DRAFT: Comments welcome. (The origins of this paper are evident in the text, but its stimulus was a question arising from the interpretation of a standard textbook on international trade.)
The paper was revised in April 2003.
Keywords: Globalisation & Trade; Macroeconomics & Money; Statistics
Over a decade ago I investigated to what extent the CPI could be used to represent the prices of the economy of the whole (GDP), as a part of the study which led eventually to In Stormy Seas. In the process of reducing the vast quantity of material that was produced into the book for publication, the material was left out. (The first draft of the book was about twice as long.)
I must have thought the issue as a methodological curiosum at the time, but since the book’s publication on a number of occasion during public discussions I have wished the material had been publically available. As the next section explains, the section was an illustration of one of the general issues with which In Stormy Seas was concerned, and it is also – as a later section explains – crucial to the understanding how monetary policy works, and how the current management regime may inhibit economic growth.
Prices in the Macro-economy
It is common for an economic exposition to posit the existence of a ‘price level’ (usually designated by ‘P’) which, if any consideration is given to its definition and measurement, is defined as the cost of a representative basket of goods and services. Little attention is given to what might be in the basket, as if it hardly matters. This paper shows it does.
I illustrate this (and other parts of the) conventional wisdom and another points from International Economics: Theory and Policy (5ed) by Paul Krugman’s and Maurice Obstfield. The choice is because it is one of the standard – centre of the paradigm – textbooks rather than because it is particularly remiss. The text first uses an aggregate price measure on page 34 introducing is simply as ‘any price of manufactures PM’ in a microeconomic context. The macroeconomics begins from Chapter 12, and the first reference to an economy wide price index may be on page 340 where the definition of the real rate of return includes ‘measuring asset values in terms of some broad representative basket of products that savers regularly purchase’. On page 367 the aggregate demand for money is introduced with one of the ‘three main factors’ determining it as
The Price Level. The economy’s price level is the price of a broad reference basket of goods and services in terms of currency. If the price level rises, individual households and firms must spend more money than before to purchase their usual weekly baskets of goods and services. To maintain the same level of liquidity as before the price level increase, they will therefore have to hold more money.
Households and firms face different ‘baskets of goods and services’ – conceptually different because households are spending while firms are producing. Yet the textbook draws no attention to the issues.)
They textbook then expresses – quite orthodoxly – the aggregate demand for money as
Md = P X L(r.Y)
with no further attention paid to the choice of P, the price level.
This, as we shall, see becomes important later in the text. Because the ratio Md/P is effectively fixed, various key results are controlled by it.
That there is a stable demand for money equation is not being disputed here. At issue is what is the appropriate price variable in the equation. Econometric estimates of the equation, based on a particular P (and also the particular definition of money, M) are subject to a random error (if not econometric bias), which in part might be due to the wrong choice of price index (and issue further complicated by multicollinearity among the explanatory variables). Moreover the size of the random error (let alone the potential bias) is usually sufficient to make the equation of little value for forecasting GDP (‘Y’ in the above equation: ‘r’ is the interest rate), an error large enough to make a difference to the growth path.
In contrast, In Stormy Seas pointed out that price relativities have varied considerably between sectors, and that at least six different sectors were necessary to describe the behaviour of the New Zealand economy over the last fifty years (pastoral products; energy; other exportables; importables; private non-tradeables; government services). Since the prices of these moved differently it followed that different baskets – that is with different proportions of goods and services from each sector – would reflect different price levels and movements in price levels.
The ‘one-price’ level characterisation of the economy is widespread. In effect the economy is assumed to consist only of a single commodity. For short term macro-economic forecasting this assumption is usually satisfactory, but it rarely makes sense in terms of the medium run economy. Indeed, the assumption is inadequate as it fails to explain one of the most important characteristics of the New Zealand economy – its participation in international trade. In that sense, international economics texts which rely heavily on a single measure of the price level (such as Krugman and Obstfeld) miss the whole point of open economies. Why does a one commodity economy need to import (and export), since apparently the same (composite) commodity is imported (and exported)? Heuristically one might assume a rigid price equivalence between the exports and imports (that is assume that the terms of trade are fixed), and between the tradeables and non-tradeables (that is assume that the real exchange rate is fixed). But that means that the international economics textbook is unable to explain what happens when such relativities change – and thus unable to explain the central experiences of a small open multi-sectoral economy such as New Zealand.
