In a classical Walrassian framework of smoothly functioning markets and comparative static analysis, the beneficial effects of trade liberalisation are well known. In the real world, production decisions develop along time. They are fringed with uncertainty, and subject to unfulfilled expectations. Depending upon demand elasticity, such phenomena lead to converging or diverging cobwebs. Yet, even if demand is inelastic, diverging cobwebs are rarely observed, because there exist also many return strings which call systems back in the vicinity of (unstable) equilibrium. Among the latter, as already noticed by Knut Wicksell in the 1930’s, attitudes toward risk and investment functions play a large role. In effect, introducing such mechanisms into a dynamic market equilibrium leads to “chaotic motion”, a now well documented mathematical being, with very specific characteristics. In this context, market price fluctuations no longer occur because of the “hand of God”, from completely external sources, such as climatic events. They are endogenous, generated by the market itself. While external risk is subject to the “law of large number”, thus allowing for the benefit of risk pooling through insurance mechanisms, endogenous is not. In particular, any effort to lower individual decision maker exposition to risk affects the values of the key model parameters, such as supply elasticity, thus changing the risk regime itself. Now, while many studies (especially by Hertel et al) have been undertaken in order to elicit the consequences of external risk for the magnitude and distribution of the trade liberalisation benefits, the endogenous risk case has generally been benignly ignored by the world research community. The present paper aims at filling this gap. To this end, a GTAP model along that line is developed, with and without agricultural liberalisation. It is shown that, after a while, the tendency to divergence is smoothed out by risk considerations. Yet, because of a greater price uncertainty, over 60 years, long run world growth is significantly affected by liberalisation. Increased price volatility plays the role of a negative technical progress, which offset the benefits from a more efficient use of comparative advantage. Distributional effects are discussed, both between regions and within. Results suggest that liberalisation is not likely to reduce poverty, quite the contrary, because rich are less risk averse than poor, and thus, can accumulate more reinsvestible benefits. However, richest nations do not benefit in the whole, because of the investment slowing down mechanism outlined above. Efforts have been done to test the model, taking opportunity of its dynamic character, which allows for the comparison between “predicted” and “actual” series. Although results in this respect can still be very much improved, price regimes from this model are compared with a few actual long run observed series, and found to be similar. The paper briefly discuss the difficulty of such comparisons, which, because of the “sensitivity to initial condition”, cannot be reduced to the simple point per point measurement of the discrepancy between “predicted” and “observed”. Series must be characterized by global indices, such as moments or Fourrier spectrums. Most of these results very much contradict common wisdom. This is why they are interesting. More research is thus needed to specify the deep sources of this outcome and their political significance.