Statistical methods for early estimation of short term indicators

Economic indicators provide important information for the evaluation of the state of the economy of a country and they are used for business cycle analysis. Timeliness of the indicators is of fundamental interest for analysts of government authorities, national banks, politicians or economic researchers, because of the short term nature of the indices. The methods discussed in this thesis were developed in order to increase timeliness of key-economic short-term indicators of Austria. There is a strong need to publish these indicators timely in order to enable policy makers to react adequately to changes at the current end of the data. Nevertheless, historically, the availability of short term indicators lagged considerably behind time-schedules. The calculation of the indices is based on a large enterprise survey for which the information of around 10 000 enterprises is collected. Enterprises included in this survey are obliged to report their data, with reporting deadlines of about 15 days after the end of a reference month. However, reporting delays are much longer than this time span and therefore data providers (in this case Statistics Austria) cannot deliver the information as early as intended. Several attempts were undertaken to accelerate data delivery by motivating respondents by incentives, but the respondents are often just not able to provide the necessary information earlier. Therefore, missing data has to be nowcasted. In contrast to conventional forecasts, rapid estimates presented subsequently, refer to periods which have already passed. Hence, partial information on the target population is available for the reference periods at the time of estimation. Employing this partial information in the best possible way is the intention of the methods discussed in this thesis.