| Summary: | <p> My dissertation consists of three essays on econometric forecasting and forecast evaluation. Each essay contributes to the literature in its own way. In the first essay, I employ a common qualitative analysis framework on a widely-available consensus forecast. I evaluate the consensus forecast's performance using the Predictive Failure statistic. I find that the survey respondents provide statistically significant directional forecasts or signals. The second essay evaluates forecasts of a prominent energy policy model. This includes analysis of the forecasts, developing a rival forecasting model, and attempting to identify information to improve the policy model's forecasts. This analysis finds that the policy model generally performs well against a sophisticated time-series model. Specifically, the model appears to incorporate data dynamics appropriately, and this technique can serve as a new benchmark to evaluate the forecasts. The third essay examines the role of data revisions on models employing error correction relationships. These error correction relationships are rooted in economic theory and thus should be robust to changing economic conditions and data revisions. This essay finds that an older, popular model breaks down using newer data vintages, suggesting that data revisions and data issues can influence the stability of vector error correction models.</p>
|
|---|
