The Balassa-Samuelson hypothesis in the developed and developing countries revisited

•A new cointegration test for panel data with multiple heterogeneous unknown breaks is proposed.•This test also captures the cross sectional dependence by incorporating non-stationary factors.•Applying new test to the Balassa-Samuelson (BS) hypothesis shows the BS hypothesis holds only for developed economies.

This paper aims at examining the Balassa-Samuelson (BS) hypothesis in 20 developed and 20 developing countries, respectively. Given the cross-sectional dependence and structural breaks, we develop a new panel cointegration technique which allows for multiple heterogeneous unknown breaks and non-stationary factors. The empirical results show that the BS hypothesis holds in the developed countries, implying that higher productivity growth leads to a real appreciation, but they are cointegrated up to a number of cross-sectional unobserved stochastic trends as factors are non-stationary. However, we find little evidence to support the BS hypothesis in the developing countries and a further research is needed for the reason.