Comparison of US Median Family Income with Canada Through Simple Tables and Age-Period-Cohort Models with Interesting Stories Behind

• Activity Number: 167 - Data Mining and Econometrics
• Type: Contributed
• Date/Time: Tuesday, August 10, 2021 : 10:00 AM to 11:50 AM
• Sponsor: Business and Economic Statistics Section
• Author(s): Wenjiang Fu* and Li Gan and Jiming Jiang
• Companies: University of Houston and Texas A&M University and University of California, Davis
• Keywords: Age-period-cohort model; identifiability problem ; linearly dependent covariates; unbiased estimation

Abstract

Modeling family income and estimating the trend are important for the studies of economic growth. Family income can be obtained through the Survey of Consumer Finances (SCF) data, which often summarized into tables with 10 year age groups as rows and single survey year as columns with 3 years apart. In each cell of the table is the mean or median family income of a specific age group in a specific calendar year. The temporal trend can be estimated by with summarizing the table into single row to examine the trend over calendar year, or single column to examine the trend over age groups. Alternatively, one may fit a regression model with fixed effects of age, period, and also cohort. The full age-period-cohort (APC) model has the potential to examine the effects of age, period, and cohort (birth cohort) simultaneously, but suffers from the identifiability problem due to a linear dependence Period – Age = Cohort. Such identifiability problem has been studied. Until recently, a final resolution has been discovered through the intrinsic estimator method. This talk will compare age, period, cohort trends between US and Canada and reals interesting stories behind.