Opinion: Rodney Dickens examines "the fundamental flaw in the way OCR decisions are made"

By Rodney Dickens

In the first report on monetary policy released last week I reviewed New Zealand’s experience with an independent central bank since 1989.

I concluded that the experience since 1989 should teach us to be wary of what ill-conceived experiment the Reserve Bank (RBNZ) will come up with next.

The analysis contained in that report suggested that under Dr Alan Bollard’s governorship monetary policy has been more a disease than a cure.

Most importantly, he could have achieved the same average inflation outcome without playing a major part in sponsoring the boom-bust economic and housing market cycles experienced between 2002 and 2009.

In today’s report I expose the fundamental flaw in the way OCR decisions are made by the RBNZ.

The flawed approach used by the RBNZ means monetary policy will inevitably be a source of volatility in economic activity and especially housing market activity, in the incomes of retired people who often have a sizeable portion of their wealth in interest bearing investments and probably also in the NZD.

Economic forecasts are a key ingredient in the framework the RBNZ uses to make OCR decisions. But the forecasting-based approach to making OCR decisions is flawed because of the RBNZ’s generally poor and at times abysmal forecasting track record. This report contains some examples of how bad the RBNZ’s forecasts can be.

Lacking quality economic forecasts the governor ends up largely adjusting the OCR in response to economic upturns and downturns. Fine tuning the OCR on a reactive basis is a sure route to monetary policy causing unnecessary volatility in economic activity, meaning people face a more volatile economic environment than is necessary for the RBNZ to achieve its inflation target.

The final report on monetary policy in this series for three reports will be released next week. It will present an alternative and superior approach to making OCR decisions.  The last thing we need is another ill-conceived monetary policy experiment. The new approach I will launch next week has many things going for it.  It is in tune with the RBNZ’s flexible inflation target - that is, keeping annual CPI inflation in the 1-3% target range on average over the medium term.  It doesn’t rely on dubious economic forecasts. It will enable the RBNZ to achieve the inflation target without monetary policy being the primary source of volatility in economic activity. What I am proposing will deliver more stability in interest rates, economic activity and possibly also the exchange rate, but it will also allow the RBNZ to respond quickly and appropriately to the impact of major shocks, like the global financial crisis.
 
The current forecasting-based approach to making OCR decisions

Decisions about the OCR are made by the RBNZ every six weeks.

The March, June, September and December OCR decisions are accompanied by a Monetary Policy Statement (MPS). Each MPS contains a policy assessment, an overview and key policy judgments, and assessments of financial market developments, current economic conditions and the macroeconomic outlook - that is, the economic forecasts. They also often have special features on important issues like the impact of the Canterbury earthquakes. In the other months the OCR announcements are accompanied by one page releases that summarises the key factors the RBNZ took into account in making the decisions.

A number of years ago the RBNZ made public the process it goes through in the build up to releasing a MPS. Things might have changed a bit since then, but probably not much.  It was something like an eight week process. The RBNZ forecasters are involved early in the process, producing forecasts that provide a focus for debate over what should be done with the OCR. The economic forecasts play a key part not just in the OCR decision at hand but also in the RBNZ’s assessment of what should be done with the OCR in the future. The linkage between the RBNZ’s view on the economic outlook and what it plans to do with the OCR is confirmed by the concluding paragraph in the policy assessment of the June 2011 MPS:

As GDP growth picks up, underlying inflation is expected to rise. A gradual increase in the OCR over the next two years will be required to offset this, such that CPI inflation tracks close to the midpoint of the target band over the latter part of the projection. The pace and timing of increases will be guided by the speed of recovery, but for now the OCR remains on hold.

The current approach is systematic and rigorous in its own way, and is aptly described as a forecasting-based approach to making OCR decisions. It represents a huge step forward from the first Monetary Policy Committee meeting I attended, which was in either late-1986 or early-1997. The quality of the debate in that meeting appalled me, with it being limited to a series of 'I think this' and 'I think that', with no supporting analysis. Thankfully I have stopped having nightmares about being back working at the RBNZ.

Conventional wisdom, which is well founded in this instance, is that it takes around a year for changes in the OCR to impact fully on economic growth and around another year for the impact on economic growth to filter through to inflation. This means decisions about the OCR today are relevant to what will happen to inflation in two years’ time.  In the ideal world the RBNZ would be able to use its inflation forecasts as a basis for deciding the appropriate current level of the OCR and the outlook for the OCR.

