Check missing scores before testing the gap between digital transformation and inclusion
Show which countries and periods are observable before interpreting a relationship between transformation and inclusion scores.
A blank score can change the conclusion
Plotting digital transformation against social inclusion is an appealing first step. Yet excluding countries without one score, or filling that score with 0, can change the picture before any relationship is examined. Truthound can help check missing and non-finite inputs, not decide whether transformation causes an inclusion outcome.
The new dashboard's question
UN DESA introduced its World Social Report and Socially Inclusive Transformation dashboard on October 15, 2026. The report questions whether economic, digital and energy transformations translate automatically into inclusion. Relationships between indicators must remain separate from evidence of policy effects.
Map the comparison population first
Keep country, period, pillar, component and observation state in the first table. A total score alone hides which information contributed to it. For a time comparison, identify the periods with compatible composition and coverage. Selecting that population is part of the analytical decision, not a neutral preprocessing detail.
Zero is not non-observation
A 0 is an observed value on a defined scale; an unobserved score is a data state. Combining them can make countries with less information appear to perform worse. Keep inapplicable components separate too. Even when exclusion is appropriate, show which countries and periods were removed.
The Truthound checks
Truthound Core Python SDK 3.3.4, operating on a prepared Polars table, provides NotNullValidator for required fields and FiniteValidator for floating-point NaN and infinity. The operator first defines which indicator rows are expected to be observed. Neither check establishes that a score is true or that a composite indicator's weights are suitable.
Keep the useful pandas mask
An existing pandas isna mask can remain useful. Align its treatment of empty strings and infinity with the indicator definition. Creating a mask and approving the comparison population are different steps. A Truthound workflow must not turn missing-country data into an automatic classification of national failure either.
Test a smaller hypothesis
A small experiment can examine only a fixed country-period population while displaying the difference from the full possible population. That allows a reader to ask how exclusions affect the result. Comparing component profiles of similar totals can also reveal how coarse a single score is.
Revisit the contract when scope changes
Revisit missingness policy when adding a country or period. A changed component set or wider disclosure may produce an apparent improvement through observation rather than actual performance. Preserve the earlier population and version so a new analysis cannot silently rewrite the old chart.
Show the result with its population
The artifact should pair the relationship with the population it represents. If the interpretation changes under another missing-data policy, that is a substantive finding, not a reason to assert policy impact. Understanding the conversion gap starts by explaining who entered the comparison.
Conditions
Verify the official indicator population, periods, components and observability states.
Alternatives
Keep pandas missing-value masks while separately defining indicator applicability and finite-value requirements.
Limitations
Missingness and finite-value checks do not establish that transformation causes inclusion.
Evidence
- Event checked
- Subject checked
- Modified
