GRADE Visualisations
What could curtailing tax expenditures provide for children, families and communities?
What are tax expenditures?
Tax expenditures are departures from a country's standard tax system that reduce the tax some people or firms pay: exemptions, deductions, credits, reduced or zero rates, deferrals and tax holidays. They are public spending delivered through the tax system. Instead of paying money out of the budget, the government forgoes revenue it would otherwise collect.
Why governments grant them
The policy objectives recorded in GTED:
- Attracting investment and developing priority sectors: corporate tax holidays, investment allowances, special economic zones, reliefs for mining, agriculture or manufacturing.
- Making goods and services more affordable: VAT exemptions or zero rates on basic food, medicines, education or energy.
- Supporting particular groups or regions: reliefs for pensioners, families, people with disabilities, or less developed regions.
- Encouraging employment, saving and other behaviour: pension and savings reliefs, employment credits.
- Environmental and social goals: reliefs for renewable energy, charitable giving or housing.
Many have no stated purpose: in the GTED data used here, many tax expenditures have no recorded policy objective. These appear as “Not stated/unclear” under Policy objective in the scenario builder (see the example “No stated objective”).
Why scrutiny matters
Tax expenditures receive less scrutiny than direct spending: they are often not appraised or reviewed annually, and their cost is less visible in the budget. The question this tool helps to answer is what the same revenue could do for children, families and communities if it were spent on public services instead. (Masiya et al. 2024, Tax expenditures and progress to the Sustainable Development Goals)
Methods
We use government-reported revenue forgone from the Global Tax Expenditures Database (GTED v1.4.0, September 2026), averaged over three recent years for each country (; government estimates where available, otherwise government projections — see Data treatment) and expressed in constant 2015 US dollars, the unit GRADE uses. Build a scenario below by choosing categories of tax expenditure; the revenue that curtailing them would return is then entered into the GRADE model as a constant share of government revenue. GRADE runs start in 2021 with a five-year lead-in; the additional revenue applies from 2026 and results are reported for 2026–2035. GRADE allocates it across the budget in line with recent patterns and estimates the consequences for children over that period.
How the scenario builder works
In step 2, pick a ready-made scenario or open Customise by category. Each category rule selects tax expenditures that match all the categories chosen within it (for example, beneficiary = businesses and objective = attract investment). Leave a field on “Any” to ignore it. Add further rules to widen the scenario: a tax expenditure is included if it matches any rule, and is never counted twice. The recovery slider sets the share of the selected revenue forgone that a government could realistically recover.
How to use: work through steps 1–4; the results alongside (below on small screens) update as you go. Below the results, explore by map, compare countries in a table, or see what the scenario is made of; click any country for its details.
1Choose countries
Pick a single country, or a group of countries by region and/or income group.
2Choose which tax expenditures to include
Pick a scenario, or customise it by category.
Customise by category (tax base, tax, type, beneficiary, policy objective, time frame)
Within a rule, a tax expenditure must match every category chosen. Add a rule to include further tax expenditures (rules are combined with OR).
4Choose outcomes to show
Tick the indicators of interest. GRADE assumes the additional revenue is spread across the budget as governments have spent in the past, so the figures are simultaneous effects of the same revenue, not alternative uses of it. Ticking or unticking an indicator changes only what is shown. Use the Population of interest menu at the top right to show basic services and social protection for all people, children under 18, children under 5 or women aged 15–49. In GRADE each outcome is modelled from government revenue per capita, with the six Worldwide Governance Indicators moderating the response: the same revenue achieves more where governance is stronger. Click the i beside each group for how its figures are calculated and the published model behind them.
Disclaimer: Tax expenditure data from the Global Tax Expenditures Database (IDOS and CEP). The modelling, results and interpretation are the authors' own and are not endorsed by IDOS or CEP.
Source: Global Tax Expenditures Database v1.4.0 (IDOS & CEP, 9 September 2026; DOI 10.5281/zenodo.22031048; CC BY 4.0). GRADE simulations, Universities of St Andrews and Leicester. Builds on Masiya et al. (2024), Tax expenditures and progress to the Sustainable Development Goals, Sustainable Development 32(6): 6144–6162.