Moreover, the single commodity assumption in economics teaching models is as almost as pernicious as it is widespread. Far too much policy and analysis unconsciously assumes that there is no problem over the choice of the measure to reflect the aggregate price level, and so even more unconsciously it assumes that relative price changes are unimportant in long run economic performance. This is not true for relativities over which New Zealand policy has little practical influence (such as the terms of trade). They have despite their having an integral impact on the growth performance. Even more insidiously, it is not true for relativities over which New Zealand has some influence (most importantly the real exchange rate). If they are either ignored or analysed at a such a simplistic level that the policy conclusions are valueless.
Are Changes in Price Relativities Significant?
In Stormy Seas draws attention to substantial variations in
– the relative price of pastoral exports to imports (the pastoral terms of trade, pp.77-81);
– the relative price of energy relative to production prices (pp.158-160);
– the relative price tradeables to non-tradeable (the real exchange rate, 87-88; 103-106).
As the prologue to this paper mentioned, the research study on which the book was based also looked at the relative price of consumer goods to productions goods.
Consumer Prices
The price of consumer goods is usually measured by the Consumer Price Index (CPI). This is the longest official price series available, beginning annually in 1914 (it is linked back to 1891 – see NZOYB 1990:614), and quarterly from 1925. Over the years the regimen (the items in the index and the weights given to them) has changed. More controversially the treatment of housing and interest payments has undergone major changes over the year, so although it is presented as a continuously defined series it strictly is not. Each figure is published a fortnight after the end of the quarter whose price level it measures, and is not revised.
(There is a second potential (SNA) consumer price series, the ratio of actual nominal consumer outlays divided by the volume (or real or constant price) consumer outlays (as defined and estimated in SNA – System of National Accounts – terms). The available official series is much shorter (annually from 1982?). It becomes available about six months after the end of each quarter and may be revised (for years) thereafter as new data becomes available. This series is particularly useful for international comparisons, because while that the consumer price indexes of most countries have different assumptions, the SNA series are more consistent.)
Producer Prices
The particular producer price series used here (and in In Stormy Seas) is GDEF, which is a price index of all goods and services produced in an economy. It is calculated as the ratio of nominal GDP divided by the volume (or real or constant price) GDP, as defined and estimated in SNA terms.
(Note that the index is – in the conventional notation
Σpnqn/Σp0qn
The base prices (p0) have been updated every year since 1988. This is called ‘chain linking’.)
The available official series is much shorter (annually from March Year 1954/5 and quarterly from 1982?). It becomes available about six months after the end of each quarter and may be revised (for years) thereafter, as new data becomes available.
GDEF also experiences regimen changes over time, just like the CPI. It has the additional complications that until 1971/2? there was no Inventory Valuation Adjustment made to the nominal series. This reduces the recorded stock change by the increase due to inflation.
In summary, the CPI is a longer series, a more timely one, and within its (changing) conceptual framework, it measures more precisely. The advantage of GDEF is that it spans all produced goods and services, and it is not conceptually complicated by the treatment of the housing asset and interest rates. Another distinction – which becomes important later in this paper – is that the CPI basket consists of expenditure items including imports, whereas GDEF is based on a basket of items which are produced.
(An alternative to GDEF would be the producer price indexes. At various stages in the study I used these, and their predecessor wholesale price indexes. This is discussed further below.)
The Relationship Between the CPI and GDEF
The following graph shows the ratio of the CPI to GDEF for the period for which the latter is available. (Note that to make it comparable with the GDEF, the CPI is the year average, not the March quarter on March quarter.)
CPI/GDEF Ratio – Percent Deviation from Trend
It appears that the CPI increases on average about .06 percent a year (.67 percent a decade) faster than GDEF. This hardly matters, despite being econometrically significant. The source of this difference may be that the CPI basket has a higher proportion of services than GDP, or a lower proportion of pastoral products whose terms of trade suffered over the period. In order to assess the interpretation of the graph this trend has been removed.
