Traded intermediate goods, non -homothetic consumer demands, and the factor content of trade

Jeffrey Jon Reimer, Purdue University

Abstract

The burgeoning field of empirical international trade has recently identified two “mysteries” associated with the poor performance of the well known Heckscher-Ohlin-Vanek model, which makes predictions about the factor content of trade. The first mystery is the “Case of the Missing Trade,” and concerns the model's tendency to over-predict trade flows. The second mystery is the “Endowments Paradox,” and concerns a little understood bias in the model's predictions that is related to per capita income. This dissertation examines the extent to which these mysteries can be explained by the model's assumption that consumer demands are homothetic. In doing so, the dissertation also addresses the question of how to measure the factor content of trade when intermediate goods are traded, and countries employ different technologies. With regard to the measurement issue, a generalized factor content of trade framework is developed which tracks the sourcing of all goods used as intermediates and in final demand. It is demonstrated that not using this framework (in particular, using the approach to intermediates of two recent studies) leads to biases in both the prediction and measurement of factor content. A numerical exercise demonstrates that employing this newly developed framework reduces the amount of “missing trade” by 13%, and reduces prediction errors concerning factor content trade by about 20%. In order to address the non-homotheticity issue, an implicit, directly additive demand system (AIDADS) is estimated using two alternative data sets. Both sets of results make clear that the HOV assumption of homotheticity is very much at odds with international data on consumption, since consumers' budget shares vary systematically across income levels. This is particularly true for services (budget shares rise with per capita income) and agriculture (budget shares fall with per capita income). Given the similarity of results from the two data sets, those more easily reconciled with production and trade data are used to incorporate the non-homothetic demand structure into the factor content of trade model. This alternative model of demand is found to reduce the amount of “missing trade” by 20%, and eliminate about 30% of the bias associated with the “Endowments Paradox.” It is concluded that for a model of trade to reasonably account for actual patterns of trade, it will need to incorporate non-homothetic consumer demands and trade in intermediate goods.

Degree

Ph.D.

Advisors

Hertel, Purdue University.

Subject Area

Economics|Agricultural economics

COinS