Model II, which uses both current and prior Form ADV filings, is more likely to perform better at predicting fraud. This is because it incorporates a broader range of data, including both historical and current information. Research has shown that prediction models that rely on all prior Form ADV filings, including the most recent ones, perform better than models that only rely on a cross-section of ADV filings at a particular point in time. The panel model that uses data submitted in all prior years can predict 31.4% of frauds, compared to only 25.9% for the cross-sectional model. This suggests that the inclusion of historical data can significantly enhance the predictive power of the model, making Model II the most effective of the three.

Choice A is incorrect. Model I only uses the current Form ADV filings, which may not provide a comprehensive view of the firm's history and potential fraudulent activities. It lacks historical context which is crucial in identifying patterns or trends in fraudulent behavior.

Choice C is incorrect. While Model III incorporates past Form ADV filings, it excludes the current ones. This could potentially miss out on recent changes or anomalies that might indicate fraud.

Choice D is incorrect. The predictive accuracy of models can vary based on the data they incorporate and how they use it to identify potential frauds. In this case, Model II has a higher predictive accuracy as it uses both current and past Form ADV filings, providing a more holistic view of the firm’s activities.

Things to Remember

• Model selection in regression analysis is crucial for the accuracy of predictions.