Bayesian Cohort Models Research Articles

Interpolation and forecasting of population census data

Since population censuses are not annually implemented, population estimates are needed for the intercensal period. This paper describes simultaneous implementations of the temporal interpolation and forecasting of the population census data, aggregated by age and period. Since age equals period minus cohort, age-period-cohort decomposition suffers from the identification problem. In order to overcome this problem, the Bayesian cohort (BC) model is applied. The efficacy of the BC model for temporal interpolation is examined in comparison with official Japanese population estimates. Empirical results suggest that the BC model is expected to work well in temporal interpolation. Regarding the age-period-cohort decomposition of the Japanese census data, it is shown that the cohort effect is the largest while the other two effects are very small but not negligible. With regard to the forecasting of the Japanese population, the official population forecast considerably outperforms the BC forecast in most forecast horizons. However, the pace of increase in root mean square error for longer-term forecasting is larger in the official population forecast than in the BC forecasts. As a result, a variant of the BC forecast is best for 10-year forecast.

A cohort analysis of household vehicle expenditure in the U.S. and Japan: A possibility of generational marketing

This paper shows the usefulness of cohort analysis for generational marketing. Aggregate data classified by age and period are decomposed into age, period, and generational cohort effects. We compare two cohort-analysis models, the constrained multiple regression model and the Bayesian cohort model. The empirical results that are common to the household vehicle expenditure ratio in the U.S. and Japan are as follows: (1) among a total of three effects, the period effect is the smallest; (2) with the exception of the latest birth cohort, the cohort effect shows a clear upward trend; (3) the age effect decreases in the 20s and 30s, and next increases with a peak detected in the late 50s, and finally decreases. We provide marketing implications for cohort segmentation and forecasting.

A Cohort Analysis of US and Japanese Homeownership Rates

Abstract Aggregate data of homeownership rates in the US and Japan, classified by period and by age, are decomposed into age, period, and cohort effects using the Bayesian cohort models which were developed to overcome the identification problem in cohort analysis. The main data-based findings are threefold. First, in both the countries, the age effects are the largest and the period effects are the smallest. Second, US and Japanese age effects are almost identical and consistent with some life cycle theories. Third, cohort effects are larger in the US than in Japan. In particular, the downward trend in the US cohort effects in the case of baby boomers is explained by two hypotheses. Furthermore, some policy implications are provided.

Empirical evidence on intergenerational inequality of tax burdens in the U.S. and Japan

Aggregate data of tax burdens in the U.S. and Japan, classified by period and by age, are decomposed into age, period, and cohort effects using the Bayesian cohort models which were developed to overcome the identification problem in cohort analysis. Main findings are that in both countries the age effects are the largest and the cohort effects are obscure or negligible and that in both countries significant intergenerational inequality is not observed.

A COHORT ANALYSIS OF US AGE–EARNINGS PROFILES

ABSTRACTAggregate data on US earnings, classified by period and by age, are decomposed into age, period and cohort effects, using the Bayesian cohort models, which were developed to overcome the identification problem in cohort analysis. The main findings, obtained by comparing college and high school graduates, are threefold. First, the age effects show a downward trend for the age group of 45–49 onwards for high school graduates but do not show any such trend for college graduates. Second, the period effects show a downward trend for high school graduates but reveal no such trend for college graduates. Third, the cohort effects are negligible for both college and high school graduates.

Age–Period–Cohort Decomposition of U.S. and Japanese Birth Rates

Aggregate data on birth rates in the U.S. and Japan, classified by period and by age, are decomposed into age, period, and cohort effects using Bayesian cohort models that were developed to overcome the identification problem in cohort analysis. Main findings are fivefold. First, age, period, and cohort effects movements are all larger in Japan than in the U.S. Second, in both countries, age effects are the largest and are roughly consistent with the life-cycle movements showing an inverted U shape. Third, Easterlin’s cohort size hypothesis roughly fits U.S. birth rates but not Japanese birth rates. Fourth, despite rapid decline of total fertility rates in Japan in last three decades, period effects have been on an upward trend since the early 1990s. Finally, upward and downward cohort effects movements in Japan are derived by rapid economic growth and the Equal Employment Opportunity Law, respectively.

Age-period-cohort decomposition of social security taxes and benefits in the USA and Japan

Aggregate data of social security taxes and benefits in the USA and Japan, classified by period and by age, are decomposed into age, period, and cohort effects using the Bayesian cohort models which were developed to overcome the identification problem in cohort analysis. Regarding to US results, it is shown that period and cohort effects have negligible impact on US social security taxes and benefits which are measured as the share of household expenditure and income, respectively. Regarding to Japanese results, it is shown that period effects show a moderate trend in social security taxes and that cohort effects indicate small up-and-down movements.

A cohort analysis of female labor participation rates in the U.S. and Japan

Aggregate data of female labor participation rates in U.S. and Japan, classified by period and by age, are decomposed into age, period, and cohort effects using innovative Bayesian cohort models that were developed to overcome the identification problem in cohort analysis. The main findings are that in both countries, age effects are the largest and period effects are the smallest; in both countries, age effects are roughly consistent with life-cycle movements expected by labor economics, but the negative effects of marriage and/or childbearing on women?’s labor supply in Japan are much larger than those observed in the U.S.; and in both countries, upward movements of cohort effects during 1930s–1960s were found. However, cohort effects are larger for the U.S. than for Japan. All the cohort results are roughly consistent with the marriage squeeze hypothesis and the Easterlin hypothesis.

An empirical analysis of US and Japanese health insurance using age–period–cohort decomposition

Aggregate data on household health insurance expenditure in the US and Japan that are classified by period and age are decomposed into age, period, and cohort effects by using the Bayesian cohort models. These models are developed to overcome the identification problem involved in cohort analysis. Despite the differences between the health insurance systems of the two countries, three interesting empirical findings are obtained. First, in both the countries, the age effects are the most influential, and the cohort effects have negligible influence. The latter provides a striking policy implication since the generational imbalance in social security expenditures is widely recognized in developed countries. Second, in both the countries, the period effects show a roughly upward trend. Finally, the age effects exhibit a roughly upward movement for all age groups in the US; however, in Japan, these effects show a downward movement for the 55-59 age group due to the changes in the health insurance system on retirement.

Bayesian cohort models for general cohort table analyses

New Bayesian cohort models designed to resolve the identification problem in cohort analysis are proposed in this paper. At first, the basic cohort model which represents the statistical structure of time-series social survey data in terms of age, period and cohort effects is explained. The logit cohort model for qualitative data from a binomial distribution and the normal-type cohort model for quantitative data from a normal distribution are considered as two special cases of the basic model. In order to overcome the identification problem in cohort analysis, a Bayesian approach is adopted, based on the assumption that the effect parameters change gradually. A Bayesian information criterion ABIC is introduced for the selection of the optimal model. This approach is so flexible that both the logit and the normal-type cohort models can be made applicable, not only to standard cohort tables but also to general cohort tables in which the range of age group is not equal to the interval between periods. The practical utility of the proposed models is demonstrated by analysing two data sets from the literature on cohort analysis.