RDP 9612: External Influences on Output: An Industry Analysis 4. Some Implications

There are four implications which flow from the analysis above.

4.1 New Information?

The first concerns the information that can be used to improve our understanding of the economic process and economic prospects. It is well accepted that foreign growth contains information about Australian growth. The analysis here suggests that US industrial production may contain information not only about US economic growth more generally, but also that knowledge of particular manufacturing sectors can help in forecasting domestic production. This would seem to hold for the production of basic and fabricated metals, chemicals, machinery and miscellaneous manufacturing, non-metallic minerals and paper products. This is a proposition to be tested.

4.2 The Relative Importance of Foreign and Domestic Influences

As the analysis above showed, while foreign output is important in some sectors, so too are domestic influences, particularly in the short run. Figure 3 summarises the relative importance of domestic and foreign shocks at the most general industrial level. It is a scatter-plot of the contemporaneous correlation of the one-digit Australian sectoral output growth rates with the growth in total Australian output (excluding the particular sector) and with growth in the corresponding US sector. When the outcome is above the 45 degree line, Australian sectoral output is affected more by contemporaneous events in the home economy than in the corresponding US sector. It appears that growth in Australian sectoral outputs is more related to growth in the rest of Australia than with growth in the corresponding industrial sector in the United States, at least on a contemporaneous basis.

This can also be explored at a more technical level. The international literature on the topic of the relative importance of international and domestic effects on output is extensive, largely because it has developed in response to the question raised by so-called ‘real business cycle’ economists of whether ‘shocks’ are explained by national fiscal and monetary policies or by technological change. The literature indicates that, for the G7 countries at least, cross-country sectoral output links are weak, at least in comparison to domestic influences. Stockman (1988) reported that changes in industrial production in European countries seem to be tied to what is happening in the home country itself, rather than what is happening to the industry in a range of other countries. Other papers have reached a similar conclusion (Costello 1993; Engle and Issler 1995; Helg et al. 1995; and Cecchetti and Kashyap 1996).

Stockman's (1988) panel data estimation method, explained in Appendix C, is used to assess the relative importance of national domestic and foreign sectoral shocks on domestic sub-sectoral manufacturing output from the OECD's STAN database from 1975 to 1994. The set of countries used for this test includes Australia, Canada, France, Germany, Korea, Japan, the United Kingdom and the United States. Using all the countries in our sample, more of the variation in sectoral output is explained by home-country effects (43 per cent) than foreign-industry-sector effects (30 per cent), even though the latter are clearly important. The inference is, therefore, that even though foreign output seems to explain a lot about domestic sectoral output, domestic influences like monetary and fiscal policy are also critically important. The results are similar (41 per cent and 32 per cent respectively) when Australia is excluded from the set of countries, which, on the face of it, suggests that Australia is much the same as other countries in terms of the relative importance of domestic and foreign influences. More formal tests fail to reject the hypothesis that Australia exhibits the same relationship as the other countries in the sample.

One caution in interpreting these results is that they are based on an analysis of growth rates, and so are restricted to short-run relationships. The times-series analysis in Section 3 indicated that the United States is central in the long run, either because of aggregate demand effects or because of direct sectoral linkages. But, in the short run, domestic demand seems to dominate.

4.3 Controlling for Other Influences

To test the robustness of the link with US sectoral outputs, alternative specifications of two-digit manufacturing output were estimated. Apart from the influence of domestic demand and foreign (aggregate and sectoral) output, industrial output may also be sensitive to the real exchange rate, the terms of trade and the real interest rate (Gruen and Shuetrim 1994). Unfortunately, these other variables could not be included in the analysis in Section 3, because there were too few degrees of freedom. To gain these extra degrees of freedom, we use quarterly data. But, since US industrial output is not published on a quarterly basis, we only include US aggregate output.^[8] The preferred specification for each industry is derived from the following unrestricted error-correction model:

where notation is the same as for equation (1), r is the real cash rate, farm is farm output, rer is the log real exchange rate, tot is the log terms of trade for all goods and services and eraⁱ is the effective rate of assistance for industry i. The analysis is restricted to manufacturing sub-sectors. Full results are reported in Appendix D.