The graph shows that the ratio of the CPI to GDEF fluctuated between 4.2 percent above and 6.0 percent below the trend. The fluctuations are not entirely random. (The Durban Watson statistic which measures autocorrelation is .87, some distance from the 2.1 value which would indicate there was no year to year correlation between the random errors of the equation).
The fluctuations are sufficient to be horribly misleading if the CPI is used instead of GDEF to deflate nominal GDP to estimate volume GDP. The worst case would have been between the 1968/9 and the 1972/3 year, where using of the CPI would have over-estimated GDP growth by 2.7 percent p.a. While volume GDP grew 16.7 percent over the four year period, the use of the CPI as a deflator would have shown a 29.7 percent increase.
Considerations such as this were sufficient to suggest the CPI was not a good measure of overall inflation. In Stormy Seas used GDEF as the prime price measure indicator (a decision reinforced by the need to make international comparisons).
In passing it should be noted that if inflation is defined as a ‘general (and continuing) increase in prices’ then there was inflation in every year on either measure. The issue is not whether there was inflation, but how much.
The Reserve Bank Target
This work was first done at the end of the 1980s, and has informed my thinking since. However, until this paper I have not consciously applied its implications to the Reserve Bank of New Zealand’s inflation targeting regime. On a number of occasions I have criticised it (see The Reserve Bank Index, although it is not complete). But this is the first occasion that I have explicitly involved this research into the criticism.
The Reserve Bank of New Zealand Act requires the Bank to pursue price stability which it does not define. This is provided for in the ‘Policy Targets Agreement’, which the Minister of Finance and the Governor of the Reserve Bank sign. The PTA of 17 September 2002 says:
Policy Targets Agreement 2002
This agreement between the Minister of Finance and the Governor of the Reserve Bank of New Zealand (the Bank) is made under section 9 of the Reserve Bank of New Zealand Act 1989 (the Act). The Minister and the Governor agree as follows:
1. Price stability
a) Under Section 8 of the Act the Reserve Bank is required to conduct monetary policy with the goal of maintaining a stable general level of prices
b) The objective of the Government’s economic policy is to promote sustainable and balanced economic development in order to create full employment, higher real incomes and a more equitable distribution of incomes. Price stability plays an important part in supporting the achievement of wider economic and social objectives.
2. Policy target
a) In pursuing the objective of a stable general level of prices, the Bank shall monitor prices as measured by a range of price indices. The price stability target will be defined in terms of the All Groups Consumers Price Index (CPI), as published by Statistics New Zealand.
b) For the purpose of this agreement, the policy target shall be to keep future CPI inflation outcomes between 1 per cent and 3 per cent on average over the medium term.
3. Inflation variations around target
a) For a variety of reasons, the actual annual rate of CPI inflation will vary around the medium-term trend of inflation, which is the focus of the policy target. Amongst these reasons, there is a range of events whose impact would normally be temporary. Such events include, for example, shifts in the aggregate price level as a result of exceptional movements in the prices of commodities traded in world markets, changes in indirect taxes, significant government policy changes that directly affect prices, or a natural disaster affecting a major part of the economy.
b) When disturbances of the kind described in clause 3(a) arise, the Bank will respond consistent with meeting its medium-term target.
4. Communication, implementation and accountability
a) On occasions when the annual rate of inflation is outside the medium-term target range, or when such occasions are projected, the Bank shall explain in Policy Statements made under section 15 of the Act why such outcomes have occurred, or are projected to occur, and what measures it has taken, or proposes to take, to ensure that inflation outcomes remain consistent with the medium-term target.
b) In pursuing its price stability objective, the Bank shall implement monetary policy in a sustainable, consistent and transparent manner and shall seek to avoid unnecessary instability in output, interest rates and the exchange rate.
c) The Bank shall be fully accountable for its judgements and actions in implementing monetary policy.
