The Treasury’s economic scenarios

The Treasury this morning released a report to the Minister of Finance offering several scenarios for how the economy might develop over the next few years.  Before getting into the substance, there are a few points worth noting:

  • these are not forecasts, but are best seen as conditional projections (mostly on unchanged macro policy –  fiscal and monetary –  and then various different assumptions about the extent of the (a) the policy restrictions, and (b) the state of the world economy,
  • note that “unchanged fiscal policy” here includes the fact that the current wage subsidy scheme expires in June (in fact, much of what will be paid has already been paid, since it is a lump-sum scheme from government to firms),
  • two of the scenarios allow for some fairly significant additional government spending,
  • none of the scenarios seems to explore different assumptions about the how the private sector responds (all the variation arises from government anti-virus measures only),
  • none of the scenarios really seems to explore the economic implications of a mitigation strategy
  • with one exception, the scenarios all seem to have real GDP back to around the previous path (HYEFU 2019 forecasts) by the year to June 2024.

These are the five basic scenarios, all with unchanged macro policy.

tsy covid 1

I noted in my post this morning that it was relatively easy to get to forecasts of a 20-25 per cent fall in GDP in the June quarter even on assumptions that the restrictions are wound back relatively promptly after next week, and that I was surprised how (relatively) modest many of the falls forecasts by bank economists had been.   I was, therefore, pleased to find that on The Treasury’s scenario 1 they project that GDP would fall by 25 per cent in the June quarter.  They are perhaps a bit less pessimistic than I am on how much smaller the economy is under the current Level 4 –  they assume 40 per cent while my stab in the dark is 50 per cent –  but relative to many of those other private sector forecasters I was encouraged to see where they had got to.  In this scenario, Treasury projects that GDP would be back to 90 per cent of normal by the September quarter.   That seems excessively optimistic (when overseas tourism alone accounts for 5.5 per cent of GDP).  On their telling, alert level 1 –  borders still closed, world economy still pretty deeply depressed – only results in a loss of New Zealand GDP, relative to normal, of “5-10%”.  Perhaps.

I reckon The Treasury is generally understating the scale of the world economic losses.  But Scenario 5 deals with that in part –  scenario 1 locally, but with a deeper and more protracted world economic downturn.  The 9 percentage point hit to GDP growth in calendar 2020 and 4 percentage point hit in calendar 2021 seems more in the ballpark of what Treasury expects to see for New Zealand (across the range of scenarios).    In scenario 5, however and somewhat remarkably, the unemployment is only 10.5 per cent by the June quarter of next year despite no material further policy support/stimulus.  That would be a surprise.

Unsurprisingly, the very worst economic outcomes are under Scenario 3 –  six months of “Level 4” and six more months of “Level 3”.  On that scenario, the unemployment rate is 22 per cent a year from now, with GDP having fallen another 23.5 per cent in the full year to June 2021 relative to the June year 2020.

As I said earlier, Treasury does not explore differences in assumptions about how the private sector respond. And at this stage they don’t seem to provide any information on how quickly they’ve assumed foreign travel (inward and outward) would revive.  The document is what it is, but I hope they will before long offer some more detail as to how they are thinking about such issues/risks.

The paper is written up in a way designed to make the government and the Reserve Bank pretty happy.   As I’ve said, the five basic scenarios assume no further macro policy support.   That is without precedent in a post-war New Zealand recession, when monetary policy would usually be eased very very substantially, and the additional stimulus that easing provides is part of what gets the economy back towards full employment as fast as possible.  In these scenarios the exchange rate does not fall far at all .  And the Treasury makes no mention of the Reserve Bank’s refusal to cut the OCR further or deal with the physical cash related constraint on just how far they could usefully cut right now.  And they have bought into the idea that the Bank’s announced large scale asset purchases are easing monetary conditions relative to what is implied normally by an OCR level of 0.25 per cent.    That is, at best, a heroic assumption, and even much of the Bank’s own commentary has talked in terms of the announced asset purchases limiting a tightening in overall monetary conditions.

Frankly, with no additional macro policy support at all the base scenarios, which assume a return to full employment by the year to June 2024, seem optimistic to say the least.  There is, for example, no mention of how many years it took to get back to full employment after the much-milder 2008/09 recession, even with the support of big cuts in interest rates and the demand impetus from the Christchurch repair and rebuild process.

Still on monetary policy, Treasury publish the inflation forecasts consistent with their scenarios and with unchanged monetary policy.   In none of those scenarios does inflation get back up to even 1.5 per cent until the year to June 2024 (whereas Treasury had been forecasting inflation at around 2 per cent from here on out).  In all the base scenarios inflation drops below 1 per cent –  in some cases goes negative –  in each of the next two years.  There is, then, the small matter of real interest rates –  which will be rising, since Treasury assumes no change in nominal rates.  That represents another downside risk to the economic outlook.

