This paper introduces two different non-parametric tests for panel unit root based on the wavelet decomposition of time series which may be used in the presence of cross-sectional dependency and an unknown structural break in the data. These tests are compared with the parametric IPS test proposed by Im, Pesaran and Shin (1997) and the Wald test suggested by Taylor and Sarno (1998). By means of Monte Carlo simulations, the results shown that the size and power properties of the new non-parametric tests are robust to cross sectional dependency of the error terms. Furthermore, it is shown that the tests may be used when the time series has an unknown structural break. These tests have also shown to have high power against the alternative hypothesis under the above mentioned conditioned, whiles the IPS and the Wald did not have any power to reject the alternative hypothesis in the presence of structural break in the data.