RDP 2001-02: Changes in the Determinants of Inflation in Australia 3. Influences on Inflation

Inflation in Australia has typically been considered in the context of a mark-up model so that, in the long run, the domestic price level is a mark-up on total unit costs of production.^[11] For an open economy, these costs include imported inputs to production as well as domestic inputs. Consequently, estimated mark-up models present us with a set of key variables, both foreign and domestic, that have played a significant role in explaining actual inflation outcomes. Prominent in this set are import prices, wages and productivity. We explore the behaviour of these variables over the 1990s and identify changes that are conducive to low and stable inflation.

3.1 Import Prices

In recent years, import prices have contributed surprisingly little to consumer price inflation. This has invited claims of a structural change in the pass-though relationship that has increased the immunity of the inflation process to external shocks. The basis of these claims can be well summarised in Figure 3, which shows how import prices have moved during three episodes of currency depreciation. (The exchange rate index comprises the currencies of Australia's major trading partners, weighted by import shares. It is expressed in Australian dollars per unit of foreign currency so that a rise indicates depreciation.)

In each episode of depreciation, import prices ‘at the docks’ moved approximately in line with changes in the exchange rate. In the 1980s, these price movements at the docks were also translated into sharply rising retail import prices that provided considerable impetus to domestic inflation. In the early 1990s, however, import prices at the docks appeared to have a much smaller effect on retail import prices (measured here by the imported component of the CPI). By the late 1990s, they appeared to have little or no effect on retail import prices so that, despite a significant depreciation, domestic inflation remained undisturbed.

While these stylised facts provide a strong prima facie case of a change in the pass-through relationship, careful examination of events leads to more modest conclusions. In this section, we present estimates of the pass-through of changes in the exchange rate to import prices over the docks, or ‘first stage pass-through’. We also present estimates of the responsiveness of final consumer prices to changes in import prices over the docks, or ‘second stage pass-through’. In both cases, we pay particular attention to the experience of the 1990s.

3.1.1 First stage pass-through

First stage pass-through is, in essence, an application of the law of one price. In its absolute form, the law states that the price of a traded good should be the same in both domestic and foreign economies, when expressed in a common currency, and can be written as:

where P is the domestic price of imports at the docks, P^* is the corresponding foreign price, E is the exchange rate (a basket of rates expressed in units of domestic currency per unit of foreign currency). The extent of first stage pass-through is represented by the elasticity of the domestic at-the-docks import price with respect to the exchange rate. It is complete when this elasticity is unity so that all of a change in the exchange rate is passed on to a change in the import price at the docks.

Since Australia is a small open economy, theory predicts that import price pass-through should be complete.^[12] This prediction is usually borne out with aggregate import data (Dwyer, Kent and Pease 1993).^[13] For a given world price, changes in the exchange rate are fully passed on to changes in import prices at the docks. Furthermore, the typical finding is that first stage pass-through is completed rapidly, with most of the adjustment occurring within one year.^[14] However, a recent challenge to estimating the extent of first stage pass-through is associated with the Asian financial crisis.

Most often, there is a high degree of co-movement between each of Australia's bilateral exchange rates so that most measures of an effective exchange rate move similarly, regardless of country coverage or weighting systems. But during the Asian financial crisis, the Australian dollar appreciated against the currencies of the troubled Asian economies and depreciated markedly against the major currencies, making the choice of effective exchange rate important. To assess the impact of these divergent currency movements on domestic import prices it is necessary to properly control for changes in the foreign prices of goods exported from each trading partner. While bilateral exchange rates are readily available for all of Australia's trading partners, timely or reliable estimates of the relevant export prices are not. These are largely confined to the G7.

Our approach is to view the G7 countries as price-makers that set a notional world price for goods and services and estimate first stage pass-through using the currency and export prices of these G7 countries, rather than those of a broader group. In other words, our approach is to investigate whether P = P_G7 E_G7.

