Statistical methods for estimating and characterizing social welfare recipient payments

摘要

Social welfare programs constitute public assistance that provides benefits to eligible populations. While the Federal Government regulates most social programs, they are generally administered at the State level. From a State perspective, the administration of these programs requires that two types of information be tracked: (1) Information about the magnitude of recipient expenditure. This is necessary to claim these expenditures to the Federal Government. (2) Information about the amount of time that recipients receive benefits from the social programs. This is useful in evaluating social program performance and potential program administration issues.;Adopting a state perspective, this research develops statistical methods to effectively estimate and characterize recipient payments in social welfare programs that provide periodic payment benefits. Specifically, this is accomplished by:;(a) Developing multi-period retrospective sampling approaches to estimate recipient payment expenditures. The proposed sampling methods improve the multi-period reliability of elementary sampling estimates by progressively incorporating sampling information obtained from previous sampling periods. Contrary to existing rotation panel approaches and to repeated survey time series methods, the proposed methods improve multi-period sampling reliability estimates without requiring special data assumptions, such as covariance-matrix non-singularity and data stationarity.;(b) Developing a nonparametric univariate technique to estimate recipient missing payment amounts in poor quality electronic databases. The proposed technique uses cross-sectional recipient information and it allows relative reductions in payment amount over-estimation penalties as compared to other univariate methods. The proposed technique differs from existing missing data techniques which require explicit-modeling, because it does not depend on either covariates or restrictive modeling assumptions.;(c) Developing a nonparametric multi-episode duration analysis approach. The proposed approach provides knowledge for understanding how recipient populations use a social program. The proposed analysis approach extends existing multi-episode analysis approaches by allowing the comparison of duration patterns for recipients with different number of service episodes.;In addition, this research provides a data set valuable for future comparative research in social sciences.