As I noted, all those base scenarios assume current announced policy only.  That includes the expiration of the wage subsidy scheme in June.  I don’t suppose anyone really expected that the government would simply let the scheme run off, with nothing to replace it, in June.  If they were to do so, it is likely that the measured unemployment rate would go much higher.

So The Treasury offers us two more stylised scenarios.  Under scenario 1 (the optimistic one about how quickly alert level restrictions are eased) they toss in another $20 billion of fiscal support aimed at households and businesses.  This variant boosts GDP by 2 per cent over the next year but –  somewhat remarkably –  lowers the unemployment rate by 3 percentage points.  This is the somewhat heroic scenario in which the unemployment rate peaks at only just over 8 per cent and is back at 5.5 per cent by the middle of next year.

Under scenario 2 (see above) we have two more months at Level 4 restrictions and only then drop back to levels 1/2.  In a variant, The Treasury throws in another $40 billion of idscal spending.  Despite the much-worse backdrop, this package is assumed to raise GDP by 4 per cent but only lowers the unrmployment by 3.5 per cent.  Again, the unemployment rate is only 6 per cent by June next year –  all thanks to the power of fiscal policy.   After the last recession, supported by monetary policy and the rebuild, it took five years to get the unemployment rate down by a full two percentage points.

As a final observation, as far as I can tell there is no allowance in any of the scenarios for future fiscal policy tightening, even in those with much much more fiscal stimulus frontloaded.   Given the likely level of public debt that will emerge from all this, it seems unlikely that by 2024 there would be no pressure for a tighter fiscal policy to start lowering again the ratio of public debt to (normal) GDP.   After all, bad as this crisis has been, all the arguments  – about vulnerability to future shocks – used to support low levels of debt in the last 25 years will be just as valid then as they ever were.   And if fiscal policy is tightening, and monetary policy –  here and probably abroad –  is assumed to be doing little, it is even harder to see how we get back to full employment relatively quickly.

It is much better to have the scenarios than not to have them.  The important differentiation they attempt to illustrate is between different levels of government restrictions on economic and social activity.  But the basic level of shock underlying them all still looks as though it is a bit optimistic –  less perhaps about the June quarter (though see above) but about things like the pervasive level of economic uncertainty that seems likely to persist (and be reflected, for example, in a willingness to invest and to lend).   Perhaps one way of seeing that is to do the thought experiment of what if domestic interest rates were assumed to fall by another 500-600 points (and some associated fall in the exchange rate).    Then the speed with which economic conditions get back towards normal in these scenarios might have seemed more plausible –  consistent with past serious recessions.  As it is, it looks a bit as though they have –  consciously or not –  assumed away some of the severity of the problem, in turning keeping the spotlight off the Reserve Bank and off their Minister who has the power to compel the Bank to act.   Monetary policy is supposed to be the primary stabilisation tool.  Failure to use it –  going off gold –  greatly exacerbated the Great Depression; failure to use it now (getting wholesale interest rates materially negative) risks greatly exacerbating this slump.




Measuring the slump

One of the current challenges for economists and others is making sense of the scale of what is happening to economic activity, employment, unemployment, underemployment etc at present.

It isn’t helped by the persistent refusal of successive governments to fund Statistics New Zealand adequately for core functions –  you could think of the Census debacle, but I’m more focused on basic macroeconomic data.  We and Australia are the only two OECD countries without a monthly CPI, our GDP estimates (quarterly only, as with most countries) come out only with a very long lag, we don’t have a monthly industrial production series, and we still don’t have an income-based measure of GDP.   One could add into the mix the degraded state of our timely net migration data too, although for the time being I guess that won’t matter much to anyone (largely closed borders and all that).

Those failings can’t be fixed in short order, although I hope that as we emerge from this crisis the weaknesses in our statistical base will prompt some fresh thinking and a willingness to spend more on these core public services (ones for which there are few votes).

I suggested a couple of weeks ago that for now Statistics New Zealand look at hosting some sort of dashbboard pulling together, and making openly available, all manner of formal and informal economic indicators.  There have to be lots, and I’m sure various government agencies are either producing or collecting all sorts of bits of information, but there are real gaps in what is available to analysts.

There are also some temporary initiatives SNZ could look at implementing.   For example, New Zealand’s unemployment data are drawn from the quarterly Household Labour Force survey.  At present, we can expect no information on anything from the June quarter until early August –  almost four months from now.  But the HLFS is conducted by surveying thousands of households each month, steadily through the quarter.     The full sample will, obviously, tend to produce more accurate estimates than, say, a third of the sample. But there doesn’t seem to be any obvious or good reason why, for now, SNZ could not calculate and publish a subset of the key HLFS series each month, within a couple of weeks of the end of that month, drawing on the sample of households surveyed during that month (still thousands).  The regional numbers might be not worth publishing, and perhaps the fine age breakdowns too, but the headline numbers (unemployment rate, employment rate, participation rate, and perhaps each by male and female) would be very much worth having.  At present, the issue is not whether the unemployment rate is, say, 12.7 or 12.9 per cent, but whether it is more like 10 per cent or more like 15 per cent.  Without the HLFS we have no good way of knowing.   (And while this expedient would be unusual, SNZ has shown some willingness to do the unusual with early reads on foreign merchandise trade data.)  Having this sort of timely monthly data would be likely to be useful for at least the next year.

