Preliminary Estimates with Design-Based Methods (Method)


Timeliness is a particularly critical component of quality for producing short-term business statistics at the National Statistical Institutes (NSIs) of the European Community, as each Member State has to meet the standard quality requirements of the Regulation No 1165/98 – amended by the Regulation No 1158/2005 – about terms for transmission of the results and details of the information provided on statistical indicators, particularly on short-term statistics. The Amendment EU Regulation on Short Term Statistics requests all the statistical institutes of the EU Member States to transmit preliminary short term indicators to EUROSTAT with a reduced delay comparing to the timeliness set in the original 1998 Regulation (Eurostat, 2000, 2001, 2005). In OECD context, also, research projects were settled and useful documentation produced (Di Fonzo, 2005).

Frequently, in the NSIs short term statistics are based on fixed panel surveys of enterprises or rotating panels with a partial overlap from one year to another. More precisely, the amended regulation provides for a substantial improvement of timeliness for the production of the most important short term indicators.

A common approach for dealing with preliminary estimates focuses essentially on the study and the definition of efficient estimators, exploiting almost exclusively auxiliary information in the estimation phase. Often preliminary estimation merely involves the use of the quick respondent units. In fact, in order to obtain "good" preliminary estimates, standard survey strategy often aims to achieve high quick response rate by means of a well-structured plan of follow-up. In some surveys the "largest" units are carefully supervised.

The main theoretical problem to be faced in a short-term preliminary estimation context concerns the possible self-selection of quick respondents that can lead to biased estimates of the unknown population mean and variances.

A useful documentation on preliminary estimation problems (even though not comprehensive) can be downloaded from the OECD web site.

This module focuses on estimation methods referring to the design-based approach. In particular describes a method proposed in Rao et al. (1989) which uses, to produce estimate referred to time t, data pertaining both to time t and t-1, with the aim to minimise the mean square error of the estimate.

Apart from this particular method, design-based (or model-assisted) estimation methods for preliminary estimates using quick respondents refer to the class of non-response weighting adjustment procedures, which are used in general when the Theoretical Sample (TS) is not achieved in practice in the Observed Sample (OS). In the case of preliminary estimates the observed sample coincides with the quick respondent set of units, available at the point of time when preliminary estimation has to be performed.

It is worth noting that in the context of preliminary estimate production the two most frequent situations at NSIs are: (i) using for the preliminary estimates the same estimator used for the final estimates computed on quick respondents, or (ii) referring to model-based estimators and ignoring the sampling design which generated data.


To read the entire document, please access the pdf file (link under "Related Documents" on the right-hand-side of this page).


Your feedback is appreciated. Please send your remarks, suggestions for improvement, etc. to memobust@cbs.nl.