In Appendix B, we present the estimated import price equation from the small model of the Australian economy presented in Beechey et al (2000). This import price equation is of the standard error-correction type, but has two special features. These are designed to control for the fact that export prices from non-G7 countries may deviate from the notional world price. Beechey et al (2000) include a dummy variable to capture price undercutting by Asian exporters following the Asian crisis, and a time trend to capture the secular shift in Australia's imports towards lower-priced goods from non-G7 countries (particularly those in Asia).

Incorporating these two variables into an otherwise conventional pass-through equation returns the results found in previous studies: changes in the exchange rate (and world prices) are completely passed through to changes in domestic import prices in the long run. Also consistent with earlier findings is the rapid adjustment to equilibrium, shown in Figure 4 by the response of import prices ‘at the docks’ to a 1 per cent depreciation.

In fact, this rapid adjustment has been a stable feature of the pass-through relationship, even during the 1990s. This is demonstrated by estimating the pass-through relationship recursively (that is, we estimate the equation up until the March quarter 1990 and successively re-estimate by extending the sample period by one quarter). The lines in Figure 5 trace the extent to which pass-through of a permanent 1 per cent depreciation is estimated to have occurred by the quarter shown. Clearly, the path of adjustment towards long-run equilibrium has remained remarkably stable. In other words, the relative stability of inflation in Australia cannot be attributed to a reduction in exchange rate pass-through, at least at the first stage. This makes second stage pass-through central to assessments of the direct inflationary consequences of currency movements.

3.1.2 Second stage pass-through

If prices are set as a mark-up on costs, the price of the retail import will be determined by the ‘over the docks’ cost of the import itself and the cost of domestic inputs used in the process of distribution and sale:

where R is the retail import price, C is the cost of domestic inputs, λ is the mark-up and α is the share of the import in total costs.^[15] In this framework, the extent of second stage pass-through is represented by the elasticity of a retail import price with respect to an over-the-docks price. Although the full increase in P (and C) will be passed on to R, the proportional change in R will be less than unity because the imported good is only one element in the total bundle of costs faced by the retailer. In other words, complete pass-through is defined by the share of the imported item in total costs. For Australia, this share appears to be around two-thirds (Prices Surveillance Authority 1989; Dwyer and Lam 1994).

This characteristic of second stage pass-through is important, because in popular discussion, when movements in retail import prices are observed to fluctuate by less than those at the docks, there has been a tendency to claim that pass-through is incomplete. This need not be so. Investigations of second stage pass-through in Australia, that include the experience of the early 1990s, have found it to be complete in the long run. That is, around two-thirds of a change in imports prices (equal to the estimated share of imports in total costs), is eventually passed on at the retail level. But adjustment is very slow, implying that distributors vary their mark-ups sometimes substantially and for considerable periods of time. Furthermore, Dwyer and Lam (1994) found that the mark-up is usually inversely related to changes in the exchange rate so that, in the short run, there is some tendency to absorb the effects of currency depreciation.^[16]

We estimate the second stage pass-through relationship, with retail import prices modelled as a mark-up on landed import prices and unit labour costs; the mark-up is allowed to vary over the cycle.^[17] Again we use a standard error-correction model, as detailed in Appendix C. Our analysis ends in June 1999 because the imported component of the CPI, which we use as our measure of retail import prices, was discontinued at that time; it was replaced with a broader measure of tradeables prices that is not directly comparable.

Initially, we make no allowance for special factors that may have affected the pass-through relationship and our model yields broadly similar results to those in Dwyer and Lam (1994). The actual and fitted values from this basic model are illustrated in Figure 6. The model explains retail import prices reasonably well (even during earlier episodes of exchange rate shocks). However, since mid 1998, actual prices have been less than predicted.

