In this essay we aim at finding an appropriate flash estimator of the quarterly Swedish private consumption (PK). With the aid of the statistics program TRAMO we study if monthly data from the consumer survey (HIP) and retail industry (DH) can be used in a transfer function model (TFM) to forecast PK. In the work of assessing the state of the market and the business trend, fast information from the national accounts is needed for making decisions for the economic politics in Sweden. A way to speed up the information process is to use leading economic indicators to asses this development. Another way to get information faster is to use a flash estimate. Such an estimate is made by investigating whether a change in one variable can be approximated by another. The idea of flash estimates is that it should be available earlier than the variable thatâ??s estimated. This method is used in the UK, Italy and Portugal. If a flash estimate can be found, then a forecast for PK is possible between 40 and 100 days earlier than the ordinary quarterly report, depending on whether one, two or three months were used to be in the model. To make one-step ahead forecasts of PK we use the program TRAMO. The forecasts are evaluated by comparing the absolute mean percentage error (MAPE) and by applying the forecast accuracy tests of Granger-Newbold and Diebold-Mariano. In the analysis of the residuals of the transfer function models with one, two and three months of data we saw that the Jarque-Bera test for normality, Durbin-Watsons and Ljung-Boxâ??s test for autocorrelation all gave good values. MAPE for the forecasts all gave a lower value for the TFM than the univariate model. On the basis of these results all the TFM improved the forecasts of PK. PK +(Framåt2mån,DH2mån) gave the lowest MAPE (0,58%) of the models. The Granger-Newbold and Diebold-Mariano tests resulted in non-significantly better forecasts made by TFM with one month data. For the TFM with two and three months of data only PK+CCI2mån,DH2mån and PK+Framåt3mån,DH3mån respectively were found nonsignificantly better. All other forecasts made by TFM were significantly better than the forecast made by the univariate model at the 5%-level. The four TFM (PK+Samtida2mån,DH2mån, PK+Framåt2mån,DH2mån, PK+CCI3mån,DH3mån and PK+Samtida3mån,DH3mån) can be used as a flash estimate of the Swedish private consumption. The most accurate forecast was made with the PK+Framåt2mån,DH2mån, which could present a forecast of PK 70 days before the quarterly forecast of the Swedish private consumption.
2355 visningar
uppladdat: 2006-03-31
uppladdat: 2006-03-31
Inactive member
Nedanstående innehåll är skapat av Mimers Brunns besökare.
Kommentera arbete
...läs fortsättningen genom att logga in dig.
Medlemskap krävs
För att komma åt allt innehåll på Mimers Brunn måste du vara medlem och inloggad.Kontot skapar du endast via facebook.
Källor för arbetet
Saknas
Kommentera arbetet: In this essay we aim at finding an appropriate flash estimator of the quarterly Swedish private consumption (PK). With the aid of the statistics program TRAMO we study if monthly data from the consumer survey (HIP) and retail industry (DH) can be used in a transfer function model (TFM) to forecast PK. In the work of assessing the state of the market and the business trend, fast information from the national accounts is needed for making decisions for the economic politics in Sweden. A way to speed up the information process is to use leading economic indicators to asses this development. Another way to get information faster is to use a flash estimate. Such an estimate is made by investigating whether a change in one variable can be approximated by another. The idea of flash estimates is that it should be available earlier than the variable thatâ??s estimated. This method is used in the UK, Italy and Portugal. If a flash estimate can be found, then a forecast for PK is possible between 40 and 100 days earlier than the ordinary quarterly report, depending on whether one, two or three months were used to be in the model. To make one-step ahead forecasts of PK we use the program TRAMO. The forecasts are evaluated by comparing the absolute mean percentage error (MAPE) and by applying the forecast accuracy tests of Granger-Newbold and Diebold-Mariano. In the analysis of the residuals of the transfer function models with one, two and three months of data we saw that the Jarque-Bera test for normality, Durbin-Watsons and Ljung-Boxâ??s test for autocorrelation all gave good values. MAPE for the forecasts all gave a lower value for the TFM than the univariate model. On the basis of these results all the TFM improved the forecasts of PK. PK +(Framåt2mån,DH2mån) gave the lowest MAPE (0,58%) of the models. The Granger-Newbold and Diebold-Mariano tests resulted in non-significantly better forecasts made by TFM with one month data. For the TFM with two and three months of data only PK+CCI2mån,DH2mån and PK+Framåt3mån,DH3mån respectively were found nonsignificantly better. All other forecasts made by TFM were significantly better than the forecast made by the univariate model at the 5%-level. The four TFM (PK+Samtida2mån,DH2mån, PK+Framåt2mån,DH2mån, PK+CCI3mån,DH3mån and PK+Samtida3mån,DH3mån) can be used as a flash estimate of the Swedish private consumption. The most accurate forecast was made with the PK+Framåt2mån,DH2mån, which could present a forecast of PK 70 days before the quarterly forecast of the Swedish private consumption.
Kommentarer på arbetet
Inga kommentarer än :(
Liknande arbeten
Källhänvisning
Inactive member [2006-03-31] In this essay we aim at finding an appropriate flash estimator of the quarterly Swedish private consumption (PK). With the aid of the statistics program TRAMO we study if monthly data from the consumer survey (HIP) and retail industry (DH) can be used in a transfer function model (TFM) to forecast PK. In the work of assessing the state of the market and the business trend, fast information from the national accounts is needed for making decisions for the economic politics in Sweden. A way to speed up the information process is to use leading economic indicators to asses this development. Another way to get information faster is to use a flash estimate. Such an estimate is made by investigating whether a change in one variable can be approximated by another. The idea of flash estimates is that it should be available earlier than the variable thatâ??s estimated. This method is used in the UK, Italy and Portugal. If a flash estimate can be found, then a forecast for PK is possible between 40 and 100 days earlier than the ordinary quarterly report, depending on whether one, two or three months were used to be in the model. To make one-step ahead forecasts of PK we use the program TRAMO. The forecasts are evaluated by comparing the absolute mean percentage error (MAPE) and by applying the forecast accuracy tests of Granger-Newbold and Diebold-Mariano. In the analysis of the residuals of the transfer function models with one, two and three months of data we saw that the Jarque-Bera test for normality, Durbin-Watsons and Ljung-Boxâ??s test for autocorrelation all gave good values. MAPE for the forecasts all gave a lower value for the TFM than the univariate model. On the basis of these results all the TFM improved the forecasts of PK. PK +(Framåt2mån,DH2mån) gave the lowest MAPE (0,58%) of the models. The Granger-Newbold and Diebold-Mariano tests resulted in non-significantly better forecasts made by TFM with one month data. For the TFM with two and three months of data only PK+CCI2mån,DH2mån and PK+Framåt3mån,DH3mån respectively were found nonsignificantly better. All other forecasts made by TFM were significantly better than the forecast made by the univariate model at the 5%-level. The four TFM (PK+Samtida2mån,DH2mån, PK+Framåt2mån,DH2mån, PK+CCI3mån,DH3mån and PK+Samtida3mån,DH3mån) can be used as a flash estimate of the Swedish private consumption. The most accurate forecast was made with the PK+Framåt2mån,DH2mån, which could present a forecast of PK 70 days before the quarterly forecast of the Swedish private consumption.Mimers Brunn [Online]. https://mimersbrunn.se/article?id=13773 [2024-05-03]
Rapportera det här arbetet
Är det något du ogillar med arbetet?
Rapportera