But there are also some real issues around measurement that are probably more specific to the immediate extreme dislocation.  Take unemployment as an example.  To be counted as unemployed in the HLFS, you have to be without a job, actively looking for a job, available to start work the following week.  And “actively looking” means more than skimming through adverts on job websites.  But right now, in the midst of the partial lockdown, opportunities for search are extremely limited (as are actual vacancies) and many people are extremely constrained in their ability to start work next week even if they wanted to.  There are issues around “employment” too.   To be employed, in an HLFS sense, you have to have done at least an hour’s work for pay in the reference week (assuming you weren’t on annual leave or similar).    But one wonders if the SNZ interviewers have been issued with good guidance as to how to treat people who may be at home at present, still being paid (to some extent or other) but who have done no work at all in the past week (that might encompass people normally doing a job that just can’t be done at home, or public servants never equipped to work from home, or…).  Some of those people might be being only partially paid, through the wage subsidy scheme, but who are pretty certain their current job won’t be there at the end of all this, and who are to all intents and purposes (if not for HLFS purposes) “unemployed” and doing whatever (perhaps little) they can to search for another job.

When we get the full quarterly data there will be a somewhat richer picture of the extent of labour underutilisation (for example, there are questions about hours worked in the reference week relative to usual hours worked), but getting a good read on this month –  which may be the worst of the economic slump –  will always be a bit problematic because we haven’t invested enough in a full monthly HLFS.   I’m not so cynical as to suppose this was a motive (even a month ago when the scheme was designed), but it is certainly convenient that with an election scheduled for September, the current wage subsidy scheme –  which will keep down headline official unemployment numbers, even if beyond that it is little more than (important as that is) a more generous income support scheme – runs into June, encompassing most of the last full HLFS to be out before the scheduled election.

(There are going to be some related sorts of issues with the other main labour market series, the Quarterly Employment Survey, which captures people employed and the hours they are paid for –  both valuable in normal times –  not the amount of work actually being done.  In some areas of the GDP estimates, QES numbers are used.)

I’ve been quite surprised by how small the forecasts for the fall in GDP over the June quarters from a couple of banks have been  (although also not sure how long ago they were finalised).  As noted yesterday, I struggle to see how right now the economy is not running at perhaps 50 per cent of normal (even if a larger percentage of the workforce than that may still be on full pay).  Even if there was a considerable rebound in May and June, if the government decides to allow more activities to occur, it is easy to see a 20-25 per cent fall in GDP in the June quarter.     But whatever the “true” scale of the fall, it seems unlikely that SNZ will have a close-to-accurate read on that fall (and then only quarterly, on official measures) for a very long time, perhaps ever.  We may be more reliant on academic estimates, generated in research papers in years to come, for a “true” read.  And without quarterly (business) income data at present, whatever official estimates SNZ does publish later this year are likely to have substantial margins of error, and be subject to substantial revisions for years to come (it takes several years for GDP numbers to settle towards finality at the best of times).  Indicators that serve reasonably well in normal times may be little use in this exceptional period.

Thus, it will be easy enough to get data on who has been paid, and how much, during this quarter, but not on what they are actually producing.    Working from home is generally less productive –  if it were not so, it would be done more often in normal times.  Working at physical distance is generally less productive (ditto).  But if, for example, much of the public sector component of GDP is estimated from employment/paid hours data….and many of those people are doing little, or even doing quite a bit but much less productively, how will the official statistics ever capture that as a reduction in real GDP?  Nominal GDP data should be subject to fewer distortions, and perhaps we will need to focus more on that than usual over the next couple of quarters.

No doubt there are all sorts of smaller issues, many of which perhaps don’t matter as much.  Clearly the effective cost of groceries has risen this month –  between waiting times, limited choice, no packers etc –  but I’m sure none of that will end up in the CPI.  The actual potential consumption basket is also very different now than in normal times – and in some respects may stay different for some time to come (overseas travel components of the CPI anyone?).

There are real challenges here for analysts and for statistical agencies, in our case SNZ.    In some cases, there are no easy answers (in others, there are some possible remedies).  What would be helpful early on would be some proactive communication from Statistics New Zealand on how they are planning to respond, and to meet the reasonable public demands for information, partial as much of may for a time inevitably be.    It isn’t a time for insisting on utter accuracy: that is important in normal times, but at present the urgency of the situation outweighs that, with more of a need for timely indicative numbers.  As an example of what can be done, see the recent estimate by the French national statistical agency, stepping outside their normal formal frameworks but using the data they, and only they, have overall access to, to produce estimates of the real-time loss of GDP (to which I was pointed by an SNZ manager).  There are gaps here that really only SNZ can adequately fill.