We present a Kaldor-Thirlwall theoretical and empirical framework on the basic driving forces of the behaviour of productivity and economic development in the long-run. From the Kaldorian view, we anchor on the hypothesis that, by operating under static and dynamic economies of scale, the main sources of the growth of productivity of the economy as a whole come from the manufacturing sector. Yet, according to the so-called Thirlwall's Law, for a country to prevent from facing balance of payments contraints to growth in the long run and successfully catch up with the levels of income per capita and well-being of developed countries, it must maintain an income elasticity of demand for exports above the income elasticity of demand for imports. Based on these main hypotheses and some stylised facts observed and pioneered by Kaldor, we show some empirical evidence based on both descriptive statistics and econometric regressions for Brazil between 1970 and 2010. There are at least four clear indicators to suggest that Brazil has entered into a process of early de-industrialisation: i) the sharp drop in participation of the Brazilian manufacturing industry in total value added in the last decades; ii) its declining average yearly growth rate of labour productivity since the end of the 1990s; iii), the dramatic augmentation of the technological gap in all subsectors of the manufacturing industry, classified by technological intensity, since the end of the 1990s and; iv) the significant trade deficits in the manufacturing subsector of more technological intensity between 2006 and 2008. In addition, since our econometric estimates also show that there was a dramatic increase in the income elasticity of demand for imports (from 1.97 to 3.36) and a decrease in the income elasticity for exports between 1999 and 2010, compared with the 1980-1998 period (from 1.36 to 1.33), we can conclude that Brazil has already embarked on a trajectory of falling-behind and has been facing balance of payments constraints to growth in the long run. However, since our econometric estimation of the Kaldor-Verdoorn coefficient revealed that the Brazilian manufacturing industry operates under dynamic economies of scale, this result means that the Brazilian manufacturing industry has, in principle, potential to sustain economic growth in the long run. All these results have important economic policy implications. The most important is that there is still time to revert both the early de-industrialisation and falling behind path in favour of a catching up trajectory in the Brazilian economy, since there is a fine coordination between the long-term policies (industrial and technological policies, infrastructure and education policies, among others) and the short-term macroeconomic policies (monetary, fiscal and, especially, exchange rate policies) oriented to boost the technological content of both Brazilian manufacturing goods and exports.