A model of income-verified discounts, its relation to the recent screening and segmentation literature, and empirical designs that could distinguish the main mechanisms.
Technical comments are especially useful here. Select a passage, or open the Hypothes.is panel at the right edge, to flag an omitted paper, an incorrect claim, or an identification problem.
Status. This is a research specification, not a completed result. It separates conclusions from the old two-type model, extensions that need to be derived, and empirical designs that first require data-access checks.
The original mechanism
The archival model combines a public income group with a private valuation. Let income group be y in {P,R} and valuation be theta in {L,H}, with high valuation more common among richer consumers. The firm cannot observe theta, so within each income group it offers a nonlinear price-quality menu.
Without income segmentation, the firm uses one menu. As in standard screening models, it distorts the low type's quality downward to reduce the information rent left to the high type. With income segmentation, it offers a menu to each income group. The changed proportion of high types changes the screening distortion within each group.
Result in the two-by-two model. Segmentation reduces the low-type distortion in the poorer group and increases it in the richer group. It raises information rent for high-valuation poor consumers and lowers it for high-valuation rich consumers. Firm profit rises; consumer, distributional, and total-welfare effects depend on parameters and welfare weights.
This is still a useful example. It is no longer enough as a novelty claim. Recent work directly studies screening and segmentation in richer environments. A new paper should begin from that work and state what the income credential and public-policy objective add.
Sketch of the screening logic
For a given segment, the low valuation's incentive constraint is slack while the high valuation's constraint binds. The firm leaves the high type enough rent not to imitate the low type, then lowers the low type's quality because doing so reduces that rent. A segment with a smaller high-to-low ratio puts less weight on rent extraction relative to low-type trade, so its low-type distortion is smaller. An income signal matters here because income predicts valuation without revealing it.
Proposed model extension
A useful extension would make four features explicit: income is an imperfect signal of valuation; the government has distributional weights; disclosure is voluntary and may be costly; and firms face a discount-only constraint tied to a public standard offer.
Environment
A consumer has income y and private valuation theta, jointly distributed according to F(y, theta).
The government or verifier chooses a credential rule pi(s | y), where s is a coarse eligibility signal.
The consumer decides whether to disclose s and bears a disclosure, hassle, or stigma cost kappa(s,y).
After observing the disclosed signal, the firm offers a menu of quality and payment pairs (q,t). Production cost is c(q).
Consumer utility is theta*q - t - kappa. The public objective can place larger welfare weights on lower incomes and subtract administration, privacy, and labor-supply costs.
Choose pi subject to Bayes plausibility, firm best response, IC, IR, voluntary disclosure, and policy constraints.
The formal work should compare a binary credential, several income bands, and no credential. It should also compare a discount-only rule with unrestricted personalized pricing. The baseline price cannot simply be treated as fixed: firms may change it when the policy is introduced.
Questions for the theory note
Optimal coarseness. When does another income band improve targeting enough to justify greater rent extraction, privacy exposure, and administration?
Voluntary disclosure. Which consumers reveal the credential, and how does selected take-up change the firm's posterior?
Discount-only pricing. Does a ban on markups protect consumers once the firm can adjust the public standard price, quality, or product range?
Distribution within income groups. Which low-income consumers gain or lose once valuation and outside options vary within the eligible group?
Endogenous income. Do thresholds alter work, reporting, or benefit-claiming decisions when income is produced under a nonlinear tax schedule?
Competition and repeated interaction. Does the credential intensify competition for eligible customers, or create a new route for data-linked price extraction?
Candidate propositions and comparative statics
A signal that lowers the high-to-low valuation ratio in the eligible group reduces the standard screening distortion for its low valuation type, holding other features fixed.
Finer segmentation need not monotonically improve consumer welfare. It improves fit but can increase extraction and disclosure costs.
Under voluntary disclosure, take-up costs can undo targeting and can make observed users unrepresentative of eligible consumers.
Under a discount-only rule, adoption requires the additional margin from eligible demand to cover the discount and implementation cost. The firm's response in the standard offer is part of the equilibrium.
Income thresholds can create labor-supply or reporting distortions; smoothing the credential or using lagged eligibility may reduce them at the cost of weaker targeting.
Relation to existing work
The closest papers already establish that segmentation, disclosure, and nonlinear pricing can divide surplus in many ways. The defensible contribution is therefore narrower: a public eligibility credential based on noisy income, evaluated with distributional weights and constrained use.
Dubé and Misra (2023), “Personalized Pricing and Consumer Welfare”field evidenceFinds higher profit, lower prices for a majority, and lower aggregate consumer surplus in one field setting. It warns against using the share receiving lower prices as a welfare measure.
