Nowcasting is not Just Contemporaneous Forecasting

Abstract

We consider the reasons for nowcasting, the timing of information and sources thereof, especially contemporaneous data, which introduce different aspects compared to forecasting. We allow for the impact of location shifts inducing nowcast failure and nowcasting during breaks, probably with measurement errors. We also apply a variant of the nowcasting strategy proposed in Castle and Hendry (2009) to nowcast Euro Area GDP growth. Models of disaggregate monthly indicators are built by automatic methods, forecasting all variables that are released with a publication lag each period, then testing for shifts in available measures including survey data, switching to robust forecasts of missing series when breaks are detected.

Robust Implementation of a Parsimonious Dynamic Factor Model to Nowcast GDP

“…We proceed widely in line with the applied literature (see, e.g., Castle et al, 2009) in judging the out-of-sample forecast potential of the derived common factor. In order to assess the one-period ahead forecast properties, we treat the period starting the second quarter of 1975 to the first quarter of 2003 as given and successively extend this window by an additional quarterly observation to make one-quarter ahead projections for each successive step of prolonging the in-sample-"window" up to the penultimate quarter.…”

Untitled

“…Nowcasting as Castle et al (2009) said implies to forecast the present or the recent past, that means that with quite recent data (i.e. weekly or monthly) we are able to make accurate predictions for a determined variable.…”

Short-term forecasting with mixed-frequency data: a MIDASSO approach

“…Koenig et al (2003) use the logarithmic approximation expressing quarterly growth rates of the dependent variable in terms of underlying monthly growth rates of coincident economic indicators available at the monthly frequency. 1 A slight modification of the approach undertaken in Koenig et al (2003) is suggested in Castle et al (2009), where each monthly variable is split into three quarterly time series, each corresponding either to the first, second, or third month of the quarters in the sample. The approaches of Koenig et al (2003) and Castle et al (2009) in dealing with variables observed at mixed frequencies is that a standard OLS regression can be used in order to estimate model parameters at the cost of a slight inflation of a number of explanatory variables.…”

“…Our MIDASSO approach is a combination of the unrestricted MIxed-frequency DAta-Sampling approach (U-MIDAS) (see Foroni et al, 2015;Castle et al, 2009;Bec and Mogliani, 2013), and the LASSO-type penalised regression used in Bai and Ng (2008), called the elastic net (Zou and Hastie, 2005). We illustrate our approach by forecasting the quarterly real GDP growth rate in Switzerland.…”