The forecasting-based approach to making OCR decisions in practice

In the previous report on monetary policy I exposed the 'go for growth' approach pursued by Governor Bollard in the early to mid 2000s.

Alan’s experiment with low interest rates in this period culminated in inflation exceeding the top of the RBNZ’s 1-3% target range for 10 of the 14 quarters between the 2005 September quarter and the 2008 December quarter, as shown in the chart. Eighteen months before the first breach of the target range the RBNZ was predicting that inflation would remain in the 1-3% target range over the next three years (see the blue line, which shows the inflation forecasts released by the RBNZ in March 2004). 

The forecasts released by the RBNZ two years before the first breach didn’t cover the whole period of the first breach, which lasted from the 2005 September quarter to the 2006 September quarter, which is why I have used the RBNZ’s 2004 March forecasts. In the 2003 September forecasts the RBNZ also wasn’t predicting a breach.  Using the 2004 March forecasts gives the RBNZ forecasters a better chance of getting it right.

Based on the predictions that inflation would not breach the top of the 1-3% target range over the subsequent three years, in March 2004 the RBNZ was predicting that the 90-day bank bill yield (which can be viewed as a proxy for the OCR) would drift sideways over the subsequent three years (the blue line in the adjacent chart). The 90-day bank bill yield was supposed to average 5.6% in the 2007 March quarter when it ended up averaging 7.8% on its way to peaking at 8.8% in the 2008 March and June quarters.

This starts to reveal the fundamental flaw with a forecasting-based approach to making OCR decisions. But it is instructive to look further behind the RBNZ’s 2004 March predictions.

At the time these were prepared the RBNZ had delivered only one of the thirteen 0.25% OCR hikes it would deliver between January 2004 and July 2007.

In March 2004 the RBNZ was predicting that the unemployment rate would increase from 4.5% to 5.5% (green line) and that annual inflation in the adjusted labour cost index would moderate from 2.25% to 2% (blue line). The chart shows a near perfect inverse relationship between the unemployment rate (right scale) and the measure of labour cost inflation that the RBNZ focuses on most because it is adjusted to remove the good part of pay increases - that is, the part justified by productivity growth. It measures the bad part of pay increases that is purely inflationary and will result in firms experiencing rising costs per unit of production and lead to them putting up prices - that is, it won’t translate into real income increases but merely fuel a damaging wage-price spiral.

The best fit in the chart is with the unemployment rate advanced or shifted to the right by three quarters, which reflects how long it takes on average for changes in the unemployment rate to filter through to this measure of labour cost inflation. For the mathematically minded the correlation is shown (-0.86 versus a maximum possible negative correlation of -1.0).  This is a beautiful example of Economics 101 at work.

Via an act of magic or some miracle, the sustained low level of the 90-day bank bill yield the RBNZ was predicting was not going to result in an inflation problem according to the RBNZ forecasters. The RBNZ forecasters were not alone, with most economic forecasters participating in similar acts of wishful thinking at the time, although with one exception - that is, me.

But the economy takes no notice of economic forecasts that are based on wishful thinking.

The unemployment rate duly fell from 4.5% to a trough of 3.2% in the 2005 December quarter, while the measure of labour cost inflation shown in the chart above duly accelerated from 2.25% to a peak of 3.7% in the 2008 September quarter. I visited my ex-colleagues at the RBNZ twice during this period armed with the chart above and tried to warn them that their forecasts were founded on sand, but they responded with a list of reasons why this time would be different.

Part of the problem with a forecasting-based approach to making OCR decisions is that even the best available economic forecasting model struggles to capture the dynamics of how the economy works.  But even if the RBNZ had a forecasting model that encapsulated how the economy actually works there would still be problems. Feed any forecasting model with dubious assumptions and it will spit out unrealistic predictions. But wishful thinking isn’t just the domain of economic forecasters.  It is practiced in a wide range of areas, including too often in business projections.

Forecasting what inflation will be over the next 2-3 years faces more challenges than just the 'rubbish in, rubbish out' problem with forecasting models exposed above.  The other challenges include:

• Inflation can be highly volatile because of the significant impact large changes in energy prices have on the Consumers Price Index (CPI).

