Productivity gains, technology spillovers and trade: Indonesian manufacturing, 1980-96

Abstract

The importance of R&D spillovers for productivity growth has been well documented in the literature. While studies in the developed country context have focused extensively on sectoral linkages, research on North-South spillovers has largely been confined to the aggregate level. This paper assesses the contribution of international R&D spillovers to manufacturing performance in Indonesia at the sectoral level. Drawing on OECD and Indonesian data sources, we consider two distinct phases: a phase characterised by intense policy regulation (1980-87), followed by a phase of liberalisation and reform (1988-96). Our results indicate that international technology spillovers made a significant contribution to the performance of Indonesian manufacturing, especially after liberalisation. The contribution of technology spillovers to productivity change is influenced by sectoral characteristics and industrial market structure.

Notes

We thank Colin Webb, Agne`s Cimper (OECD), Marcel Timmer (Groningen Growth and Development Centre), Haryo Aswicahyono (Centre for Strategic and international Studies, Jakarta), Yahya Jammal (Boston Institute for Developing Economies) and Sadayuki Takii (International Centre for the Study of East Asian Development, Kitakyushu) for provision of and advice on data. We are grateful to Bart Los, Bart Verspagen and Eddy Szirmai for encouragement and discussion, to participants at the conference on ‘The Empirical Implications of Technology-Based Growth Theories', SOM (Systems, Organisation and Management Research School), University of Groningen, August 2002, and to two anonymous referees for comments and suggestions. Financial support from the Netherlands Organisation for Scientific Research (NWO) is acknowledged.

For examples see Hill (1996) and Simatupang (1996).

Technology contracts between domestic and foreign firms are an important channel of North-South technology diffusion. Data on these contracts are not available for Indonesia, however.

Spillovers are ‘leakages’ of a firm's or industry's R&D efforts to one or more other firms or industries. They may result from economic transactions such as trade, from the presence of foreign firms in the local industry (FDI), or from access to technical and trade journals, for example. See the next section for a discussion on the two types of technology spillovers.

Hill (1996) provides a detailed account of Indonesia's industrial technology landscape.

For a detailed explanation of the two forms of spillovers, see the following section.

We do not include the period before 1980 because of the poor quality of the Indonesian data for these years. In order to avoid the distortions that resulted from the crisis of the late 1990s, we also exclude the years after 1996.

We assume that knowledge spillovers from imported inputs have only limited relevance as a technology transmission mechanism in Indonesia, given the lack of complementary technological and human capabilities (Hill 1995; Lall 1998). Thee (2003) also rules out the possibility of reverse engineering by Indonesian firms.

A detailed discussion of endogenous technological change in growth theories can be found in Schneider and Ziesemer (1995), for example.

This can be by such means as reverse engineering, the exploitation of knowledge from academic and trade journals, turnover of researchers, and licensing. Even though licensing is an economic activity, it can generate knowledge externalities if the borrower builds on the licensed technology to generate technologies that are new to it.

Fagerberg and Verspagen (2000) illustrate why econometric attempts to measure spillovers, until the work of Coe and Helpman (1995), failed to capture the presence of international spillovers. They show that international spillovers have been found to exist by studies that used panel data estimations with country-specific dummies. The bottom line, as they point out, is that the countries that possess high levels of absorptive capacity benefit technologically from their ‘backwardness’, while those with low levels do not.

^anec = not elsewhere classified.

We note that any measure of concentration as calculated from plant level data understates the true concentration. This is because it ignores the fact that many of Indonesia's large conglomerates have vast holdings across many industries, so that nominally distinct firms may be owned by the same group (Hill 1996).

We do not report the changes in the number of plants, owing to space constraints.

We do not consider domestic R&D, because data are not available for all the years covered by the study. This omission does not appear to prejudice the results, because of the small degree of domestic R&D, and because most R&D spending in Indonesia is undertaken by public research laboratories, mainly for product certification, training and testing (Thee 1998).

We are aware of the critique by Keller (1998) of the usefulness of trade weights. However, the results of Coe and Hoffmaister (1999) suggest that technological spillovers are greater when a country trades with another that is technologically more advanced.

We define foreign-controlled plants as those with a foreign ownership of 10% or more. This is based on the International Monetary Fund (IMF) view that ownership of at least 10% implies that the direct investor is able to influence, or participate in, the management of an enterprise. Absolute control by the foreign investor is not required.

These countries are Australia, Canada, Denmark, France, Germany, the UK, Italy, Japan, the Netherlands and the USA.

We followed this two-step merging procedure rather than stage two alone because the plant identification codes are not completely accurate.

The two plant-level data sets are beset with flaws such as duplicate observations, and even duplicate plant-identification codes. Most of these result from the BPS practice of accounting for the missing data of plants that do not report data for some years by using the data of plants with similar characteristics. We removed observations with the same variable values for output, value added and labour.

Aswicahyono (1998) and Timmer (2000) follow the same approach.

The calculated F-statistic is highest when the cut-off year is 1987 rather than 1985, 1986 or 1988.

These results are not reported, but are available on request from the authors.

We excluded the time trend in the EY estimation to avoid any bias resulting from a fixed effect in the second stage. Note that, during this stage, explanatory variables are presented in their first differences, and therefore the time trend takes a constant value.

The OLS results can be found in our working paper (Jacob and Meister 2004).

The four subsamples for which OLS estimation yielded negative values for the returns-to-scale parameter are the medium technology samples for the full period, the preliberalisation and post-liberalisation phases, and the low technology sample for the pre-liberalisation phase.

Griliches and Mairesse (1984) report negative returns to scale and put forward a number of explanations. Los and Verspagen (2000), in line with our own findings, have demonstrated that non-stationarity of the variables is a possible cause for negative returns to scale under OLS estimation.