DellaVigna and Gentzkow (2019), “Uniform Pricing in U.S. Retail Chains”retail evidenceDocuments nearly uniform chain pricing and estimates that flexible pricing could lower relative prices in poorer areas. It motivates testing why firms currently leave geographic or income-linked differentiation unused.
Cremer and Gahvari (2002) links nonlinear pricing to optimal tax design, relevant when income and labor supply are endogenous.
Empirical designs
Each design below names a target quantity and the assumptions needed to interpret it. The first step in the administrative-data studies is a feasibility audit; a convenient policy date is not itself an identification strategy.
1. UK benefit verification for broadband social tariffs
Estimand
The effect of lowering verification friction on eligible households' take-up, bills, switching, and retention.
Preferred design
Customer-level randomized encouragement or phased implementation with a provider. A provider-by-month event study is a weaker fallback.
Data
Eligibility checks, successful matches, offers, take-up, plan, bill, churn, and a defensible eligible denominator.
Assumptions
For staggered adoption, timing must not be driven by unobserved changes in provider demand or marketing; untreated units must supply a credible counterfactual.
Main threats
Endogenous provider adoption, concurrent campaigns, mismeasured eligibility, selective consent, and switching between providers.
The effect of automatic matching on enrollment, bills, consumption, arrears, and disconnection among eligible households.
Preferred design
Linked household records with variation in pre-reform non-enrollment or matching exposure across distributors, supplemented by event-time diagnostics.
Data
Administrative eligibility, tariff enrollment date, billing and consumption histories, distributor, location, and disconnection outcomes.
Assumptions
Exposure differences must not proxy for differential shocks. Any threshold design requires continuity of potential outcomes and no simultaneous programs at the cutoff.
Main threats
A national reform date, pandemic-era shocks, changes in other transfers or tariffs, record-linkage errors, and endogenous meter-account ownership.
3. Merchant field trial
Estimand
Incremental effects of a verified discount relative to an untargeted promotion and to an equivalent voucher or cash benefit.
Preferred design
Cluster randomization at the store or market level, with pre-specified standard prices and enough time to observe firm adjustments.
Data
Transactions, eligibility and take-up, product quality and availability, prices to ineligible customers, merchant margin, resale, complaints, and administration cost.
Assumptions
Clusters must limit customer and pricing spillovers; treatment arms must have comparable transfer values; measurement must include nonparticipants.
Main threats
Cross-store shopping, retailer repricing, short-run novelty, staff discretion, stigma, substitution across products, and insufficient power for distributional outcomes.
4. Survey experiment on take-up and legitimacy
Estimand
Effects of verifier, disclosure method, income threshold, and discount-only language on stated take-up, perceived fairness, and privacy concern.
Preferred design
Randomized vignettes with comprehension checks, open-text explanations, and oversampling near plausible eligibility thresholds.
Use
Choose implementation language and detect obvious objections before a field test.
Limit
It measures stated reactions, not purchases, firm responses, equilibrium welfare, or actual disclosure behavior.
5. Transaction-data structural exercise
Estimand
Income-specific demand and counterfactual prices, quantities, quality choices, and surplus under a discount-only credential.
Preferred design
Estimate demand using transaction or scanner data linked to coarse income measures, then impose alternative firm-response models.
Use
Screen markets and parameter regions where a pilot is plausible; clarify which elasticities drive the result.
Main threats
Price endogeneity, weak outside-option measurement, unmodeled supply responses, income measurement error, and extrapolation beyond observed pricing.
Minimum pre-analysis outcomes for a field test
Purchases and expenditure by eligibility and take-up status.
The posted standard price, discounts outside the program, quality, stockouts, and product removal.
Merchant revenue, variable margin, acquisition, and retention.
Application completion, false rejection, appeal, time cost, perceived stigma, and privacy incidents.
Effects on ineligible customers and nonparticipants, not only credential users.
Program administration and verification costs.
What would change the recommendation?
The case for further work becomes stronger if the technical note finds a genuine gap, an administrative-data audit identifies a credible counterfactual, and a field partner can enforce limited disclosure while recording standard-price and quality responses.
The case becomes weaker if the remaining theory is already covered, take-up costs dominate, sellers mainly adjust the standard offer, verification cannot be separated from other data uses, or a cash or voucher comparison performs as well at lower cost.
Practical threshold. Do not move from a research pilot to promotion as a general policy unless eligible consumers gain after firm responses and participation costs, ineligible consumers are not materially harmed, and the result is competitive with a simpler transfer.