A statistical definition of information utilization for policy making decisions and an evaluation impact test to determine its occurrence are proposed. A univariate time series analysis is used to identify the internal trend for a given policy output variable and to control its effect. Two problems are identified in implementing an evaluation impact test: sample size and the nature of the time series significance test, which is used to make inferences about a series shift but actually measures goodness-of-fit of the forecasts. Other studies, involving utilization rather than quality as the dependent variable, are cited. Two general classes of evaluation characteristics related to utilization are suggested: political or the nature of the organization which will use the evaluation results and the state-of-the-art in evaluation. It is suggested that utilization might be increased if certain elements were included in evaluations, such as: a statement of the premises on which a program is based; specification of the program's process model; and analysis of the effectiveness of program components or alternative approaches. (MH)