Export-Led Growth in Chile: Assessing the Role of Export Composition in Productivity Growth

Studies about the export-led growth hypothesis (ELG) acquired criticism when it used cross-country correlations. According to Abu-Quarn and Abu Bader (2004), an observed positive correlation of exports in the growth equation can be compatible with causality running from growth to exports. It also assumes a common economic structure and similar production technologies across countries. In response to such criticisms, more recent econometric studies used time-series data from individual countries to investigate the causal relationship between exports and growth by means of Granger-type causality tests.

The objective of this paper is to carefully investigate the long-run relationship between exports and growth. It applies time series techniques and estimates an export-augmented production function – instead of traditional two-variable causality relationship. Chilean time series data (1960-2001) were chosen as a case study because it experienced high long-run growth, its exports (mainly composed of primary products) grew very rapidly, and it is extremely vulnerable to fluctuating commodity prices.

Theoretical Background and Empirical Model

The ELG hypothesis suggests that export expansion is a key factor in promoting long-run economic growth. From a demand side perspective, it can be argued that export markets are limitless and they do not involve restrictions. They can be a catalyst for income growth. In addition, Riezman (1996) proposed that export expansion may indirectly affect growth by providing the foreign exchange that allows for increasing levels of capital goods imports.

Exports can increase productivity 1) by promoting specialization in sectors in which a country has comparative advantage, and lead to a reallocation of resources from the relatively inefficient non-trade sector to the more productive export sector; 2) by offering larger economies of scale (Helpman and Krugman 1985); and 3) through dynamic spillover effects on the rest of the economy (Feder 1983). The ELG hypothesis implies that export growth will lead to economy-wide productivity growth.

In contrast to this, authors have argued that primary exports are an obstacle to greater productivity growth because: they do not offer sustainable potential for knowledge spillovers while they draw resources away from the externality-generating manufacturing sector (Matsuyama 1992); and because they are subject to extreme price and volume fluctuations.

The empirical model starts with a simple neoclassical production function where the aggregate production of the economy at time t is equal to the product of the level of total factor productivity, the capital stock, and the stock of labour. To investigate if and how manufactured and primary exports affect economic growth via increases in productivity, total factor productivity was expressed as a function of manufactured exports, primary exports, and other exogenous factors. The natural log was taken and the ‘economic influence’ of exports on output was separated from the influence incorporated into the ‘growth accounting relationship’ by using the aggregate output, net of primary and manufactured exports instead of total output. The final equation has the form: LNYt = c +αLKt + βLLt +δLCMt +γLIXt + ρLPXt + et. This is estimated to determine the impact of increasing manufactured exports and primary exports on economic growth via increases in productivity.

Data and Econometric Methodology

The data spans from 1960 to 2001. They were gathered from the Indicadores económicos y sociales de Chile 1960-2000 and the Boletínes mensuales published by the Chilean Central Bank. The variables CMt, IXt, and PXt represent real imports of capital goods, real exports of manufactured goods, and real exports of primary products respectively. The non-export output, NYt, is measured by real Chilean GDP net of primary and manufactured exports. Kt is the Chilean capital stock in real terms. Non-export GDP, capital stock, capital goods imports, exports of manufactured products, and primary products are evaluated in Chilean pesos at constant 1996 prices. The labour variable, Lt, is represented by the total number of people employed each year.

The variables were tested for unit roots to verify their order of integration. The unit root test developed by Perron (1997) was used because standard unit root tests are biased in favour of identifying data as integrated if there are structural changes. This procedure permits a formal evaluation of the time series properties in the presence of structural breaks at unknown points in time. If all variables are found to be I(1), the second step is to test for the existence of a cointegration relationship between them by applying the standard Engle-Granger (1987) two-step estimation procedure. The full information maximum likelihood (FIML) cointegration approach developed by Johansen (1995) was used in addition to the Engle-Granger method since no final judgement can be passed on the significance of the estimated coefficients due to non-normality of the distribution of the estimators. The next step is to test for weak exogeneity of the long-run parameters by employing a weak exogeneity testing approach. The last step involves checking the robustness of the cointegration estimates.

Empirical Analysis

The results of both the Perron (1997) and the Kapetanios (2002) unit root test show that LNYt, LKt, LLt, LCMt, LIXt and LPXt are integrated of order one. To test for cointegration, the Engle-Granger (1987) and Johansen (1995) methods were both used. The cointegration regression Durbin-Watson and the ADF test statistics suggest that the null hypothesis of no cointegration can be rejected at least at the 5% significance level. Under the Johansen method, the trace and the maximum-eigenvalue statistics with the corresponding critical values suggest that the null hypothesis of no cointegration can be rejected at the 5% and the 1% significance level, but not the null of at most one cointegrating vector.

The long-run relation that was obtained implies that Chilean non-export GDP increases by 0.033 percent in response to a one percent increase in manufactured exports. In contrast, a one percent increase in primary exports leads to a 0.428 percent decrease in non-export GDP. This result suggests that manufactured exports promote economic growth via increases in productivity. In contrast, primary exports appear to have a negative impact on total factor productivity.

The weak exogeneity tests suggest a long-run feedback relationship between non-export GDP, capital stock, capital goods imports, exports of manufactured products and primary products since all these variables are endogenous to the system as these variables can be regarded as endogenous to the system. The only variable that is weakly exogenous to the long-run relationship is labour. The robustness of the cointegration estimates was checked and results show that statistically valid inferences can be drawn from the estimated long-run elasticities.

The effect of manufactured exports on non-export GDP is significantly positive. A one percent increase in manufactured exports causes a 0.042 percent increase in non-export GDP. The effect of primary exports on non-export GDP is found to be strong and negative, implying that Chilean non-export GDP decreases by 0.459 percent in response to a one percent increase in primary exports.

Summary and Conclusions

The results of the production function estimation indicate that (1) there is a long-run relationship between capital, labour, capital goods imports, manufactured exports, primary exports and non-export GDP; (2) there is also a long-run Granger causality running from capital stock, aggregate employment, capital goods imports, and exports of manufactured products and primary products to non-export GDP, where capital stock, capital goods imports, exports of manufactured products and primary products are also endogenous; (3) However, primary-product exports have a statistically negative impact, while manufactured-product exports have a statistically positive impact on non-export GDP.

This result is robust to different estimation techniques. The outcomes can be used as evidence of productivity-enhancing effects of manufactured exports and of productivity-limiting effects of primary exports. The first result may be due to the problem of unstable commodity export prices and earnings. Manufactured exports might offer greater potential for knowledge spillovers and other externalities than primary exports.

While primary and manufactured export earnings contributed to the Chilean national income, exports of manufactured products have been important for productivity and for long-run economic growth. However, before policy recommendations can be made, further research on the relationship between composition of primary and manufactured exports, structural change in the composition of exports, and economic growth must be done first.

Source:

Herzer, Dierk, Felicitas Nowak-Lehmann and Boriss Siliverstovs, “Export-Led Growth in Chile: Assessing the Role of Export Composition in Productivity Growth” The Developing Economies, XLIV-3 (September 2006), pp. 306-328.