arXiv subjects

Lars Vilhuber

Publications and source records attributed to Lars Vilhuber.

8 recordsLinked to original sources

arXiv2026

Randomness in large language models: What researchers need to know (and report)

Large language models (LLMs) are increasingly used to generate data for research. Typical use cases are classifications, annotations, information extraction, and generation of numerical scores. Unlike conventional measurements, LLM outputs can vary across repeated requests even when the prompt and apparent model settings remain unchanged. This variation arises from deliberate sampling, silent model updates, numerical rounding, or expert routing. Setting a dedicated temperature parameter to zero removes deliberate sampling when that option is available, but it does not eliminate the other sources of randomness. Exact reproduction is therefore generally not possible when using proprietary application programming interfaces. Local execution of open-weight models offers greater control, but reproducibility still depends on the complete hardware and software stack. We illustrate these issues through sentiment classifications of corporate filings and examine their consequences for downstream regression results. We then propose a reporting standard for articles and replication packages, as well as guidance for data editors and authors. Together, these findings and recommendations establish that LLM outputs should be treated as draws from a distribution rather than as fixed measurements.

econ.GN↗

arXiv2026

Assessing Utility of Differential Privacy for RCTs

Randomized controlled trials (RCTs) have become powerful tools for assessing the impact of interventions and policies in many contexts. They are considered the gold standard for causal inference in the biomedical fields and many social sciences. Researchers have published an increasing number of studies that rely on RCTs for at least part of their inference. These studies typically include the response data that has been collected, de-identified, and sometimes protected through traditional disclosure limitation methods. In this paper, we empirically assess the impact of privacy-preserving synthetic data generation methodologies on published RCT analyses by leveraging available replication packages (research compendia) in economics and policy analysis. We implement three privacy-preserving algorithms, that use as a base one of the basic differentially private (DP) algorithms, the perturbed histogram, to support the quality of statistical inference. We highlight challenges with the straight use of this algorithm and the stability-based histogram in our setting and described the adjustments needed. We provide simulation studies and demonstrate that we can replicate the analysis in a published economics article on privacy-protected data under various parameterizations. We find that relatively straightforward (at a high-level) privacy-preserving methods influenced by DP techniques allow for inference-valid protection of published data. The results have applicability to researchers wishing to share RCT data, especially in the context of low- and middle-income countries, with strong privacy protection.

stat.AP↗

arXiv2025

A Simulated Reconstruction and Reidentification Attack on the 2010 U.S. Census

We show that individual, confidential microdata records from the 2010 U.S. Census of Population and Housing can be accurately reconstructed from the published tabular summaries. Ninety-seven million person records (every resident in 70% of all census blocks) are exactly reconstructed with provable certainty using only public information. We further show that a hypothetical attacker using our methods can reidentify with 95% accuracy population unique individuals who are perfectly reconstructed and not in the modal race and ethnicity category in their census block (3.4 million persons)--a result that is only possible because their confidential records were used in the published tabulations. Finally, we show that the methods used for the 2020 Census, based on a differential privacy framework, provide better protection against this type of attack, with better published data accuracy, than feasible alternatives.

stat.AP↗

arXiv2023

Reproducibility and Transparency versus Privacy and Confidentiality: Reflections from a Data Editor

Transparency and reproducibility are often seen in opposition to privacy and confidentiality. Data that need to be kept confidential are seen as an impediment to reproducibility, and privacy would seem to inhibit transparency. I bring a more nuanced view to the discussion, and show, using examples from over 1,000 reproducibility assessments, that confidential data can very well be used in reproducible and transparent research. The key insight is that access to most confidential data, while tedious, is open to hundreds if not thousands of researchers. In cases where few researchers can consider accessing such data in the future, reproducibility services, such as those provided by some journals, can provide some evidence for effective reproducibility even when the same data may not be available for future research.

econ.GN↗

arXiv2022

Teaching for large-scale Reproducibility Verification

We describe a unique environment in which undergraduate students from various STEM and social science disciplines are trained in data provenance and reproducible methods, and then apply that knowledge to real, conditionally accepted manuscripts and associated replication packages. We describe in detail the recruitment, training, and regular activities. While the activity is not part of a regular curriculum, the skills and knowledge taught through explicit training of reproducible methods and principles, and reinforced through repeated application in a real-life workflow, contribute to the education of these undergraduate students, and prepare them for post-graduation jobs and further studies.

cs.CY↗

arXiv2020

Total Error and Variability Measures for the Quarterly Workforce Indicators and LEHD Origin-Destination Employment Statistics in OnTheMap

We report results from the first comprehensive total quality evaluation of five major indicators in the U.S. Census Bureau's Longitudinal Employer-Household Dynamics (LEHD) Program Quarterly Workforce Indicators (QWI): total flow-employment, beginning-of-quarter employment, full-quarter employment, average monthly earnings of full-quarter employees, and total quarterly payroll. Beginning-of-quarter employment is also the main tabulation variable in the LEHD Origin-Destination Employment Statistics (LODES) workplace reports as displayed in OnTheMap (OTM), including OnTheMap for Emergency Management. We account for errors due to coverage; record-level non-response; edit and imputation of item missing data; and statistical disclosure limitation. The analysis reveals that the five publication variables under study are estimated very accurately for tabulations involving at least 10 jobs. Tabulations involving three to nine jobs are a transition zone, where cells may be fit for use with caution. Tabulations involving one or two jobs, which are generally suppressed on fitness-for-use criteria in the QWI and synthesized in LODES, have substantial total variability but can still be used to estimate statistics for untabulated aggregates as long as the job count in the aggregate is more than 10.

econ.EM↗

arXiv2020

Applying Data Synthesis for Longitudinal Business Data across Three Countries

Data on businesses collected by statistical agencies are challenging to protect. Many businesses have unique characteristics, and distributions of employment, sales, and profits are highly skewed. Attackers wishing to conduct identification attacks often have access to much more information than for any individual. As a consequence, most disclosure avoidance mechanisms fail to strike an acceptable balance between usefulness and confidentiality protection. Detailed aggregate statistics by geography or detailed industry classes are rare, public-use microdata on businesses are virtually inexistant, and access to confidential microdata can be burdensome. Synthetic microdata have been proposed as a secure mechanism to publish microdata, as part of a broader discussion of how to provide broader access to such data sets to researchers. In this article, we document an experiment to create analytically valid synthetic data, using the exact same model and methods previously employed for the United States, for data from two different countries: Canada (LEAP) and Germany (BHP). We assess utility and protection, and provide an assessment of the feasibility of extending such an approach in a cost-effective way to other data.

econ.EM↗

arXiv2019

Suboptimal Provision of Privacy and Statistical Accuracy When They are Public Goods

With vast databases at their disposal, private tech companies can compete with public statistical agencies to provide population statistics. However, private companies face different incentives to provide high-quality statistics and to protect the privacy of the people whose data are used. When both privacy protection and statistical accuracy are public goods, private providers tend to produce at least one suboptimally, but it is not clear which. We model a firm that publishes statistics under a guarantee of differential privacy. We prove that provision by the private firm results in inefficiently low data quality in this framework.

econ.TH↗