Does this represent a change in the pass-through relationship or is it the result of special factors? In the Australian case, there would appear to be at least some role for special factors. Motor vehicles account for a substantial share of retail imports,^[18] and their prices have been depressed by increased domestic competition in the automotive industry following the efforts of Asian suppliers to expand their share of the Australian market. This culminated in aggressive discounting of motor vehicles sourced from Asia, particularly during the Asian financial crisis.^[19]

We model the second stage pass-through relationship excluding motor vehicles, also detailed in Appendix C. The results are summarised in Figure 7, which compares actual and fitted retail import prices. The extent of over-prediction has been reduced, so that the experience of the late 1990s now looks less unusual. But although the prediction error is not exceptionally large, it remains slightly more persistent than was previously the case. We find that if a dummy variable is employed over the recent period of prediction error (from the June quarter 1998) it is statistically significant.^[20] Since these developments occur at the end of the sample period, it is difficult to determine whether there has been a temporary disturbance to the pass-through relationship, or a permanent change. Indeed, it is too soon to tell.

One interpretation is that there has been a temporary disturbance to the pass-through relationship, perhaps stemming from a widespread but short-lived discounting of goods sourced from Asia following the Asian financial crisis. Alternatively, there may be more pervasive forces at work that are placing sustained downward pressure on either the domestic costs involved with the distribution and sale of imports, or the mark-ups expected by retailers. These forces would imply a structural change in second stage pass-through and diminish the inflationary consequences of a given shock to import prices at the docks.

We can obtain some insights into the possible behaviour of the distributors' mark-up. Deviations of retail import prices from their long-run equilibrium, which are captured by the error-correction term in our equations, imply a variation in the mark-up. We use the error-correction term as a measure of the mark-up.^[21] These terms, from both models, are plotted in Figure 8.

A protracted or unusually large change in the mark-up could suggest the possibility of structural change in the pass-through relationship. However, there is no suggestion of a downward drift in the mark-up during the 1990s. Neither is it clear that the mark-up is now established at a rate below historical norms. The two recent episodes of sharply falling mark-ups follow a period in which the mark-up was high.^[22]

The recent behaviour of retail import prices has been clearly helpful for inflation. While the behaviour is unusual, it is too soon to tell whether it stems from temporary influences or is the result of a more fundamental change in the pass-through relationship.

3.2 Wage Developments

In the mark-up framework, unit labour costs have the largest long-run effect on prices. Furthermore, in the Australian experience, they are passed on more quickly than changes in other costs.^[23] So, for a given rate of labour productivity, wage developments are central to inflation performance. The 1990s witnessed significant changes in the wage-setting system in Australia which have implications for the propagation of wages shocks and thereby inflation.

For much of the past century, wage determination in Australia had two defining features: the bulk of wages were centrally determined, and they were indexed (either partially or fully) to the cost of living.^[24] In the 1990s, however, there was a move towards enterprise-based agreements.^[25] Consequently, at present, roughly 40 per cent of employees are covered by enterprise agreements, 40 per cent are covered by individual contracts while roughly 20 per cent remain in the centralised system (DEWRSB 2000).^[26] The resultant changes in the process of wage bargaining served to undo some longstanding traditions, the first of which was the tendency to preserve wage relativities between workers.

The tradition of preserving wage relativities, known as ‘comparative wage justice’, had the result that an increase in wages in one sector was usually quickly transmitted to other sectors and resulted in a generalised wage increase. However, under the more recent decentralised system, there has been a greater tendency for wage rises in a given sector to be ‘quarantined’ and not lead to a generalised wage increase. The effects of this can be summarised by the coefficient of variation in industry wages (that is, the standard deviation divided by the mean) shown in Figure 9.^[27] By this measure, variation in wages growth between industries appears to have increased in the 1990s.^[28] Consequently, the pockets of high wages growth during the mid 1990s did not become generalised, leaving aggregate wages growth at a rate consistent with low inflation.

Another important consequence of more decentralised wage bargaining is the reduced tendency for automatic indexing of wages to prices, removing a mechanism for the direct transmission of prices to wages. For much of the period of centralised wage determination, there was some form of regular indexation of wages to the cost of living (usually measured by the CPI). In the current system, though, indexation does not generally occur. Rather, it occurs if there is a cost of living adjustment clause in the enterprise agreement or a clause that permits wages to be renegotiated in the event of surprise inflation. However, relatively few enterprise agreements have explicit clauses of this nature. Indexing has, in effect, been replaced by incorporating expected inflation into initial wage demands.

