PROSPECTS FOR SKILLS-BASED EXPORT GROWTH IN A LABOUR-ABUNDANT, RESOURCE-RICH DEVELOPING ECONOMY

Abstract

In an integrated global economy, specialisation in trade is an increasingly prominent strategy. A labour-abundant, resource-rich economy like Indonesia faces stiff competition in labour-intensive manufactures; meanwhile, rapid growth in demand for resources from China and India exposes it to the ‘curse’ of resource wealth. This diminishes prospects for more diversified growth based on renewable resources like human capital. Using an international panel data set we explore the influence of resource wealth, foreign direct investment and human capital on the share of skill-intensive products in exports. FDI and human capital increase this share; resource wealth diminishes it. We use the results to compare Indonesia with Thailand and Malaysia. Indonesia's reliance on skill-intensive exports would have been greater had it achieved higher levels of FDI and skills. Its performance in accumulating these endowments, and its relative resource abundance, impede diversification in production and trade. We present policy options flowing from these findings.

Acknowledgements

The authors are grateful to Ross McLeod, Thee Kian Wie, two anonymous referees and the series editor for helpful comments on earlier drafts. All remaining errors are our own.

Notes

¹‘Dutch disease’ is ‘[t]he deindustrialization of a nation's economy that occurs when the discovery of a natural resource raises the value of that nation's currency, making manufactured goods less competitive with other nations, increasing imports and decreasing exports. The term originated in Holland after the discovery of North Sea gas’ (<http://www.investorwords.com/>).

²This analysis is a precursor to endogenous growth models in which expansion of high-skill industries has positive productivity spill-overs, which raise returns to skilled labour and induce additional investments in human capital. But human capital investments are financed by profits earned from production in lower-skill industries. So faster growth in lower-skill industries accelerates growth along with structural change (expansion of higher-skill output); conversely, lower world prices for lower-skill manufactures reduce profits, and thus reduce the rate of growth and structural change.

³As Jones and Kierzkowski (2001) put it, ‘an Olympic gold winner in a mixed event, such as the decathlon … might return with no medals if the event is broken down into separate components’.

⁴Collins and Bosworth (1996) provide an excellent overview of the mid-1990s blossoming of scholarly studies on this subject.

⁵Intermediate goods in the ‘High’ category in table 2 have the following 1992 Harmonised System (HS-1992) codes: 8503, 850490, 850690, 850790, 850870, 850990, 851090, 851190, 851290, 851390, 851490, 851590, 851690, 851770, 851890, 8522, 8529, 853190, 853290, 853390, 8538, 853990, 854091, 854190, 854390, 8803, 880400, 9002, 900390, 900590, 900691, 900699, 900791,900792, 900890, 901090, 901190, 901290, 901390, 901490, 901590, 901790, 901811, 901819, 901819, 901820, 901831, 901890, 902290, 902490, 902590, 902690, 902790, 902890, 902990, 903090, 903190, 903290, 903300, 911190, 9114, 9209, 9305, 930690, 930700.

⁶Total merchandise export value is the sum of a country's total merchandise exports to the rest of the world, using SITC codes 00–97.

⁷This measure does not discriminate by type of post-secondary qualification or by quality. We were unable to obtain more detailed data with coverage adequate for a panel of countries.

⁸Obviously this adjustment is very crude: it ignores sectoral capital intensity, bundling of investments with technology or training, and many other relevant phenomena including the allocation of FDI within manufacturing. Some difficulties in constructing cross-country comparisons of FDI are raised in Anderson and Rand (2003). The validity of our results using this adjustment method depends on the adjusted figures being correlated with the true data.

⁹There are in general no unambiguous ex ante reasons to prefer FE or RE estimators, but in the current context there is a case for using RE. The FE estimator uses only ‘within’ varia tion, i.e. variation in the dependent and explanatory variables after removing observation-specific means. In the case where much of the variation in the data is ‘between’, i.e. information contained in the means (across individual observations or time), the FE estimator ignores a relatively large share of the information in the data set. Moreover, because FE attenuates the variance in the regressors, measurement error imposes a greater degradation in the signal to noise ratio; hence a variable measured with error will be more likely to be biased toward zero. On the other hand, RE assumes that the independent variables are uncorrelated with the error terms. The FE estimator will be inefficient under the null hypothesis that RE is correctly specified, because it discards the information in the group means. If the error terms are correlated with the independent variables, then the RE estimator will be inconsistent. Another reason to prefer RE is that we want to include some time-invariant dummy variables because we believe these are important sources of explanatory power and as such will help us to get a better prediction for Southeast Asian countries’ net skilled-export intensity. Accordingly, we use the Hausmann test; for the fully specified static model the null is not rejected, so we prefer the more efficient RE estimator.

¹⁰To save space, the full set of estimation results is available in an online archive at <http://www.aae.wisc.edu/coxhead/papers/CoxheadLiBIES/>.

¹¹Empirically, there are two ways to control for an Indonesia-specific effect. We can include a dummy variable for Indonesia in all of our regressions; doing so barely changes our estimation or counter-factual analysis results (for these results see the URL cited in footnote 10). Alternatively, in our counter-factual analysis in the next section of this paper, we take the difference between Indonesia's actual net skill-intensive export share and our prediction of that share as a country-specific effect that varies over time (see below).

¹²We are grateful to an anonymous referee for suggested improvements in the model specification.

¹³Several recent studies of Indonesian growth also identify important supply-side constraints, including labour market regulations, infrastructural inadequacies and an unstable legal and political setting (e.g. Athukorala 2006; Takii and Ramstetter 2007). Such considerations undoubtedly apply widely in the developing world. Our econometric analysis cannot capture such effects for a panel of countries without the addition of much more (and more detailed) data. It is likely that the FDI measure in our data set reflects at least part of the effect of this set of constraints, as all must serve to discourage inward FDI flows in manufacturing sectors.

¹⁴‘Indonesian manufacturing is steadily climbing the technology ladder. However, global levels of TFP have also improved over the past decades. Hence, when viewed from an international perspective, Indonesia's ascent resembles a standstill on the global escalator’ (Timmer 1999: 93).

¹⁵For details of estimation, see the URL cited in footnote 10.