Country-specific demand elasticities for forest products: Estimation method and consequences for long term projections

Abstract Long-term projections for the global forest sector are particularly sensitive to the parameters of the demand equations for the end products. To get more precise estimators, and test statistics with more power than with pure time-series, the elasticities of demand are typically estimated from panel data, pooling time-series across countries, and thus assuming that the elasticities are the same in all countries. The objective of this study was to recognize potential differences between countries, while using the prior information obtained by pooling. The proposed method estimated with quadratic programming country-specific elasticities that minimized the sum of squares of the errors across countries and over time, while keeping the country elasticities within the confidence intervals of the pooled elasticities. The method was applied to international data for seven product groups from 1992 to 2016. Compared with pooling, country-specific elasticities reduced the root mean square error of in-sample predictions by 7% to 43% depending on the product. The country-specific elasticities had smaller standard errors than the pooled elasticities, and they tended to cluster near the bounds of the confidence intervals of the pooled elasticities. With country-specific elasticities in a global sector model the projected world prices in 2065 were 3% to 13% higher, depending on the product, than with pooled elasticities. World consumption in 2065 with country-specific elasticities was from 2% lower to 42% higher depending on the product, with large differences across countries and product groups.