Impute non-dom status

[MaxGhenis]

We should impute non-dom status to enable more accurate modeling of potential reforms to the non-dom tax regime. This would allow us to better estimate the distributional and revenue impacts of proposals such as Labour's plan to abolish non-dom status.

Background:
Recent research by Advani et al. provides valuable data on the UK's non-dom population that we can use to inform our imputation:

1. "The UK's 'non-doms': Who are they, what do they do, and where do they live?" (CAGE Policy Brief no. 36)
2. "Reforming the non-dom regime: revenue estimates" (CAGE Policy Briefing no. 38)

These papers offer detailed information on the demographics, income, wealth, and geographic distribution of non-doms, as well as estimates of their unreported offshore income and gains.

Proposed approach:

1. Imputation stage:

   • Assign a probability of non-dom status to each individual in our enhanced FRS dataset. This could be uniform or tied to predictors from Advani's work (calibration occurs at reweighting):
     • Income and wealth levels
     • Nationality/country of birth (if available)
     • Geographic location
     • Occupation and industry
     • Age and gender

2. Reweighting stage:

   • Include relevant non-dom aggregates from Advani et al. in our reweighting targets.
   • Use gradient descent to adjust weights, minimizing deviations between our microdata and the non-dom aggregates, alongside our existing targets.

Key statistics to target in reweighting (from Advani et al.):

• Total non-doms: ~238,000 (2018)
• Total unreported income and gains: ~£10.9 billion (2018)
• Distribution of non-doms by income bands (e.g., 41% of those earning £5 million+ have claimed non-dom status)
• Geographic distribution (e.g., concentration in specific London areas)
• Industry distribution
• Distribution by length of stay in UK
• Nationality distribution

Additional considerations:

1. Impute unreported offshore income and gains for those assigned non-dom status, based on distributions in the research.
2. Consider adding parameters to adjust the non-dom probability and unreported income/gains imputations, allowing for sensitivity analysis.
3. We may need to synthesize some high-wealth individuals to accurately represent the top end of the distribution where many non-doms are concentrated.

Next steps:

1. Review the full Advani et al. papers and extract all relevant statistics for imputation and reweighting targets.
2. Implement the non-dom status and unreported income/gains imputation in the first stage of our enhanced FRS production.
3. Add the non-dom related aggregates to our reweighting targets.
4. Validate results against the statistics and revenue estimates provided in the Advani et al. papers.
5. Add parameters to model various reform proposals to the non-dom regime.

This enhancement will significantly improve PolicyEngine-UK's ability to model reforms to the non-dom regime and provide more accurate distributional analyses of UK tax policy changes.