AStA Wirtschafts- und Sozialstatistisches Archiv最新文献

For many attempts to inform evidence-based policymaking (or policy-makers in general) researchers have to rely on already available (instead of newly collected) data. These data have to be reliable, accessible (at best, without high hurdles, and with low or no fees to be paid) and findable. One way that helps to find suitable data that are easily accessible (and hopefully reliable) is to look at the contributions published in the Data Observer series described in this paper.

Internet data pose a challenge to the traditional system of official statistics, which relies on more conventional sources such as surveys and registers, not readily adaptable to rapid changes. Expanding this system to include internet data is currently at an experimental stage, exploring these sources’ potentials and benefits. This paper describes a project conducted within the ESSnet Trusted Smart Statistics – Web Intelligence Network framework. It investigates the use of online apartment listings to analyze the rental market. We used web scraping to extract information from two online real estate portals for flats in the city of Berlin. Using this data, we developed a model to predict rental prices per square meter based on the accommodation’s features and location within the city. We detected offers which appear in both portals by means of statistical matching and removed duplicate offers. Missing values were treated by multiple imputation. The prediction model is a semi-parametric approach where the postal districts are used to describe the location effect. Comparisons with microcensus results and the local rent index reveal significant differences between the market of online flat offers and the stock of existing flat contracts. Interested readers will find the commented programming code in the internet supplement.

Following Russia’s invasion of Ukraine in early 2022, more than one million refugees have arrived in Germany. These Ukrainian refugees differ in many aspects from Germany’s past forced migration experiences and there exists an urgent need for sound data and information for politics, practitioners, and academics. In response, the IAB-BiB/FReDA-BAMF-SOEP study was established to provide high-quality longitudinal data following a register-based probability sample. We detail on an approach for sampling refugees in brief time, making use of two different registers—the German population register and the central register of foreigners—and discuss the quality of the final sample with respect to potential selectivity of participation in the panel. Overall, we demonstrate the benefits and feasibility of establishing register-based samples even in the context of a geopolitical crisis and the necessity of sound data within brief time horizons. We provide guidance that can be followed for similar events in the future.

In this paper we discuss the use and potential advantages and disadvantages of machine learning driven models in rental guides. Rental guides are a formal legal instrument in Germany for surveying rents of flats in cities and municipalities, which are today based on regression models or simple contingency tables. We discuss if and how modern and timely methods of machine learning outperform existing and established routines. We make use of data from the Munich rental guide and mainly focus on the predictive power of these models. We discuss the “black-box” character making some of these models difficult to interpret and hence challenging for applications in the rental guide context. Still, it is of interest to see how “black-box” models perform with respect to prediction error. Moreover, we study adversarial effects, i.e. we investigate robustness in the sense how corrupted data influence the performance of the prediction models. With the data at hand we show that models with promising predictive performance suffer from being more vulnerable to corruptions than classic linear models including Ridge or Lasso regularization.

Machine learning (ML) is increasingly being used in official statistics with a range of different applications. The main focus of ML models is to accurately predict attributes of new, unlabeled cases whereas the focus of classical statistical models is to describe the relations between independent and dependent variables. There is already a lot of experience in the sound use of classical statistical models in official statistics, but for ML models this is still under development. Recent discussions concerning the quality aspects of using ML in official statistics have concentrated on its implications for existing quality frameworks. We are in favor of the use of ML in official statistics, but the main question remains as to what factors need to be considered when using ML models in official statistics. As a means of raising awareness regarding these factors, we pose ten propositions regarding the (sensible) use of ML in official statistics.