• Inflation is impacted significantly by the major cycles in the exchange rate because New Zealand has a large traded goods sector - that is, imports and exports make up sizeable shares of economic activity.  Import prices are particularly impacted by the vagaries of the exchange rate.

• The government has the habit of changing the rate of GST and/or other government charges, which can at times have a significant short-term impact on inflation, although this isn’t relevant to monetary policy decisions.

• The RBNZ faces a moral hazard.  It has an incentive to forecast that inflation will remain within the target range to try and influence inflation expectations, and wage and price setting behaviour. This potentially introduces a “political” element or bias in the RBNZ’s inflation forecasts.

How the RBNZ copes in the face of not being able to rely on its inflation forecasts

After adjusting the OCR the RBNZ doesn’t want to wait two years to see if it has the desired impact on inflation. This would run the risk of letting inflation get well down the wrong path before a corrective adjustment in the OCR is made, which is effectively what Governor Bollard did with his 'go for growth' experiment.

Instead, the RBNZ and central banks in general respond to this problem by focusing mainly on economic growth prospects. As mentioned above, conventional wisdom is that it takes around a year for changes in the OCR to impact fully on economic growth and around another year for the impact on economic growth to filter through to inflation. Economic growth provides insights into the outlook for underlying inflation - that is inflation excluding the impact of the volatile factor like petrol prices and the exchange rate. This means the RBNZ has an option other than relying on dodgy inflation forecasts or waiting to see inflation outcomes.

The following policy assessment from the 2011 June Monetary Policy Statement shows the emphasis the RBNZ puts on the outlook for economic growth:

The Reserve Bank today left the Official Cash Rate (OCR) unchanged at 2.5 percent

The outlook for the New Zealand economy has improved since the publication of the March Statement.

Economic activity has been significantly disrupted by the Christchurch earthquake. However, while many firms and households – particularly within Canterbury – continue to be adversely affected, it appears the negative confidence effect of the earthquake on economic activity throughout the rest of the country has been limited.

The early signs of recovery noted in the March Statement have continued. Despite some continuing signs of weakness in the world economy, commodity prices remain very strong and firms expect to increase their hiring and capital investment. Reconstruction in Canterbury is projected to add about 2 percentage points to GDP growth over 2012, and boost the level of activity for several years thereafter.

Despite the strong outlook for export earnings, household expenditure is expected to grow only modestly. Household debt remains very high and is expected to constrain retail spending and the housing market for some time. Continued fiscal consolidation will also act to dampen activity.

The New Zealand dollar has appreciated substantially over the past two months. This appreciation, supported by high export prices for primary producers, is negatively affecting other parts of the tradable sector, constraining rebalancing of the New Zealand economy.

Headline inflation is currently being boosted by recent increases in indirect taxes, food and petrol prices, and surveyed expectations of future inflation have edged up. Despite this, indicators of capacity usage and core inflation suggest underlying inflation remains constrained.

As GDP growth picks up, underlying inflation is expected to rise. A gradual increase in the OCR over the next two years will be required to offset this, such that CPI inflation tracks close to the midpoint of the target band over the latter part of the projection. The pace and timing of increases will be guided by the speed of recovery, but for now the OCR remains on hold.
 

The first four of the seven paragraphs in the policy assessment above focus on economic growth prospects, while only one paragraph is dedicated to the inflation outlook.

And the first comment in the last paragraph that explains what the RBNZ plans to do with the OCR is on the growth outlook, followed by what it implies for inflation. So the game the RBNZ is playing in practice is trying to achieve economic growth outcomes that are consistent with its inflation target, with its forecasts of economic growth being a key input. If you can’t predict inflation accurately this makes sense in the context of the conventional wisdom of the link between the OCR, economic growth and inflation.

However, the last paragraph of the policy assessment quoted above is effectively a confession to adjusting the OCR in response to what happens with GDP growth rather than in a proactive manner - 'The pace and timing of [OCR] increases will be guided by the speed of recovery'. Lacking reliable forecasts for inflation or economic growth the governor is forced to largely wait and see what happens with economic growth and adjust the OCR reactively. The experience in 2010 is the most recent of many examples of this reactive behaviour by the RBNZ.

