The Treasury

Global Navigation

Personal tools

Treasury
Publication

Past, Present and Future Developments in New Zealand's Terms of Trade - WP 06/09

3  Trends and volatility in New Zealand’s terms of trade

The previous section described the current views in the literature surrounding the impact of terms of trade trends and volatility on economic growth and briefly looked at it from a New Zealand context. This section looks at the historical trend and volatility in New Zealand terms of trade to see if they have changed over time. It follows very closely the methodology used by Gillitzer and Kearns (2005) in their study of long-run trends and volatility in Australia’s terms of trade. The description and sources of the data for this section are explained in Appendix 1.

3.1  Long-run trend

3.1.1  Has New Zealand’s terms of trade declined over time?

Tables 2 and 3 below display the composition of New Zealand’s merchandise trade over the past ten years. It is clear that New Zealand’s exports are dominated by primary commodities while its imports are dominated by manufactured goods. For this reason, the Prebisch-Singer hypothesis implies that the terms of trade should have declined over history and should continue to decline. This would have implications for the economy, some of which were discussed in the previous section.

There have been some significant events in New Zealand’s economic history that have had an impact on the terms of trade. One of these events was when the UK formally joined the European Economic Community (EEC) in 1973. At the time the UK was New Zealand’s biggest export destination taking 72% of lamb exports, 73% of butter exports, 66% of cheese exports and approximately 20% of wool exports.[7] The UK joining the EEC had the effect of dramatically reducing the demand for New Zealand products in the UK and led to a significant reduction in export prices. It required New Zealand to identify new markets for its primary exports as well as develop new merchandise in order to remain competitive overseas. Other events at the time which may have affected the terms of trade were the commodity price boom of the early 1970s, followed by the first oil shock in late 1973.

Table 2 – New Zealand’s Merchandise Exports Composition
Exported Product
year ended March 31
1995 2000 2005
value $m % of total value $m % of total value $m % of total
Non-commodity manufactures 4,345 20.8% 5,528 22.5% 8,434 27.1%
Dairy, casein and caseinates 3,337 15.9% 4,459 18.1% 5,678 18.3%
Meat and edible offal 2,689 12.9% 3,198 13.0% 4,688 15.1%
Forestry products 2,515 12.0% 2,942 12.0% 2,957 9.5%
Fruit 764 3.6% 1,059 4.3% 1,355 4.4%
Seafood products 1,084 5.2% 1,183 4.8% 1,126 3.6%
Aluminium and aluminium articles 826 3.9% 1,013 4.1% 1,071 3.4%
Wool 1,299 6.2% 760 3.1% 698 2.2%
Other goods exports 4,065 19.4% 4,473 18.2% 5,082 16.3%
Total 20,923 100.0% 24,615 100.0% 31,088 100.0%

Source: Statistics New Zealand

Table 3 – New Zealand’s Merchandise Imports Composition

Imported Product
year ended March 31 (VFD)

1995 2000 2005
value $m % of total value $m % of total value $m % of total
Minerals, chemicals and plastics 4,322 22.6% 6,148 23.4% 8,611 26.0%
Machinery and mechanical appliances 5,092 26.6% 6,238 23.8% 7,865 23.7%
Vehicles and aircraft 3,267 17.1% 5,171 19.7% 5,910 17.8%
Other manufactures 1,528 8.0% 2,124 8.1% 2,488 7.5%
Metals and articles of metal 1,233 6.4% 1,489 5.7% 1,946 5.9%
Textiles and textile articles 1,165 6.1% 1,451 5.5% 1,554 4.7%
Other goods imports 2,529 13.2% 3,604 13.7% 4,764 14.4%
Total 19,136 100.0% 26,226 100.0% 33,139 100.0%

Source: Statistics New Zealand

Another influential period was the Korean War in 1950. This saw the demand for wool increase dramatically with the price of wool rising by over 150% in two years. The overall impact of this wool price shock saw the goods terms of trade reach a record high in 1951 before falling once the war was over.

