Synthetic Census Microdata Generation: A Comparative Study of Synthesis Methods Examining the Trade-Off Between Disclosure Risk and Utility
Why this work is in the frame
A frame that forgets how it found something cannot be audited. These are the routes that admitted this work.
Bibliographic record
Abstract
There is growing interest in synthetic data generation as a means of allowing access to useful data whilst preserving confidentiality. In particular, synthetic microdata generation could allow increased access to census and administrative data. An accurate understanding of the comparative performance of current synthetic data generators, in terms of the resulting data utility and disclosure risk for synthetic microdata, is important in allowing data owners to make informed decisions about the choice of method and parameter settings to use. Synthesizing microdata can present challenges as the data typically contains predominantly categorical variables that standard statistical methods may struggle to process. In this paper we present the first in-depth evaluation of four state-of-the-art synthetic data generators originating from the statistical (synthpop, DataSynthesizer) and deep learning (CTGAN, TVAE) communities and each capable of dealing with microdata. We use four real census microdatasets (Canada, Fiji, Rwanda, UK) to systematically validate and compare the synthetic data generators and their parameter settings in terms of the utility and disclosure risk of the resulting synthetic data using statistical metrics and the risk-utility map for visualization. Our analysis shows that the performance of the synthetic data generators considered depends on their parameter settings and the dataset.
Fetched live from OpenAlex and de-inverted. Abstracts are not stored in this database: the inverted indexes are 8.6 GB of the frame’s 9.3 GB of text, and the host has 13 GB free.
Full frame distilled prediction
Teacher imitationNot calibrated prevalence, not ground truth. Human validation pending. Learned from the 10,348 direct Codex labels and 10,348 direct Gemma labels. Candidate is the union of thresholded teacher heads; consensus is their intersection. These outputs are machine_predicted_unvalidated and are not human labels or direct frontier model labels.
Codex and Gemma teacher scores by category
| Category | Codex | Gemma |
|---|---|---|
| Metaresearch | 0.003 | 0.015 |
| Meta-epidemiology (narrow) | 0.000 | 0.000 |
| Meta-epidemiology (broad) | 0.000 | 0.000 |
| Bibliometrics | 0.000 | 0.000 |
| Science and technology studies | 0.000 | 0.000 |
| Scholarly communication | 0.000 | 0.000 |
| Open science | 0.006 | 0.006 |
| Research integrity | 0.000 | 0.001 |
| Insufficient payload (model declined to judge) | 0.000 | 0.000 |
Machine scores (provisional)
The two teacher heads of the student model, read on this work. A score orders the frame for review; it never asserts a category, and the validation status ships verbatim with every row.
Baseline scores from an immature model (maturity gate not passed, 7 training rounds). Scores rank; they never assert a category.
score_only:v0-immature-baseline · verbatim from the scoring run: score_only means the number may rank works, and no category label ships from it