Factors with retirement behaviour among retirees and pre-retirees identified with a machine learning method / Muhammad Aizat Zainal Alam

Muhammad Aizat , Zainal Alam (2023) Factors with retirement behaviour among retirees and pre-retirees identified with a machine learning method / Muhammad Aizat Zainal Alam. PhD thesis, Universiti Malaya.

	PDF (The Candidate's Agreement) Restricted to Repository staff only Download (240Kb)
	PDF (Thesis PhD) Download (1622Kb)

Abstract

The Malaysian population is greying, and most individuals were found not prepared for it. Many are implied to have low retirement savings as retirement was found considered not a top priority by Malaysians. With studies on the psychology of retirement driven by the rise of behavioural economics, the mental accounting effect where the prospect of gains and losses are weighted differently, and violation of fungibility between different categories of wealth and expenditure as per classical economic theory which impacts retirement behaviour can be anchored upon to explore further ways towards retirement preparedness among Malaysians given limited yet growing number of studies conducted to understand retirement among Malaysians. The Malaysian Ageing and Retirement Survey (MARS) Data Wave 1 is used for this study where the data was sampled via multiple-stage sampling framework where each region is stratified by urban and rural areas called enumeration blocks (EBs) which are proportionate to the population size of each region across 3,384 households. This study uses 3,067 responses which are then be coupled with a machine learning methodology (ranging from Naïve Bayesian, Generalised Linear Model, Logistic Regression, Artificial Neural Network, Decision Tree, Random Forest, and Gradient Boosted Trees) via RapidMiner Studio to expand the understanding of how categories of wealth and expenditures can affect retirement behaviour, given the increasingly important role of machine learning algorithms within the context of behavioural economics where it has been demonstrated to describe patterns and relationships in behavioural data better than standard statistical analysis. In this regard, it was found that a vast majority of individuals exhibit mental accounting behaviour (66% of total respondents weight the prospect of gains and losses differently), where it was also found that future income wealth category, such as retirement savings, have most predictive weightage on retirement satisfaction based on an artificial neural network model (ANN) with an accuracy rate of 80.33%. Moreover, a re-ranking of wealth priorities for pre-retirees who often think about retirement is called for to ensure that they achieve satisfaction later in retirement. Towards being more prepared for retirement, pre-retirees are encouraged to move from heavily saving in current assets wealth category such as in fixed deposits (which has the highest predictive weightage on the tendency of thinking about retirement based on an ANN model with an accuracy rate of 65.80%) towards more savings in future income wealth category such as in Private Retirement Schemes (PRS). It is demonstrated that not only mental accounting behaviour, but also the allocation according to mental accounting categories on top of demographic variables may have impact on retirement behaviour. Importantly, the evaluation of retirement wellbeing should equally focus on financial and psychological perspectives to ensure both are optimised accordingly, where organisations should focus on avenues to enhance an individual’s future income wealth category to enhance savings adequacy and retirement preparedness while the government can consider ways to improve income to encourage higher savings rate in tandem with the increasing provision of financial and psychological education on managing personal finances to the Malaysian population.

Actions (For repository staff only : Login required)