Testing a Poverty Trap Mechanism with Tsimane' Panel Data

Even if people have asserted the existence of a poverty trap within the forest, it has never been empirically tested that forest and poverty interactions lead to a poverty trap from which poor forest people could not escape (Angelsen and Wunder, 2003). This paper tests whether such a mechanism arises within forest communities and analyses the different livelihoods forest people could implement. Data used are from the Tsimane' Amazonian Panel Study (TAPS 2002-06) and have been complemented by qualitative data collected in June-July 2008. The poverty trap mechanism studied is inspired by Carter and Barrett's model defined for their studies of agrarian communities (Carter and Barrett, 2006). They intend to show that an S-shape exists in the asset accumulation process of households due to the presence of increasing returns to assets at an unstable equilibrium. An asset index is computed through factor analysis to estimate the asset accumulation process. To test the S-shape, different regression techniques are used: parametric, nonparametric and semiparametric. The parametric regression uses a fourth-degree polynomial regression to estimate the relationship between the change in asset index and its lagged value. For both nonparametric and semiparametric regressions, penalised splines and LOWESS estimators are used. All three regressions show that a poverty trap mechanism does not occur for the Tsimane', there is no S-shape in the asset accumulation and it seems concave. These results contradict Carter's and Barrett's results, because here Tsimane' households have a wide range of activities and assets. Households rely on diverse assets to achieve their livelihood. They use forest resources in addition to agriculture, so they are not pushed further into poverty if forest resources are depleted, they would rely more on agriculture as also land constraints are weak