders leaving.  They won’t have taken “a bit of time to find their feet”.    Second, for the other migrants, there might be something to the story (although there hasn’t been much variability in the number of actual residence approvals) through, for example, the increased number of foreign students working and people on working holiday visas.  But….the New Zealand Initiative and other business lobby groups can’t really have it both ways. They often tell us it is imperative that we have the sort of immigration policy we have now, because (for example) New Zealanders just can’t, or won’t, do the work (at a price firms can afford). There is a strong hint in that sort of argumentation that immigrants are on average actually quite highly productive relative to natives (even though the data show that for most immigrant groups it can take decades for the earnings to reach those of similarly qualified, similarly experienced New Zealanders).

I wouldn’t rule out the possibility that the compositional effects resulting from immigration are part of the explanation for the latest productivity slowdown (although we didn’t see something similar when Australia had its huge surge) but….if the Initiative is right about the general economic payoff to high immigration, we should be expecting a pretty big lift in average labour productivity (the more so to make up for four years of no growth) really quite soon.

One other lens on the composition issue is offered by our own official annual productivity data (for the “measured sector” rather than for the whole economy).   SNZ produces both labour productivity and multi-factor productivity estimates, and they also produce both series using both total hours worked and an estimate that attempts to adjust for the changing composition of the labour force.   The latter isn’t precise by any means, and won’t pick up all the sorts of issues that have been touched on in this post, or Eric’s, but they are just another angle on the question.  The MFP numbers are valuable because they help get round the question of whether, say, labour productivity is just poor because firms have substituted away from capital towards abundant labour.  Any such substitution would be less troubling if the result was showing strong MFP growth.

Unfortunately, the most recent data are for the year to March 2015.  In the labour productivity data, SNZ weren’t detecting any sign that a worsening average quality of the labour force was explaining the productivity slowdown – they reported much the same improvement in the average quality of the labour force as in earlier years.  And here is the MFP chart.


On this measure, of labour-quality adjusted MFP, there has been no productivity growth at all since around 2006.    There is some modest growth over 2012 to 2015 (a bit over 1 per cent over three years).

Where does all this leave us?

I remain a bit uneasy about the prospects the data could be revised, but then data revisions are always a risk.  But if the average real GDP per hour worked data are roughly right – and there really has been no average labour productivity growth for perhaps four years now –  I think we should be more inclined to believe that it is telling us something about overall economic underperformance, than that it is simply, or even largely, reflecting compositional changes in the labour force. To repeat:

  • the share of working age welfare benefit recipients has fallen gradually over the last few years, but then it rose in the previous few years, and there was no obvious associated productivity surge,
  • over the last few years the employment rates of the low productivity young age groups have risen, but not noticeably faster than those for, say, the rather large 25 to 44 age group.  Over 65s employment rates are rising more than those for other age groups, but that change has been underway for many years.  There was no obvious associated productivity surge (at least in the reported data) when youth employment rates dropped sharply.
  • there is nothing in cross-country comparative data suggesting employment rate changes here have been unusual, in ways that might help account for unusually weak productivity growth here.
  • compositional effects resulting from increased immigration of non-citizens (especially  working students and working holidaymakers) could be part of the story (averaging down real GDP per hour worked, even if no one individually is less productive), although it would be worth testing that story against other episodes in other countries.  If higher immigration is playing a role in dampening productivity growth, I suspect it isn’t mostly a compositional story, but one about overall pressures on domestic resources, which have contributed to holding up real interest rates (relative to those in other countries) and the real exchange rate.
  • and overall MFP growth –  whether SNZ estimated for the measured sector, with some labour composition effects accounted for, or the Conference Board’s estimates that I showed the other day – also seems to have been weak to non-existent.