Problems in Forecasting

1. Limitations to the use of data
1. Time lag in publication
2. Subjected to revise
3. Methodology changed
4. Rebased
2. Data measurement errors and biases
1. Transcription error
2. Survivorship bias
3. Appraisal data (smoothed)
3. Limitations of historical estimates
1. Because of regime changed and thus non-stationary
2. Trade-off between short time span (less data, dependent on time chosen) and long time span (more data, regime changed)
3. Make sure data are synchronized (When comparing assets, both needs to have the same amount of data available)
4. Using ex post data to estimate ex ante data (forgot the risk faced in the ex post data)
5. Data mining and time period bias – find data until a pattern is matched
1. Should test the result on the out-of-sample data to confirm
2. Make sure have economic basis
6. Conditioning information
1. Value based on market conditions
7. Misinterpretation of correlations
8. Psychological traps
1. Anchoring trap
2. Status quo trap – prediction influenced by the current data
3. Confirming evidence trap – put too much weight on evidence confirming his/her predictions
4. Overconfidence trap
5. Prudence trap – to avoid making extreme prediction and avoid being regret
6. Recallability trap – Let pass disaster weight too much in their considerations
9. Model and input uncertainty