Social Protection, and in particular social assistance, has emerged as a key area of international development policy. Recognition of the important role of social protection policies is aligned to a growing consensus on their role in reducing poverty and vulnerability and in preventing people from falling into poverty and facilitating exit from poverty. In the late 1990s, the World Bank (WB) developed the Social Risk Management (SRM) framework as a new conceptual framework for analysing social protection in developing countries. In the SRM, risk taking, the proactive management of risk at household level, was argued to be essential in enabling poor households to invest and grasp opportunities for economic development and poverty reduction. This research examines and conceptualizes the why and the how of investment by households in poverty. The research also examines the extent to which access to social assistance interventions may play a positive or negative role in the process. The thesis argues that the SRM oversimplifies and underestimates a variety of factors and processes that play a role in the household's investment behaviour in three dimensions of their life, namely savings, education and health. In the research, household decision making is conceptualised as a two stage process of 'constructing' investment preferences (what they are willing to do), and of 'realizing' or 'revealing' household choices (what they actually do). The empirical findings indicate that access to social assistance did not appear to have a role in constructing investment preferences. However, it had limited role under certain circumstances in favouring the realization of households' investment preferences. The research suggests that the SRM fails to encapsulate the complexity of household investment decisions, crucial to exiting poverty. Building on some of the concepts emerged in the research, I develop a 'behavioural' variation of the 'risk taking /poverty exit' component of the SRM in an attempt to improve the explanatory capacity of this framework.The research utilises the grounded theory framework (GT), adapted to a low income country context, and investigates the role of social assistance in household behaviour through an extensive field work in Bangladesh with urban households targeted by the Old Age Allowance Scheme (OAA), one of the largest social assistance schemes in Bangladesh.