Do judgmental adjustments of statistical forecasts of demand improve accuracy? Companies certainly believe adjustment helps: a survey by Fildes and Goodwin estimated that 34% of statistical forecasts were subsequently judgmentally adjusted. Now a large empirical study by Fildes, Goodwin, Lawrence, and Nikolopoulos (2009) has shown that on average adjustments do help, but only when something important has happened, is planned, or is expected that has not been included in the statistical model.
While management adjustments can incorporate useful information to improve forecast accuracy, they are expensive and they may introduce bias. Because the managers' adjustments in the study were made in an unstructured manner, it seems likely that further improvements in the use of managerial judgments are possible.
There has been little research on how best to forecast the effectiveness of altenative strategies for implementing government policies. It seems reasonable to assume that better forecasts of the effects of policy implementation would lead to better better policies and better implementation of those policies. Nicolas Savio and Konstantinos Nikolopoulos propose in their working paper an approach that combines two evidence-based forecasting methods: strunctured analogies (Green and Armstrong 2007) and econometric modelling.
Eighteen papers on forecasting for public sector problems have been accepted for the ISF in Hong Kong (June 21-24). The papers address forecasting issues in education, employment, hospitals, performance, population, terrorism, and transport. The hot topic is climate and related matters, about which there are eight papers. See the conference schedule for a complete list of papers (still subject to revision).
Forecasting for public policy is important because public policies can have large effects and involve coercion. The Public Policy Forecasting SIG encourages the auditing of public policy proposals using the Forecasting Audit Software available on this site (see the menu at the top of the page).