Previous Canadian studies have documented a substantial rise in private pension assets such as employer sponsored pension plans and tax-sheltered retirement savings accounts for prime age workers, suggesting that private savings are becoming an increasingly important component of retirement financing. So far, however, very few studies have examined the post-retirement profiles of income generated by private savings. This study aims at making a contribution to the current retirement literature by providing insights into the overall structure of the income sources of older Canadians and, more specifically, the role of tax-sheltered savings programs such as RRSP in retirement financing. Using Canadian administrative data this paper examines changes in the income structure of older Canadians over the period from 1990 to 2005 and, in particular, changes in the roles of RRSP and private pensions during this period. We ask several questions, including: (1) how big is the role of RRSP income in retirement financing and for which group of seniors is the income share attributable to RRSP income the largest? (2) how did the patterns of RRSP withdrawals change in the past two decades? and (3) are RRSPs viewed mostly as a retirement asset or an “emergency fund” to be used in case of a health downturn or death of the spouse?
In our multivariate analysis section, the probability of RRSP withdrawals in later life is estimated using a conditional fixed-effects logit model (FEL) developed in Chamberlain (1980). This model has several important advantages which make it particularly suitable for our analysis. An important feature of FEL is that it permits controlling for unobserved household fixed effects without restricting their distribution, and the parameters estimates based on a conditional FEL are consistent and efficient. The parameter identification in FEL is based on the sub-population of those who experience at least one positive outcome, so the estimation results will not be affected by people who do not make any withdrawals whether they have RRSP assets or not.

Our preliminary findings show that the share of RRSPs in retirement income remains relatively low and that the likelihood of RRSP withdrawals is strongly correlated with lifecycle events, such as the death of the spouse, consistent with the view that families might view their assets in tax sheltered savings accounts primarily as precautionary savings.