Data treatment: GTED labels each estimate as an estimate or a projection. For each country we average three years since 2015, using the best basis available, shown as Basis in the table and country panel: (1) fully estimated years, the three most recent years containing no projected provisions (101 countries); (2) estimates and projections, the three most recent years containing estimates, where every such year also includes some projected provisions (Moldova, 2019–2021); (3) projections only, the three most recent years up to 2025, where the government reports only projections (11 countries: Algeria, Armenia, Dominican Republic, Estonia, Hungary, Israel, Italy, Luxembourg, Mexico, Romania, United States). Projections for past years are typically forward-looking budget figures that the government has not revised. US dollar values are revenue forgone as a share of GDP (from GTED) multiplied by GDP in constant 2015 US$ (World Bank WDI GDP per capita × population, as in the GRADE base data); this avoids distortions from exchange-rate swings. Taiwan is excluded because it is not in the GRADE base data. Cuba, Cayman Islands, Kosovo and Puerto Rico are shown with their tax expenditure but not modelled, because the GRADE base data hold no government revenue series for them. Kazakhstan is listed but not modelled: GTED flags its estimates as unusually high and excludes them from its own calculations. Rows GTED flags as dropped from all calculations are excluded, as are negative estimates. No provisions are removed as duplicates, consistent with GTED; in a few countries (e.g. Ecuador, Guatemala, Austria, Greece, Ireland, Rwanda) linked provisions reported in the same year may partly overlap. Categories follow GTED's Level 1 classifications, with minor regrouping and harmonised “not stated” labels.
Outcomes: GRADE (updated October 2026, with a five-year lead-in; base data 2026, version 55) was run for each of 113 countries at 12 shares of its total reported tax expenditure (2.5% to 100%). Results for any scenario are linearly interpolated between the two nearest runs. Direct GRADE runs for four categories (corporate income tax, VAT, investment incentives, temporary provisions) differ from the interpolated values by less than 3% for every outcome (0.1% for the two largest); interpolation slightly understates. Flow outcomes (deaths averted, years of schooling) are summed over 2026–2035; stock outcomes (access to services, staff, stunting) are for 2035. Expected deaths are GRADE's projected deaths for 2026–2035 without the additional revenue. For each country, the percentage is deaths averted divided by that country's expected deaths. The headline percentage pools countries: total deaths averted divided by total expected deaths in the countries that have at least one tax expenditure in the selected scenario (countries with none are left out of both).
Indicators and percentages: deaths averted and years of schooling are summed over 2026–2035; access to services, social protection, stunting, health workers and teachers are for 2035; governance is the change in the World Governance Indicators score by 2035 (averaged across countries). Percentages show the share of the unmet need closed: for survival, the share of GRADE's expected (projected baseline) deaths averted; for education, the share of school-age children out of school in 2035; for services, the share of people without access in 2035. For children and women, the share is the same as for the whole population, because GRADE apportions gains by population group. Where modelled survival reaches GRADE's maximum (99.9% under-5, 100% maternal), nearly all expected deaths are counted as averted; these countries are flagged (▲) and their survival figures should be read as an upper bound.
Map: country boundaries from Natural Earth via the world-atlas package (loaded from the web). Colour bands each contain an eighth of the countries shown. Grey countries have no usable tax expenditure data in GTED or are not modelled in GRADE.
Time frame: GRADE runs start in 2021; 2021–2025 is GRADE's five-year lead-in, the additional revenue applies from 2026 and outcomes are reported for 2026–2035. Base data beyond the latest observed year are GRADE's mechanical extensions (half the recent growth rate) and are not forecasts; the timing of reform is illustrative.
Limitations: This tool shows the opportunity cost of revenue forgone; it does not recommend removing any provision. Results are gross: they do not net off direct costs to households of removing a provision (for example, higher prices where a VAT exemption on basic goods is withdrawn). Revenue forgone is a static estimate and may overstate the revenue a policy change would raise. Countries define their benchmark tax systems differently, so figures are not strictly comparable across countries. Coverage is limited to central government in most cases, and 100+ countries publish nothing. Shares are of reported revenue forgone, so a country that reports few provisions will appear to have low tax expenditure. The % of tax revenue is from GTED's background data; the % of government revenue uses government revenue excluding grants from the GRADE base data 2026 (version 55; UNU-WIDER GRD, with IMF and World Bank data where GRD is missing), averaged over the same years.