Firm- and Plant-level Analysis of Multinationals in Southeast Asia: the Perils of Pooling Industries and Balancing Panels

Eric Ramstetter

December 2009


This paper uses micro data and published compilations of micro data to estimate shares of multinational corporations (MNCs) in Southeast Asian manufacturing. It first shows that MNC shares tended to be largest in Singapore, intermediate in Malaysia and (recently) in Vietnam, and lowest in Thailand and Indonesia. Shares tended to decline in Singapore and Thailand, were relatively constant in Malaysia, and increased in Indonesia and Vietnam. Shares of majority foreign MNCs also increased conspicuously in Indonesia and Thailand as MNCs bought out local partners in joint ventures after the Asian crisis. Second, it highlights how MNC shares were always lowest in terms of the number of plants or establishments, or in other words, how MNCs tended to be larger on average than local firms or plants. MNCs also tended to account for larger shares of production than employment, and even larger shares of exports. Hence MNCs tended to have relatively high labor productivity and export propensities. Because these simple comparisons do not account for other influences on productivity, wages, or exporting, for example, the paper also describes how micro-data have been used to analyze productivity, wages, and export propensities. This literature suggests that productivity differentials were generally positive but often statistically insignificant, especially at the industry level. Wage differentials were also positive and more often significant, but the largest and most consistent differentials are observed in export propensities. Third, the paper also reviewed literature suggesting positive productivity and wage spillovers in Indonesia, Thailand, and to some extent Vietnam. However, such analyses need to be treated with caution because unwarranted pooling across manufacturing industries is common and has the well-known potential to bias estimates. In addition, the paper emphasized that balanced panels can create important sampling biases because of large turnover that is particularly conspicuous among small non-MNCs in this dynamic region.

Full text

PDF Download (PDF: 320KB)