In June 2010 the RBNZ was predicting robust economic growth over the next two years (orange line, left chart) and reflecting this it was predicting that the 90-day bank bill yield would increase dramatically over the next two years (orange line, right chart). The RBNZ’s 90-day bank bill yield forecasts are proxies for OCR predictions.  Reflecting the predictions, the RBNZ hiked the OCR 0.25% in both June and July 2010. But economic activity increased only 0.14% in the 2010 June quarter, although this wasn’t made public by Stats NZ until 23 September 2010. At the time the RBNZ released its predictions of robust near-term economic growth and large interest rate increases the economy was already struggling to grow at all.

In September 2010 the RBNZ revised down its GDP growth forecasts (green line, left chart), which was only partly in response to the expected initial negative impact from the 4 September Canterbury earthquake. It also revised down its predictions for the 90-day bank bill yield (green line, right chart). In the 2010 September quarter economic activity contracted 0.2%, but even without the negative impact of the earthquake economic growth was weak. At the time the RBNZ released its September 2010 predictions economic growth had been almost stagnant for two quarters.

In December 2010 the RBNZ revised down its near-term GDP growth predictions more, although it also revised up its predictions for the second half of 2011 and for 2012, largely reflecting the expected rebuilding in Canterbury (red line, left chart). At the same time it again revised down its predictions for the 90-day bank bill yield (red line, right chart). Economic activity grew only 0.21% in the 2010 December quarter. In the 2010 December quarter economic activity or GDP was up only 0.8% from the previous December quarter. This compared to the RBNZ’s predictions of 3.6% growth in June 2010, 2.8% growth in September 2010 and 1.5% growth in December 2010 (some of these predictions for annual GDP growth are interpolated from the RBNZ’s annual average predictions).

In March 2011 the RBNZ cut its near-term economic growth forecasts even further, which was only partly in response to the expected negative impact of the 22 February Christchurch earthquake. And it revised up its predictions for GDP growth in late-2011 and 2012, mainly because of the super-sizing of the reconstruction job by the 22 February earthquake (blue line, left chart). But in March 2011 it again cut its year-ahead predictions for the 90-day bank bill yield (blue line, right chart).

This was like closing the barn door after the horse had bolted.

Even after allowing for the impact of the earthquakes, the left chart above highlights both how inaccurate the RBNZ’s near-term economic growth forecasts can be and the tendency for them to react to actual outcomes rather than provide useful insight into the future.

The magnitude of the changes in the RBNZ’s 90-day bank bill yield forecasts over a period of only nine months shown in the right chart above is frightening.

The majority of the change is not related to the earthquakes but rather because aside from the earthquakes economic growth in 2010 ended up dramatically weaker than the RBNZ was predicting.

We spent much of 2010 warning clients of our monthly economic reports that economic growth would be significantly weaker than the RBNZ was predicting and that the RBNZ’s interest rate forecasts were way off the mark. In our August 2010 economic report - that is, prior to the first major earthquake - we even warned clients that the RBNZ might be forced to reverse the OCR hikes delivered in June and July. But even with our superior framework for assessing near-term economic growth prospects I wouldn’t have the arrogance to pretend that I could fine tune the OCR in a proactive, useful way.

There was nothing even vaguely proactive about what the RBNZ did in 2010. It waited a long time after the first evidence of weak economic growth before it cut the OCR 0.5% in March 2011.

This follows a long tradition of monetary policy reacting to economic outcomes.

In my assessment the RBNZ’s overly optimistic economic growth forecasts played a significant part in the RBNZ being extremely slow to realise that the economy was struggling to grow at all in 2010.

The experience in 2010 makes a mockery of the RBNZ’s forecasting-based approach to making OCR decisions.

Adjusting the OCR in reaction to outcomes for economic growth is much better than waiting two years after adjusting the OCR to see the impact on inflation, but it means the RBNZ is at risk of chasing its tail.

If the RBNZ is confident in its predictions that economic growth will be excessively strong or weak why wait until this is confirmed before acting when it takes around a year for changes in the OCR to impact fully on economic growth? Tightening or easing monetary policy in reaction to what happens to economic growth is a recipe for exacerbating economic cycles.

A parody of the decision making process the RBNZ effectively uses runs like this:

• Hike the OCR a notch in response to the first evidence of excessively strong GDP growth - noting that the strong GDP growth may have largely been the result of earlier OCR cuts that will take around a year to be fully reflected in the data, although other primary drivers of economic growth, including export prices, the exchange rate and net migration may have played a part in fuelling the initial strong economic growth.