A crude indication of the shift away from automatic wage indexation is given in Figure 10, which attempts to capture the extent to which wages are indexed to the CPI.^[29] Two sets of estimates are presented. One set assumes that informal sector contracts (about which there is little published information) are not indexed. The second set assumes that informal sector contracts are indexed in the same way as other types of wage contracts. Regardless of the assumptions, though, there has been a clear regime change in the 1990s. No longer are the bulk of wages in the economy subject to automatic indexation. This diminishes the transmission of price shocks to wages and the potential for a wage-price spiral.

Along with these changes, there has been an increase in contract duration, which has imparted some inertia into the wage-setting process. This is highlighted in Figure 11, which plots the average duration of formal federal enterprise agreements since the introduction of the new wage-setting system. The duration of new federal agreements registered in each quarter and the duration of all currently active agreements are shown.^[30] The general increase in duration is consistent with an environment of low and stable inflation; with lower uncertainty about inflation, agents are likely to revise contracts less frequently.^[31] At the same time, the nominal rigidity introduced by longer contract duration means that wage changes may represent a more persistent shock to prices than previously. As it happens, in the second half of the 1990s, these shocks have generally been favourable ones, and so have been helpful to the maintenance of low inflation.

The permanency of the regime change in wage-setting arrangements can, however, be overstated. The economies of scale in bargaining are encouraging some labour market participants to seek more coordinated industry-wide bargaining rather than to negotiate at the work place level. If uncertainty about the future path of inflation increases, wage indexation may become fashionable again. And a reduction in macroeconomic stability may encourage more frequent resetting of wage contracts. But notwithstanding these possibilities, the containment of sectoral wage pressures, combined with an increase in the duration of contracts with modest rates of wages growth, has made it easier to maintain the low inflation outcomes of the 1990s.

3.3 Productivity

Another important influence on inflation performance in the 1990s has been the sustained strength of labour productivity growth which, like inflation, has returned to rates not witnessed for more than twenty years.

Figure 12 presents trends in market sector productivity since the mid 1960s. Measured productivity growth tends to vary during the course of a business cycle, as inputs are used more intensively when demand is increasing than when it is slowing. Consequently, we focus on average rates of productivity growth over entire economic expansions, depicted here by the trend lines. Growth rates of both labour and multi-factor productivity are clearly faster than in the economic expansions of the previous two decades, while growth in multi-factor productivity now exceeds that in the 1960s. The pick-up in labour productivity growth tends to attract most attention. However, the growth in multi-factor productivity (which abstracts from the effects of substitution between labour and capital), provides even more compelling evidence that there was an increase in the rate of technological progress in the 1990s.

Workers do not appear to have captured all of the productivity gains in the form of higher real wages.^[32] Consequently, the increase in trend productivity growth has been associated with low growth in unit labour costs which has imparted a clear disinflationary impulse. In fact, in Australia over the past decade, the weakness in unit labour cost growth has resulted in a fall in labour's share of income. This may simply be a consequence of an extended period in which unemployment was above the NAIRU. Alternatively, the strength of productivity growth in the current expansion may not have been fully appreciated by wage negotiators. In this case, the unexpectedly strong productivity growth can be interpreted as having led to at least a temporary fall in the NAIRU, again imparting disinflationary pressure as workers learn about the sustained strength in productivity growth.

While the trend increase in productivity has lowered growth in unit labour costs, it has also increased potential output in the economy, so that supply-side constraints on output, and thereby prices, are less binding. However, the extent to which this has occurred is difficult to determine. Assessments of potential output, and the attendant output gap, are better done in retrospect than in real time. There is the inherent ‘end-point problem’ involved in determining the trend rates of output growth from which we gauge potential, and a tendency for output data to be revised. Consequently, real time assessments of output gaps are often wrong (Orphanides 2000).^[33]

The structural improvement in productivity during the 1990s would appear to have been a helpful influence on inflation outcomes. However, such increases are difficult to recognise at the time they are occurring. But eventually, a structural improvement in productivity growth will no longer be a surprise source of disinflationary pressure. The period of transition, though, could take some time.

