This Ph.D. work aims at exposing, and giving sense to the dynamics of agrarian change in Southern Thailand where rubber is a major crop. It intends to provide elements for decision makers to reflect on the future directions of the Thai rubber economy. I applied a sequence of multivariate analysis and systematic clustering to characterize the trajectories of 220 rubber farms. I also analyzed the evolution of share-tapping arrangements though personal interviews, survey and literature review. A path dependence analytical framework was then applied to identify and characterize how institutional settings and institutional reproduction shaped the observed trajectories and share-tapping arrangements. I identify six significant farm transformation trajectories between 1990 and 2010. Two trajectories show a decline in landholding and/or hired labor related to medium farms (10.5%) and small farms (25%). Three trajectories show growing landholding size and/or use of hired labor. They concern growing medium family farm enterprise (14.5%), large family farm enterprise (4.1%) and farms moving towards patronal enterprise (7.7%). These opposing trajectories witness a continuing polarization of rubber farms. 38.2% of farms follow a trajectory of stability with no change in farm size and labor structure. However, these are small farms and present a risk to follow the trajectories of farm decline. Labor availability and share-tapping arrangements evolved over time but share-tapping remains the main, almost unique, labor contract for harvesting since the beginning of rubber cultivation in Thailand. Labor contracting arrangements are in an institutional lock-in situation under the current form of share-tapping through three self-reinforcing economic, functional and legitimating mechanisms. I identified a set of rules that could explain the prevalence of the share-tapping arrangement in pursuing the exploitation of available labor force and ensuring high resilience to variations and uncertainty in social and economic conditions. Path dependence explains farm polarization and the institutional lock-in of share-tapping as the results of self-reinforcement mechanisms leading to institutional reproduction. The thesis shows that the period of rubber control during 1934-1946 was a critical juncture resulting in large expansion of new plantings and an institutional setting favoring polarization that has reproduced itself through self-reinforcing mechanisms until now. Land polarization had a positive effect on the stability of share-tapping contracts. Due to polarization, farms with growing landholding employ more hired labor, inducing an increasing demand while an increasing supply of skilled tappers is provided by declining small farms. The current share-tapping arrangements ensure the availability of paid labor and simultaneously reinforce land concentration. Based on the results of this analysis of agrarian change, three alternative policy scenarios are: continuation of present policies, reverting polarization and coping with global constraints. The policy scenario of reverting polarization is suggested as an option considering its consequences on economic growth, social welfare and sustainable agricultural development in the rubber sector.