The visitor from Mars might ask why the PTA targets the Consumer Price Index. The historic explanation might be that for most of the seven decades until the passing of the Reserve Bank Act, a major inflationary consideration was the linkage between consumer prices and wages, and the Consumer Price Index became enshrined in the public as ‘the’ measure of inflation. Little consideration seems to have been given when the policy was developed as to whether a better price target would be more appropriate. (Indeed, from the papers made available under the OIA, very little attention was paid to the mechanisms by which the Bank could affect the price level. Rather there seems to have been an acceptance of the theory which underpins the simplest monetarist prescriptions, including the single commodity description of an economy.) Perhaps it might be argued that the primary issue seen to be facing the Bank was to ‘squeeze out’ inflationary expectations, and it was therefore necessary to use an index with which the public was familiar. That is not an reason for continuing with the CPI as the inflation target.
Now-a-days the Bank looks at a range of price indexes (including GDEF) as a part of informing itself of the state of the inflation. But the target remains the CPI. This would not matter if the prices all moved closely together. The CPI versus GDEF comparison shows they do not.
There are 47 years (since 1955) for which there is data on the annual price changes. In over half (28) the price changes in the two indexes differed by more than 1 percentage point. In another 13 (over a quarter) the price changes differed by between .5 and 1.0 percentage points. Thus in only 6 years (less than a quarter) was the difference less that half a percentage point. These are much larger differences than are accepted as tolerable in the public discussion on the forecasting and outcomes of the CPI. (The Standard Deviation of the difference over the full period was 2.1 percentage points.)
Is this only a short term phenomenon?. The PTA does not specify how long the medium term is, so let us take a three year horizon – about the length of the average business cycle. In 26 of the 45 observations (over half) the changes in the two indexes differed by more than 1.5 percentage points (or .5 percentage points a year).
What about the period in which the Act was in force? Excluding 1990/1 (a bad year in which the divergence was 3.1 percentage points), in the remaining 13 years, there were four in which the percentage increase between the CPI and GDEF exceed 1 percentage point, and another five when the divergences was between .5 and 1.0 percentage points. So the pattern of deviation since the Act has not been very different from before the Act. (I have not detailed the three year pattern, because the run is so short, but the long term conclusion probably applies to it too.)
In summary, there has been considerable year to year deviation between the two price indexes over the period for which we have data, and it is not evident that there has been a significant improvement in the last decade either as a result of the Reserve Bank Act or of the improvements in the measurement of GDEF.
Are There Consequences from the Choice of Price Index?
Whether there are consequences from the choice of the price index used to define price stability in the PTA is a bigger issue than this paper’s concerns. Here I simply set down some down.
The CPI and GDEF measure very different forms of economic behaviour. The CPI is an expenditure-side price index, and GDEF is a production-side one. Effective targeting on the CPI requires some theory of the determination of expenditure prices, whereas targeting on GDEF has a more production orientation.
In fact, the Reserve Bank’s account of the inflation mechanism seems to focus on the production side, especially the output gap in aggregate and in individual markets. Ii is unclear how this relates to the expenditure-side price measure which the Bank is targeting on.
In a one commodity model there is no problem. But in fact the CPI is based on a basket of which only around 60 percent of which is domestically produced, the other 40 percent being imported. (This is an old ratio and has not been updated. It is net of indirect taxes and subsidies.) Assuming, for a moment, that the price of imports (and the exchange rate) are fixed, monetary policy is targeting on only 60 percent of the CPI basket, which makes up about, say, 36 percent of GDP (since private consumption is only about 60 percent of GDP). In particular the pricing behaviour of most of the investment, almost all the public sector, and a large part of the export sector have little relevance to the CPI.
IUnfortunately the prices in that 36 percent do not move in parallel with the other prices in the economy, as the CPI to GDEF ratio shows.
Moreover, the CPI’s considerable dependence on import prices leaves another circuit to reduce its level – the exchange rate. Monetary policy can be used to push up the nominal exchange rate, which lowers import prices and thereby reduce the rate of CPI inflation. Textbooks, such as Krugman and Obstfeld, imply this is not possible in the long run, but they use a single price (that is, they are single commodity models ill suited for open economies) and the whole point of an exchange rate change is that it changes the relative price between the expenditure and production side of the economy. The analytics of aggregate price determination models become very murky when the exchange rate appears in the aggregate price level. It seems that the models may generate multiple equilibria with a high and low GDP (although the higher one may be unstable).
