RDP 8910: An Analysis of the Determinants of Imports 3. Analytical Framework

In most studies of import demand, the quantity of imports is modelled as a function of the price of imports relative to domestic prices and of some activity variable. Many studies also attempt to incorporate cyclical influences on imports.

The demand model we estimate comes from simple production theory.^[2] In this model there are assumed to be three sectors in the economy in which importables (m), exportables (x) and non-traded goods (n) are produced. The excess demand function for importables (d_m) – i.e. the demand function for imports – is thus the difference between domestic demand for (D_m) and supply of (S_m) importables. It is a function of three prices (p_m, p_x and p_n) and nominal income (Y), and is subject to a domestic production capacity constraint (CU), i.e.,

Assuming this excess demand function is homogeneous of degree zero in income and prices, it follows that we can normalize it by any one, or an index, of these prices. Thus we define a new function d_m^* by:

In many studies, p_x/p_n is excluded from the estimation because it is claimed it is not a significant explanator. However, there are two reasons why we believe it may help explain import growth. First, it picks up the effect of terms of trade movements. When the price of our exports rises, our capacity to consume imports increases even though our actual GDP has not increased.^[3] Secondly, we expect that movements in the price of exports could pick up sector specific demand effects. Since about 50 per cent of Australia's imports are intermediate goods, and a further 25 per cent are capital goods, we thought it likely that movements in the price of exports, which would make production in the exportable sector more profitable, would have a significant effect on the demand for imports.^[4]

Explanatory Variables

From the evidence presented in Section 2 above, both relative prices and activity are very important determinants of the demand for imports. However, there are a variety of measures which may be used as a proxy for these variables. For this paper several alternatives were investigated for each variable before arriving at a preferred equation.

Possible Demand Variables

The two main candidates for activity variables are gross national expenditure (GNE) and gross domestic product (GDP). There are valid reasons for choosing either measure: GNE may be preferred if it is thought that the demand for imports should be related to domestic spending on all goods rather than to the sum of domestic and foreign spending on domestic goods only. Alternatively, if imports are mainly intermediate goods used as an input to production, GDP may be a more appropriate measure as it may be more reasonable to treat imports as a function of domestic output rather than spending.

In a long run of data, GNE and GDP are highly correlated. We use both as potential demand explanators.

Possible Price Variables

The first of the relative price terms in our import demand function should compare the price of importables with the price of non-traded goods. However, reliable price indices for domestically produced importables and non-tradeables do not exist, therefore the price of all domestically produced goods is generally used as a proxy for the price of non-tradeables.^[5] We use the GDP deflator.^[6]

For the price of imports we use the endogenous import price deflator when modelling endogenous imports, and the merchandise import price deflator when modelling total merchandise imports. An alternative measure is the import price index. We did not use this measure for two reasons. First, it is only available in a consistent form since 1981/82 which reduces our sample period considerably. Secondly, the import price index is a fixed weight price index, weighted using the pattern of Australian imports during the three years to June 1981. We doubted the relevance in the late 1980s of those weights.

The price deflator for total merchandise exports was used to proxy for the price of exportables.

Possible Capacity Variables

There is reason to believe that not all growth in imports can be explained by GNE and relative price movements. As the economy reaches full capacity, there is likely to be a spillover of excess demand into imports. Overtime is used to model capacity constraints.

Footnotes

This framework is used in Hall et al.(1989). [2]

An alternative would be to use a measure of GDP adjusted for the terms of trade movements, however, this would make interpretation of the import price elasticities difficult. [3]

An alternative model of import demand, suggested by Goldstein et al. (1980), assumes that consumers first allocate their expenditure between tradeables and non-tradeables, and then allocate their expenditure between imports and domestically produced importables. It then follows that the demand for imports is independent of the price of non-tradeables and exportables. The relative price of imports to domestically produced importables is thus the sole relative price term entering the model. There are two problems with this approach. Firstly, the assumption of strong separability in preferences is very restrictive and whether it actually explains individuals' behaviour is subject to debate. Secondly, we do not have a reliable index of the price of domestically produced importables, making empirical estimation difficult. [4]

A consequence of this is that the elasticity of demand for imports with respect to the relative price of importables to non-traded goods is constrained to be the same as the elasticity of demand for imports with respect to the relative price of importables to domestically produced traded goods. [5]

An alternative is suggested by Dwyer (1988) – this involves removing movements in import prices from the CPI to isolate movements in non-traded goods prices. On preliminary estimation we found this series was not very different from the GDP deflator series. [6]