The results indicate that international links are fairly robust to alternative specifications. Foreign aggregate output has significant and economically substantial impacts in eight of the 12 manufacturing sectors, even after controlling for domestic income. There is a long-run relationship with US aggregate output in three cases (basic metals, machinery and wood products), and a short-run relationship in five cases (textiles, clothing, paper products, fabricated metals and miscellaneous manufacturing). While there are fewer long-run links at the quarterly level than at the annual level (three compared to eight), these results are not directly comparable to those in Table 2, since the relationship here is with aggregate output. When the manufacturing equations in Table 2 are estimated with US aggregate output in place of US sectoral output, the explanatory power of the equations usually falls substantially, the dynamics terms all become insignificant, and the long-run coefficients becomes less significant, and in the case of wood products and furniture, insignificant.^[9] Shifting the specification to aggregate US output itself substantially weakens the links with the United States.

The foreign impact is usually not contemporaneous, but delayed a quarter or two, which differs from models of aggregate output like Gruen and Shuetrim (1994). The lags of domestic aggregate output in the equations are relatively short, while those for foreign output are relatively long. This is consistent with the view, very loosely speaking, that domestic aggregate output captures demand effects while foreign aggregate output captures supply effects.

Other external variables, like the real exchange rate and the aggregate terms of trade, also have a substantial effect on some sectors. A real appreciation has a statistically significant, and economically substantial, negative impact in eight manufacturing sectors, including transport equipment, other machinery, and miscellaneous manufacturing. It appears to have little effect in sectors producing fairly simply transformed manufactures such as food, textiles, chemicals or basic metals. A rise in the terms of trade supports manufacturing, but the effect is concentrated in a few sectors, notably those involved in commodity processing or the production of capital goods or consumer durables, like wood and furniture, non-metallic mineral products, basic metal products and fabricated metal products.

Not surprisingly, domestic monetary policy also affects output. In the simple modelling framework used in this paper, there are three possible transmission routes by which policy affects sectoral outputs. Apart from the direct interest rate effect, there are two indirect or feedback effects. The first is through policy's influence on the rest of output, and hence on domestic demand, and the second is the effect of policy on the exchange rate. The estimates in Appendix D suggest that monetary policy has a relatively fast and large impact on manufacturing industries. In many manufacturing industries, policy affects output in the same quarter or with just one lag, unlike for aggregate output where the effect seems to take up to two quarters. There is a range of responses in the various sub-sectors. The direct effect is important in most industries, with the largest effect being in other machinery, which produces investment goods and consumer appliances. Direct effects are not found for wood and furniture, for paper, printing and publishing, for chemicals, or for miscellaneous manufacturing, with the effect of policy in these industries wholly captured in the exchange rate and income channels.^[10] The indirect income effect is strongest in food and fabricated metals, while the exchange rate effect is strongest in non-metallic mineral products and machinery.

4.4 Aggregate or Compositional Effects?

While US sectoral output clearly affects production in some manufacturing sub-sectors in Australia, it is unclear whether these links affect aggregate Australian output, as opposed to merely changing the composition of output. If the latter is the case, then total income is determined by some process other than through the international links that we have identified. This can be tested by drawing on the results from the panel data estimation. In particular, the panel data estimation provides a time series of shocks to each manufacturing sub-sector which are common to all countries. These represent ‘foreign shocks’. These shocks are added to obtain a variable measuring foreign shocks to manufacturing, where the weight is the value of the sub-sector in Australian manufacturing outputs. This annual series of shocks is converted to a quarterly series, and added as an explanatory variable in an equation for aggregate Australian output:

where y^agg is aggregate output, r is the real cash rate, farm is farm output, shock is the shock to manufacturing industry described above, and a tilde denotes the foreign sector.

If sectoral links to foreign output explain movements in aggregate output then we would expect a positive coefficient on the shock variable: an increase in manufacturing due to a foreign shock increases aggregate output. The full results of the regression are reported in Table C.3 in Appendix C. A positive coefficient is found, and the size of the coefficient is larger than the share of manufacturing in total output, although this difference is not statistically significant. This suggests that foreign-sourced shocks to manufacturing industries may, at least in the short run, have multiplier effects on aggregate output. This evidence supports our hypothesis that sectoral links do have a role in explaining movements in total Australian GDP.

Footnotes

Although manufacturing output is a component of non-farm output, each individual manufacturing industry is only a very small proportion of non-farm output, so the simultaneity problem that arises is very small. We also substituted OECD output for US output, with little effect on the results. [8]

These results are not reported but are available from the authors on reques. [9]

The perverse result for textiles is picking up a spurious correlation between recent low real interest rates and a dramatic fall in output in this industry related to tariff reductions. [10]