Survey data can be viewed as incomplete or partially missing from a variety of perspectives and there are different ways of dealing with this kind of data in the prediction and the estimation of economic quantities. In this thesis, we present two selected research contexts in which the prediction or estimation of economic quantities is examined under incomplete survey data. These contexts are first the investigation of composite estimators in the German Microcensus (Chapters 3 and 4) and second extensions of multivariate Fay-Herriot (MFH) models (Chapters 5 and 6), which are applied to small area problems. Composite estimators are estimation methods that take into account the sample overlap in rotating panel surveys such as the German Microcensus in order to stabilise the estimation of the statistics of interest (e.g. employment statistics). Due to the partial sample overlaps, information from previous samples is only available for some of the respondents, so the data are partially missing. MFH models are model-based estimation methods that work with aggregated survey data in order to obtain more precise estimation results for small area problems compared to classical estimation methods. In these models, several variables of interest are modelled simultaneously. The survey estimates of these variables, which are used as input in the MFH models, are often partially missing. If the domains of interest are not explicitly accounted for in a sampling design, the sizes of the samples allocated to them can, by chance, be small. As a result, it can happen that either no estimates can be calculated at all or that the estimated values are not published by statistical offices because their variances are too large.

Behavioural traces from interactions with digital technologies are diverse and abundant. Yet, their capacity for theory-driven research is still to be constituted. In the present cumulative dissertation project, I deliberate the caveats and potentials of digital behavioural trace data in behavioural and social science research. One use case is online radicalisation research. The three studies included, set out to discern the state-of-the-art of methods and constructs employed in radicalization research, at the intersection of traditional methods and digital behavioural trace data. Firstly, I display, based on a systematic literature review of empirical work, the prevalence of digital behavioural trace data across different research strands and discern determinants and outcomes of radicalisation constructs. Secondly, I extract, based on this literature review, hypotheses and constructs and integrate them to a framework from network theory. This graph of hypotheses, in turn, makes the relative importance of theoretical considerations explicit. One implication of visualising the assumptions in the field is to systematise bottlenecks for the analysis of digital behavioural trace data and to provide the grounds for the genesis of new hypotheses. Thirdly, I provide a proof-of-concept for incorporating a theoretical framework from conspiracy theory research (as a specific form of radicalisation) and digital behavioural traces. I argue for marrying theoretical assumptions derived from temporal signals of posting behaviour and semantic meaning from textual content that rests on a framework from evolutionary psychology. In the light of these findings, I conclude by discussing important potential biases at different stages in the research cycle and practical implications.

Non-probability sampling is a topic of growing relevance, especially due to its occurrence in the context of new emerging data sources like web surveys and Big Data. This thesis addresses statistical challenges arising from non-probability samples, where unknown or uncontrolled sampling mechanisms raise concerns in terms of data quality and representativity. Various methods to quantify and reduce the potential selectivity and biases of non-probability samples in estimation and inference are discussed. The thesis introduces new forms of prediction and weighting methods, namely a) semi-parametric artificial neural networks (ANNs) that integrate B-spline layers with optimal knot positioning in the general structure and fitting procedure of artificial neural networks, and b) calibrated semi-parametric ANNs that determine weights for non-probability samples by integrating an ANN as response model with calibration constraints for totals, covariances and correlations. Custom-made computational implementations are developed for fitting (calibrated) semi-parametric ANNs by means of stochastic gradient descent, BFGS and sequential quadratic programming algorithms. The performance of all the discussed methods is evaluated and compared for a bandwidth of non-probability sampling scenarios in a Monte Carlo simulation study as well as an application to a real non-probability sample, the WageIndicator web survey. Potentials and limitations of the different methods for dealing with the challenges of non-probability sampling under various circumstances are highlighted. It is shown that the best strategy for using non-probability samples heavily depends on the particular selection mechanism, research interest and available auxiliary information. Nevertheless, the findings show that existing as well as newly proposed methods can be used to ease or even fully counterbalance the issues of non-probability samples and highlight the conditions under which this is possible.

Allocating scarce resources efficiently is a major task in mechanism design. One of the most fundamental problems in mechanism design theory is the problem of selling a single indivisible item to bidders with private valuations for the item. In this setting, the classic Vickrey auction of~\citet{vickrey1961} describes a simple mechanism to implement a social welfare maximizing allocation. The Vickrey auction for a single item asks every buyer to report its valuation and allocates the item to the highest bidder for a price of the second highest bid. This auction features some desirable properties, e.g., buyers cannot benefit from misreporting their true value for the item (incentive compatibility) and the auction can be executed in polynomial time. However, when there is more than one item for sale and buyers' valuations for sets of items are not additive or the set of feasible allocations is constrained, then constructing mechanisms that implement efficient allocations and have polynomial runtime might be very challenging. Consider a single seller selling $n\in \N$ heterogeneous indivisible items to several bidders. The Vickrey-Clarke-Groves auction generalizes the idea of the Vickrey auction to this multi-item setting. Naturally, every bidder has an intrinsic value for every subset of items. As in in the Vickrey auction, bidders report their valuations (Now, for every subset of items!). Then, the auctioneer computes a social welfare maximizing allocation according to the submitted bids and charges buyers the social cost of their winning that is incurred by the rest of the buyers. (This is the analogue to charging the second highest bid to the winning bidder in the single item Vickrey auction.) It turns out that the Vickrey-Clarke-Groves auction is also incentive compatible but it poses some problems: In fact, say for $n=40$, bidders would have to submit $2^{40}-1$ values (one value for each nonempty subset of the ground set) in total. Thus, asking every bidder for its valuation might be impossible due to time complexity issues. Therefore, even though the Vickrey-Clarke-Groves auction implements a social welfare maximizing allocation in this multi-item setting it might be impractical and there is need for alternative approaches to implement social welfare maximizing allocations. This dissertation represents the results of three independent research papers all of them tackling the problem of implementing efficient allocations in different combinatorial settings.