Work in Progress

• "Unsupervised Clustering Techniques for Robust Regression"

  Abstract: This paper proposes a methodology to robustly estimate parameters in linear regression setting. While previous studies on outliers, on one hand, focus on reexamining the estimators and using reliable transformation techniques to variables in question, our approach, on the other hand, is to apply unsupervised clustering that has not received attention as a robust candidate for estimation. We consider two strands of clustering in this study, namely partitioning clustering (PC) and density-based clustering (DC). Our contribution to the literature is twofold. First, we develop a framework for robust regression by incorporating clusters into the linear model. Second, we compare the performances of various clustering techniques to evaluate which one of them is more reliable as the sample size grows.

• "Forecasting Equity Risk Premium with Unsupervised Clustering under Stationary Covariates"

  Abstract: This paper seeks to deal with the equity premium puzzle (EPP), which suggests that major economic and financial variables are unable to explain the dynamics of equity risk premium (ERP). A continuum of paper has allowed better prediction by means of machine learning techniques, which have shaped its own agenda and appealed to researchers in finance and artificial intelligence. We contribute to the literature in two ways. Firstly, we argue that, in presence of structural breaks, parameters are instable, and hence a remedy robust to this issue is needed. We consider expanding window (EXW) and rolling window (RLW) methods in this paper. Secondly, we deploy unsupervised clustering techniques in combination with the parameter-stable methods thereof. We apply the clustering to the ERP covariates under the assumption of weak stationarity. Two emerging advantages are that checking for no unit root in each covariate is sufficient and that there is great flexibility in choosing which estimator will be exercised. 

• "The Role of Collective Bargaining in Price Formation: Evidence from the Netherlands"

• "Overlapping Generations in a Multiagent Economy"

  Abstract: Overlapping generations (OLG) model has been the workhorse in macroeconomics to trace the dynamic changes in macro variables flavoured with lifetime behaviour of population, whereby one works and saves in period one, and retires and consummates leisure in period two. Nevertheless, the assumption of representative agent in macroeconomics is deemed restrictive, particularly when a population consists of two or more like-minded groups. As such this papers explores such heterogeneity by incorporating three groups, namely the altruists (AL), the common/representatives (RE), and the egoists (EG). Whereas the altruists discount the future at a relatively high rate, the egoists demand that the net present value of their wealth be maximised, more than what AL and RE would expect. First, the standard OLG model with only the representative agents is considered. Second, the standard model is contested with the outcome of OLG model that considers three different behaviours, i.e. AL, RE, EG. Third, to illustrate sensitivity, we consider changing the micro-founded parameters and the population composition of the three behaviours.