This paper analyses the empirical validity of Prebisch-Singer hypothesis using the time series Grilli-Yang Commodity Price Index data spanning from 1900 to 2015. The methodology employed is encapsulated in a threefold approach: a) endogenous detection of structural breaks; b) estimation of trend through piece-wise linear regression; and c) validation of the statistical significance of the trends applying the Mann-Kendall test. The four structural breaks endogenously determined, primarily, coincides with four important historical/economic events over the last century: (a) World War I (1914 to 1918) and, thereafter, the Great Depression (1929 to 1939), (b) World War II (1939-1945) and immediate post war rebuilding (1950s), (c) First Oil Crisis (1973-74) and (d) Commodity Price Boom (late 1990s). From the trend results, the inference derived on the validity of Prebisch-Singer hypothesis is mixed. If the overall period of study is considered, then the empirical evidence in support of PS hypothesis is weak. However, if the terms of trade movement over the last century is considered, especially the second half, then the evidence in favor of Prebisch-Singer hypothesis is strong. The other important finding is that it also depends on the type of the commodities i.e. the terms of trade of agricultural commodities are more prone to secular decline than metals.