Once the exchange rate becomes an intermediate target (whether consciously or unconsciously) it impacts on the CPI inflation rate through the following circuits:
– it directly lowers the price of consumer goods and services via a lower price of imports;
– it pressures domestic producers competing with imports to keep their prices as low;
– it eases wage pressures, in so far as wage demands are in part determined by changes in consumer prices.
Under the fixed exchange rate regime before 1985, New Zealand had a practice of over-valuing the exchange rate for anti-inflation purposes . Under the floating rate which followed after 1985, monetary policy restrained inflation by using the same over-valuation circuit.
The Exchange Rate and the Growth of the Economy
But as experience of the decade after the float shows, an over-valued exchange rate restricts the expansion of exportable sector, which depresses the aggregate growth rate of the economy. This was a major theme of In Stormy Seas, and the evidence since its writing confirms the conclusion that the health of the exportable sector is crucial for sustainable economic growth. (Notice this introduces an extra commodity in the story: ‘exportables’. Krugman and Obstfeld do not even cite ‘tradeables’ in their text book’s index.)
What differed between the pre- and post-float regimes of exchange rate over-valuation was that in the earlier period the government subsidised (in a variety of ways) the exportable sector (while imports were also restricted). This is not to argue for those policies, but to explain how the interventionist wedge enabled the exportable sector to grow.
Almost all the export subsidies and import restrictions were eliminated from 1985, while the exchange rate valuation was maintained. Exports now almost stagnated. while imports surged. The actual growth rates between calendar years 1985 and 1993 were 3.1 percent p.a. for goods exports and 4.5 percent p.a. for imports of goods, according to the OECD Outlook, while OECD exports grew 4.8 percent p.a. and 5.6 percent respectively. Without going into details – I have done so elsewhere – New Zealand lost world export market share, while its import penetration ratio (relative to GDP) rose. Exports did better than the pure theory might have predicted given this over-valuation, because there was a carry-over from earlier (pre-1985) policies which stimulated the export supply of petrochemicals, forest products, and horticultural products (probably others, but these impacts are measurable), while the energy/petrochemical expansion also reduced imports.
What happened was exactly as economics predicted. When the return to exporting fell from the over-valued exchange rate (and the removal of subsidies) the export industry shifted back down the aggregate export industry supply schedule until all that was left was the (diminished) industry which was viable at the higher excahnge rate. At that point it began expanding again – as did the New Zealand economy as a whole.
Much of this analysis was available over a decade ago. What this section does is tie in how the choice of target price index in the PTA can limit the poor performance of the New Zealand economy by further detailing the mechanisms involved. This is also the most detailed account I have written of the ineptitude of the single commodity model to characterise the small open multi-sectoral economy.
What should be the PTA Target Price Index?
The extension to this paper is to raise the question as to what is the appropriate price index (or indexes) to target upon, since the choice of index matters. The long run relevance of the CPI is unclear given that it is an expenditure side index, covers a relatively small part of the economy, and is overly influenced by the exchange rate.
On the other hand GDEF is to slow to be published, and subject to revisions. However there are other quarterly production based price indexes – in particular the producer price indexes – which are much more economy wide that the CPI. One might envisage Statistics New Zealand upgrading the indexes, including speeding up their publication, so that one (probably the PPI-Outputs) could be phased in to replace the CPI over a number of renegotiated PTAs. A weakness of the PPI series is that we do not have the same experience of their use (historically they are – in their current form – a shorter series), and they may be subject to greater error. But as Keynes famously said ‘it is better to be vaguely correct, than precisely wrong’.
However, it even more important to limit the use of an over-valued exchange rate as the primary circuit for controlling inflation in New Zealand. The first step would be to discard single commodity accounts of the economy, to recognise there are a variety of price relativities which are relevant, and that they move quite differently from one another.