Previsões - Eleições presidenciais em Portugal em 2011
The Portuguese presidential elections 2011 are a thing of the past and it is now time to compare the real votes with the last official polls and with our last political forecasts released on the 20/01/2011, which we remind considered four different methods:
1) Forecasts using GTI data and historical electoral results;
2) Forecasts using GTI data and historical electoral results with robust regression methods;
3) Forecasts using GTI data and past political opinion polls;
4) Forecasts using GTI data and past political opinion polls with robust regression methods;
In order to compare the different polls and our forecasts, we use the standard forecasting evaluation statistics:
- the Root Mean Squared Error (RMSE)
- the Mean Absolute Error (MAE)
- the Mean Absolute Percentage Error (MAPE)
The results are reported below:
| ELECTION RESULTS (23.01.2011) | |||||||
| Cavaco Silva | Manuel Alegre | ||||||
| Release Date | Institute | 52,94 | 19,75 | RMSE | MAE | MAPE (IN %) | |
| 21/01/2011 | Sol (média) | 58,00 | 22,00 | 3,92 | 3,66 | 10,48 | |
| 21/01/2011 | Universidade Católica | 59,00 | 22,00 | 4,57 | 4,16 | 11,42 | |
| 21/01/2011 | Eurosondagem | 56,30 | 25,00 | 4,41 | 4,31 | 16,46 | |
| 21/01/2011 | Aximage | 54,70 | 25,60 | 4,32 | 3,81 | 16,47 | |
| 20/01/2011 | Intercampus | 54,60 | 22,80 | 2,46 | 2,36 | 9,29 | |
| 19/01/2011 | Marktest | 61,50 | 15,00 | 6,92 | 6,66 | 20,11 | |
| 20/01/2011 | 1) GTI data and historical electoral results | 49,88 | 20,32 | 2,20 | 1,82 | 4,33 | |
| 20/01/2011 | 2) GTI data and historical electoral results with robust regression methods; | 50,11 | 20,60 | 2,09 | 1,84 | 4,82 | |
| 20/01/2011 | 3) GTI data and past political opinion polls | 59,48 | 22,77 | 5,09 | 4,78 | 13,82 | |
| 20/01/2011 | 4) GTI data and past political opinion polls with robust regression methods; | 60,40 | 24,36 | 6,20 | 6,04 | 18,72 | |
The previous statistical analysis allows us to reach the following conclusions:
- The use of GTI data increased considerably the precision of political forecasts: when GTI data were used with historical electoral results WITHOUT opinion polls data, we obtained the best political forecasts, with much lower RMSE, MAE and MAPE than official polls.
- Using robust regression methods resulted in minor gains when correct information was used (in our case GTI data and historical results), while they amplified the biases when incorrect information was used (in our case opinion polls, see the last row in the previous table)
These results are encouraging and stimulate us to further proceed in this line of research, but we are aware that we still have a lot of work to do.
Finally the latest trends in last 30 days reported by GTI and considering the main five candidates are reported below. The strong increase in the searches for the third candidate Fernando Nobre, during the last two weeks before the presidential elections, is clearly visible:
Gigi Bi