In order to test whether any of these periods had a large influence on the trend or level in New Zealand’s terms of trade, the Andrews and Ploberger (1994) test for a structural break was used.[8] However, it yielded inconclusive results and found no evidence of a structural break in the terms of trade (expressed in logarithms) for annual data from 1900 to 2005.[9]

Following the methodology of Gillitzer and Kearns (2005), unit root tests were performed to test the stationarity of the data. Although the Andrews and Ploberger (1994) test found no evidence of any structural break, the sample is split into two periods to display how the trend in New Zealand’s terms of trade differs depending on the period examined. These periods are 1900 to 1973 and 1974 to 2005.[10]

The results of the Augmented Dickey-Fuller (ADF) and Phillips-Perron (PP) unit root tests are displayed in Table 4. They show that generally New Zealand’s log terms of trade are stationary. However, the ADF unit root test is unable to reject that the terms of trade have a unit root for the period 1974 to 2005 while the PP test can only reject the hypothesis of a unit root when a trend is included. This may be a result of the relatively small sample size and the fact that the terms of trade has generally been increasing over this period.

Table 4 – Unit Root Tests for Log Goods Terms of Trade
  Augmented Dickey-Fuller Phillips-Perron
  Intercept Trend Intercept Trend
1900-2005 *** *** *** **
1900-1973 ** * ** **
1974-2005 ***

***, ** and * denote rejection of the null hypothesis of a unit root at the 1%, 5% and 10% significance levels respectively.

To estimate the linear trend in the log terms of trade, . , (3) is estimated below. The results are displayed in Table 5.

(3)      .

Table 5 shows that over the period 1900 to 2005 there has been a statistically insignificant downward trend in New Zealand’s goods terms of trade of 0.1% per year. This is an interesting result because it appears to reject the Prebisch-Singer hypothesis that New Zealand’s terms of trade should be trending downwards as a result of the composition of its exports and imports. If the sub-samples are examined, the period from 1900 to 1973 has a statistically insignificant trend, while the period 1974 to 2005 reports a statistically significant upward trend of 0.6% per year in the goods terms of trade. This appears to fit with the argument posed by Kellard and Wohar (2006) in that although a single trend may be present (insignificant in this case), there may be periods in which the trend - and the sign in particular - are different. The results show that for the past 30 years New Zealand has experienced an upward trend in its terms of trade. Consequently, this leads to a different interpretation of the results than if the entire 1900 to 2005 period is examined. However, it is not surprising that the terms of trade have trended upwards over the 1974 to 2005 period given that the date chosen as the break was close to the period where New Zealand experienced some significant downward shocks to the terms of trade that may have meant they were artificially low.

Table 5 – Trend in Log Goods Terms of Trade
  1900-2005 1900-1973 1974-2005
α 6.997*** 6.941*** 6.387***
(0.043) (0.054) (0.132)
β -0.001 0.001 0.006***
(0.001) (0.001) (0.001)
Q(1) 56.324*** 37.782*** 2.735*
Q(5) 78.065*** 54.359*** 4.099

***, ** and * denote rejection of the null hypothesis at the 1%, 5% and 10% significance levels respectively. Newey-West standard errors are presented in parentheses. Q(1) and Q(5) are the Ljung-Box statistics for autocorrelation at 1 and 5 lags respectively.

However as Gillitzer and Kearns (2005) found for Australia, this method of estimating the trend reports considerable autocorrelation in the residuals of (3). This is shown by the highly significant Ljung-Box statistics reported in Table 5 and suggests that there is a level of persistence in the terms of trade (with the exception of the 1974 to 2005 period). To correct for this autocorrelation in the residuals, a lagged dependent variable is added to the regression as shown by (4) where ρ is the autoregressive parameter.

(4)      .

