Regional parameters
Regional fields are optional extensions and are not part of the stable baseline mechanism.
Neutral reference fields
reference_internet_reach_rate, reference_target_audience_density, and reference_income_inequality_gini define the point at which regional context makes no relative change for that component. A reference is a modeling anchor, not automatically a worldwide average.
internet_reach_rate
The fraction of a population that uses the internet. It is an upper-level accessibility measure; it does not mean the same fraction uses an advertising platform or saw a campaign.
target_audience_density
The assumed fraction of internet users eligible for the campaign's targeting. Without platform audience measurements, this is a scenario assumption and should receive low confidence.
economic_willingness_multiplier
A positive proxy for broad willingness or ability to purchase relative to a reference. The behavior provider converts it into an additive log-odds adjustment. A national economic indicator should never be described as directly measured consumer intent.
income_inequality_gini
A value from zero to one summarizing how unequally income is distributed. In this simulator it changes conversion-noise spread, not mean demand. This is a simplified modeling choice, not an established law connecting inequality to advertising response.
Sensitivity weights
economic_willingness_log_odds_weight controls how strongly the willingness proxy shifts conversion log odds. income_inequality_noise_weight controls how strongly the Gini difference changes noise spread. A zero weight disables its corresponding adjustment.
Evidence record
Every private real-region value should record:
- source organization and indicator name;
- observation year and retrieval date;
- original unit and transformation;
- missing-data or interpolation method;
- why the indicator is relevant to the simulated decision;
- confidence and sensitivity range.
Official national statistics may be high-confidence measurements of their own indicator while their mapping to campaign behavior remains low-confidence. Record those two confidence judgments separately.