FLASH GROWTH-ESTIMATES USING CALENDAR INFORMATION

In 1991 Statistics Netherlands introduced their early flash estimates of the Quarterly National Accounts. In this article we examine a new, faster flash estimate, some three to four weeks earlier. The gain is m ade by using a simple regression technique and incomplete data. To com pensate for the lack of data, information on the number of working day s and shopping days was added to the regression. The inclusion of thes e calendar aspects significantly affect GDP growth: 0.30 percent point extra GDP growth for one extra working day, and 0.17 percent point fo r one extra shopping day. The cost of an earlier estimate is a decreas e of reliability. The probability of a forecast error of over 0.5 perc ent point will be about 26 percent, compared to 12 percent for the off icial flash estimate.