• Keep hiking the OCR until there is evidence that the hikes are starting to slow economic growth, which is effectively what the RBNZ is currently planning on doing.

• Once the economic data confirm that economic growth is moderating, halt the hikes.

• If economic growth shows signs of slowing excessively start delivering OCR cuts and continue to do so until the economic data show that economic growth is improving.

This parody of how monetary policy is operated is somewhat simplified, but it isn’t too far from the mark in terms of how it has been operated much of the time since 1989. This was especially evident during the Wild West period of monetary policy in the 1990s, as described in the first report on monetary policy.

The problem is that it takes around a year for changes in the OCR to impact fully on economic growth and the delay in getting feedback about the impact is exacerbated because the GDP data are released almost three months after the end of each quarter. This means that adjusting the OCR largely in response to the strength of the economic growth data puts the RBNZ at great risk of over hiking, resulting in an excessive slowdown in economic growth, followed by over cutting, resulting in an excessive acceleration in economic growth - the RBNZ chasing its tail. This is evidenced by the excessive volatility in economic activity and especially housing market activity since 1989, as exposed in the first report on monetary policy.

Having tried and apparently abandoned the ill-conceived, 'go for growth' experiment discussed in the first report, I can’t be certain what framework Governor Bollard is now using to make OCR decisions.

He may not even have a formal framework in his head.

For good reason he is unlikely to be putting much faith in the RBNZ’s economic forecasts. Based on the recent policy assessments in the Monetary Policy Statements he is now behaving much as described in the parody above. Even allowing for his conservative approach this puts him at risk of chasing his tail, with undesirable consequences for the stability of economic growth and especially for the stability of activity in the real estate and residential building industries, and for the wide range of firms and individuals servicing these industries.

Why does the RBNZ use a forecasting-based approach for setting the OCR?

The track record of the RBNZ’s forecasts is poor, which is the case for economic forecasts more generally. It certainly doesn’t provide a sound basis for fine tuning the OCR. So why does the RBNZ persist with the forecasting-based approach to making OCR decisions?

History explains many things. The history of an independent central bank in New Zealand started with Governor Brash having a strict 0-2% target range for annual CPI inflation that was supposed to be enforced continuously. When you have a strict target like this is makes sense to try and use economic forecasts as a key important input into your decisions in the hope that they will give you the earliest possible warnings that inflation might head outside the tight target range. But for good reasons, as discussed in the first report on monetary policy, we have moved on from this strict target range. Unfortunately, the RBNZ has continued to play a forecasting-based game despite overwhelming evidence that it is flawed.

It is well past time that a better and more appropriate framework is adopted for making OCR decisions. But the last thing we need is yet another of the ill-conceived monetary policy experiments the RBNZ has been inflicting on the country since 1989.

What I propose is a framework that is in tune with the RBNZ’s flexible target - 1-3% on average over the medium term - one that doesn’t rely on dubious economic forecasts and will achieve the medium term inflation target without being a primary source of volatility in economic activity. The third report will unveil the framework I believe the RBNZ should be using to make OCR decisions.

Just in case you are willing to give the forecasting-based approach to making OCR decisions one more chance …

What did the RBNZ forecasters learn from the major mistakes they made in the March 2004 predictions discussed above? They learnt not to make the same mistake again, but rather to make the opposite mistake this time around.

While in March 2004 they predicted no increase in the 90-day bank bill yield (blue line, chart below), which implied no increase in the OCR, in March 2010 they were predicting a major cyclical increase in the 90-day bank bill over the subsequent three years, taking it back to around the historical average level (red line). In the early to mid 2000s there were no good reasons why keeping the 90-day bank bill yield below the historical average rate would not result in an inflation problem despite the inclination of the RBNZ forecasters to argue that things would be different that time around.

Having got it wrong last time around the RBNZ forecasters were predicting a normal, large, cyclical increase in interest rates this time around. But if there was ever a time that would be different than normal it was this time around and not because of the horrific Canterbury earthquakes. In March 2010 we were still living in the shadow of the global financial crisis, which is a crisis of a lifetime if not several lifetimes. We are likely to be living under this shadow for many years to come. This means there is an extremely good reason why the 90-day bank bill yield should not return to around the pre-crisis average level any time soon.

