I investigate consumption patterns in Italy and use a PCA-biplot to discover a consumption puzzle for the elderly poor. Data from the third wave (2017) of the Eurosystem’s Household Finance and Consumption Survey (HFCS) indicate that Italian poor old-aged households boast lower levels of the marginal propensity to consume (MPC) than sug-gested by the dominant consumption models. A customized regression analysis exhibits group differences with richer peers to be only half as large as prescribed by a traditional linear regression model. This analy-sis has benefited from a visualization technique for high-dimensional matrices related to the unsupervised machine learning literature. I demonstrate that PCA-biplots are a useful tool to reveal hidden rela-tions and to help researchers to formulate simple research questions. The method is presented in detail and suggestions on incorporating it in the econometric modeling pipeline are given.
SAFE Working Paper No. 382