Footnotes

See, for example, Cockerell and Russell (1995), de Brouwer and Ericsson (1995) and Beechey et al (2000). [11]

Because in a small open economy importers face perfect elasticity of supply, foreigners will not adjust the foreign-currency price of the import following a change in the exchange rate, so that the domestic price will move in exact proportion to the exchange rate. [12]

Although, researchers examining pass-through of exchange rates to the prices of individual classes of goods find it to be incomplete (see, for example, Menon (1991) with respect to motor vehicles). [13]

And much of it after one quarter. See also Dwyer and Lam (1994). [14]

Assuming a Cobb-Douglas production function. [15]

However, Dwyer and Lam (1994) found that the inverse relationship between the mark-up and the exchange rate was less evident during the large depreciation of the mid 1980s. They concluded that the magnitude of the depreciation appeared to force many firms out of their ‘band of inaction’ so that they passed on higher import prices much more quickly. So despite the fact that the experience of the mid 1980s was itself unusual, it has influenced popular expectations about the behaviour of retail import prices. [16]

In our model, domestic costs are represented only by unit labour costs rather than the more comprehensive cost index used by Dwyer and Lam (1994). [17]

With a weight of 16 per cent in the imported component of the CPI. [18]

For a more detailed discussion, see Reserve Bank of Australia (1999). [19]

See Appendix C. [20]

The mark-up is λ_t = r_t − αp_t − (1−α)c_t, where r is the log of retail import prices, p is log of landed import prices, c is the log of unit labour costs and α and 1−α are the long-run elasticities reported in Appendix C. The mark-up term for the ‘Excluding motor vehicles’ case does not include a dummy variable. [21]

Even though the mark-ups shown here are the error-correction terms of our equations, a very similar pattern of mark-up behaviour in the 1990s can be found from independent estimates of the mark-up obtained by dividing earnings (before interest and depreciation) by sales, and measures of profitability recorded in surveys. [22]

See de Brouwer and Ericsson (1995) and Dwyer and Lam (1994). This may reflect perceptions that changes in wages are permanent while changes in other costs, such as those stemming from exchange rate changes, may be temporary. [23]

This is, of course, a gross simplification. For a detailed discussion of the wage determination system in Australia see Niland (1986) and for the period of the Prices and Incomes Accord between labour and the government that operated in the 1980s, see Lewis (1993). [24]

For a detailed discussion see Wooden (2000). [25]

Workers who are unable to secure wage rises through enterprise bargaining receive ‘safety net’ adjustments of their awards which are determined centrally by the Australian Industrial Relations Commission. [26]

A consistent series of industry-based wages data is not available for the run of years that we wish to consider. We are confined to using the average ordinary-time earnings of adults working full-time in each industry. These data are available from 1983. Before calculating the coefficient of variation, extreme wage changes (that are likely to reflect sampling problems) were trimmed from the distribution. [27]

However, over the period averages shown in Figure 9, the standard deviation of industry wages growth has increased by around a third as much as the coefficient of variation. This confirms that much of the increase in the coefficient of variation has been due to a reduction in the mean rate of wages growth rather than an absolute increase in wage variability. [28]

For details about how these estimates have been derived, see Appendix A. [29]

The Department of Employment, Workplace Relations and Small Business maintains a census of formal enterprise agreements in the federal jurisdiction. [30]

While reduced uncertainty about inflation reduces the need to revise contracts, under the new system of enterprise bargaining in Australia, increased contracts duration is also a function of increased maturity in the bargaining process. [31]

Certainly, real wages (both real product and consumption wages) have grown at a rate less than trend productivity. [32]

For example, in the 1970s, policy-makers thought that output gaps were much larger than they actually were, because of a belated recognition of a productivity slowdown (Orphanides 2000). [33]