In my early years at the RBNZ I was involved in building parts of one of the forecasting models. Dr Rod Deane, best known these days for his involvement at Telecom, was the RBNZ’s Chief Economist at the time and had led the team that built the first economic forecasting model in New Zealand. One of the pearls of wisdom Rod bestowed on us was that an important part of using a forecasting model was to make sure that your predictions were internally consistent. For example, don’t predict that both interest rates and residential building activity will increase because interest rates are the most powerful driver of residential building activity and whenever they have risen significantly residential building activity has subsequently fallen. Equally, if you predict that the unemployment rate is going to fall significantly, which means a tighter labour market and increased bargaining power for employees, you had better be predicting that labour cost inflation will subsequently increase.

Unfortunately, the RBNZ forecasters have long forgotten Rod’s sage advice.

In the June 2011 MPS the RBNZ forecasters predicted that, excluding the rebuilding in Canterbury, the residential building activity share of total economic activity would increase sharply over the next three years (blue line, left chart) at the same time as the 90-day bank bill yield will increase significantly (blue line, right chart).

Just to make sure there isn’t any doubt over this issue, the second left chart, which is reproduced from the first report on monetary policy, shows the strong inverse relationship between mortgage interest rates and residential building consents. With floating and short-term fixed mortgage rates currently the cheapest in the market, changes in the 90-day bank bill yield will have a major impact on borrowing costs. And there is no way the increase the RBNZ is predicting for the 90-day bank bill yield will result in a sharp rise in the residential building share of GDP when the rebuilding in Canterbury is excluded.

The right chart above shows that the RBNZ is currently predicting that the unemployment rate will fall significantly over the next three years (dashed red line) while annual inflation in the adjusted measure of labour costs will zigzag sideways, albeit after an initial increase (blue line). This is one area the RBNZ forecasters should have learnt the lesson about being internally consistent, but are at times immune to learning from past mistakes or the lessons they learn just lead to new mistakes, although occasionally they get some things reasonably close to the mark.

What the RBNZ does with the OCR has massive economic and human implications.

I shudder at the thought that forecasts like the ones shown above play any part in determining what the RBNZ does with the OCR. The number one rule of monetary policy should be that if you can’t forecast the economy’s future, don’t mess around fine tuning the OCR. Doing so runs the risk of operating monetary policy in a reactive manner, which will inevitably exacerbate rather than reduce volatility in economic activity.

We have a proven track record in providing clients with superior assessments of near-term prospects for economic growth, housing market activity and interest rates.

Having seen the flaws in the traditional approach to economic forecasting used by the RBNZ first hand I have spent the last 25 years building a completely different approach to assessing near-term prospects. This means we can often provide superior insights into what the RBNZ will do with the OCR in the future than can the RBNZ. We are in the game of providing clients with valuable inputs for operational and/or strategic decisions. We are not in the game of providing inept economic forecasts, although we aren’t infallible, especially when events like the global financial crisis first hit town.  Visit our website (www.sra.co.nz) or contact me for information about the reports and services available. But even with our superior ability to assess economic upturns and downturns before they occur I would not propose fine tuning the OCR based on the insights we can offer.

This is especially the case when I take into account the RBNZ’s target of achieving 1-3% inflation on average over the medium term. The flexibility offered by the inflation target and the clause in the Policy Targets Agreement signed between the RBNZ Governor and the Minister of Finance requiring that stability in output, interest rates and the exchange rate be taken into account mean the current approach to making OCR decisions should be dumped and replaced with a much more fitting approach. The new approach should also take into account some other factors, including:

• Why the RBNZ was given independence over monetary policy in the first place.

• The tool(s) of monetary policy that fit best with the underlying reason why the RBNZ is charged with operating monetary policy independently from the government.

• The merits of the governor having the final say in OCR decisions versus the committee approach favoured in a number of other countries.

In the third report on monetary policy to be released next week I will present what I believe represents a far superior approach to making OCR decisions than either the forecasting-based approach presented in the Monetary Policy Statements or the actual reactionary approach generally used by the RBNZ.

---------------------

* Rodney Dickens is the Managing Director Strategic Risk Analysis. See more detail here.