The results of (4) are displayed in Table 6 and show that the inclusion of the lagged dependent variable reduces the trend coefficient across the entire sample as well as the sub-samples. For the 1900 to 2005 period the downward trend is reduced from 0.1% per year in (8) to 0.0% per year (-0.0001%) in (4) (and it still remains insignificant). The trend is still also statistically insignificant for the period of 1900 to 1973 and the upward trend for the period of 1974 to 2005 has fallen to 0.5% per year (from 0.6%) but is still significant. The long-run parameter, defined as β/(1-ρ), shows that the above results apply. That is, the trend for the 1974 to 2005 period is positive and highly significant at 0.8% per year.

Table 6 – Trend in Log Goods Terms of Trade (Allowing for Persistence)
  1900-2005 1900-1973 1974-2005
α 1.904*** 1.889*** 4.272**
(0.403) (0.518) (1.643)
β -0.000 0.001 0.005***
(0.000) (0.001) (0.002)
ρ 0.728*** 0.727*** 0.310
(0.057) (0.074) (0.258)
β/(1-ρ) -0.001 0.001 0.008***
(0.001) (0.001) (0.002)
Q(1) 3.159* 2.231 3.923**
Q(5) 6.818 7.057 11.476**
ρ(Andrews) 0.754 0.846 0.396
Half-Life 2.455 4.145 0.748

***, ** and * denote rejection of the null hypothesis at the 1%, 5% and 10% significance levels respectively. Newey-West standard errors are presented in parentheses. Q(1) and Q(5) are the Ljung-Box statistics for autocorrelation at 1 and 5 lags respectively. ρ(Andrews) is Andrews (1993) median-unbiased estimator of ρ. β/(1-ρ)is the long-run parameter.

Also reported in Table 6 is the median-unbiased estimate of ρ based on Andrews (1993). This estimate corrects for biases associated with ordinary least squares estimation with lagged dependent variables (like the form of (4)). The Andrews (1993) estimate of ρ in all three sample lengths is higher than that estimated by least squares to correct for the downward bias. The half-lives of unit shocks are also presented in Table 6 and these show that shocks to the terms of trade do not appear to be persistent (consistent with the finding of stationarity). Shocks appear to dissipate after approximately four years for the period of 1900 to 1973 and have decreased to less than one year for the period 1974 to 2005. Applying this result to the findings of Obstfeld (1982) and Kent and Cashin (2003) suggests that shocks to the terms of trade impact on the New Zealand economy as suggested by the Harberger-Laursen-Metzler effect as they are relatively short-lived.

As reported above, there is little evidence that New Zealand’s terms of trade have declined over time, which contradicts the Prebisch-Singer hypothesis. As discussed above, there is no statistical evidence that there were any structural breaks present. However, when examining the sub-periods, there is strong evidence that the terms of trade have experienced an upward trend over the past 30 years.

Following Gillitzer and Kearns (2005) methodology, New Zealand’s terms of trade is compared with the ratio of world commodity prices and world manufacturing prices. This series is taken from Grilli and Yang (1988) and is extended to 2005 using IMF commodity price data. Figure 4 below displays the series since 1900.

Prior to the mid 1970s, New Zealand’s goods terms of trade and the ratio of world commodity prices to world manufacture prices (relative commodity prices) had a similar downward trend as well as similar cycles. However, following this period the series have diverged. The terms of trade have increased while relative commodity prices have continued decreasing at a more rapid pace. This development has important implications for the New Zealand economy because it suggests that over the last 30 years New Zealand has benefited from a terms of trade that is higher than implied by relative world commodity prices.

Figure 4 – Relative Commodity Prices and New Zealand’s Terms of Trade
.
Sources: Grilli and Yang, IMF, NZIER, Statistics New Zealand

Notes

  • [7]These figures were taken from Dalziel and Lattimore (2004).
  • [8]I thank Kam Szeto for performing this test for me.
  • [9]However, using quarterly data from 1950-2005, the test found evidence of breaks in the trend as well as other coefficients. The results of this test are available from the author.
  • [10]These sub-periods were chosen by graphical observation and also with a belief that the UK entering the EEC in 1973 had a large impact on the terms of trade. However, as mentioned above, the oil shocks of the 1970s also occurred at this time, therefore not all of the initial downward movement in the terms of trade can be attributed to the UK joining the EEC.
Page top