When a Lower Remittance Fee Has a Higher Quoted Cost

A dated comparison of India-bound transfer quotes from the United States, United Kingdom and United Arab Emirates.

On 20 August 2025, a Western Union quote for sending USD 200 from the United States to India listed a USD 1.99 fee; a Ria quote listed USD 2. Both specified a debit-card payment through the internet, bank-account receipt in less than an hour, and high receiving-network coverage. Western Union was cheaper by one cent on the fee, but dearer by the World Bank's reported total cost: 2.48% of the stated amount against Ria's 1.27%. That 1.21-percentage-point gap is USD 2.42 on the Bank's USD 200 cost basis, not a measured payout difference. [1]

Currency conversion explains the reversal. Ria's recorded exchange-rate margin was 0.27%; Western Union's was 1.48%. Each uses its own recorded reference rate: INR 87.0725 and 87.0675 per dollar, respectively, despite the common collection date. The ranking compares dated quotes under the dataset's convention. It does not establish simultaneous availability or an exact difference in what a recipient would receive. [1]

That distinction matters whenever the fee is the easiest number to see. An explicit charge is paid or accounted for at one point in a transfer; a weaker exchange rate changes how much receiver currency the converted amount commands. World Bank Remittance Prices Worldwide (RPW) reports a percentage total combining the fee as a share of the transfer amount with an exchange-rate margin. This decomposition is established RPW practice. RPW's SmaRT measure also recognises that cost rankings should consider whether a service qualifies on transparency, speed and access. [2] The question for this article is more specific: in one historical set of India-bound service records, how often did the fee and reported total disagree, and what remained after restricting comparisons to similar recorded services?

A ranking that changes when the comparison changes

The underlying Q3 2025 workbook contains 46 US-to-India, 28 UK-to-India and 24 UAE-to-India service records selected for this analysis. Each has two surveyed transfer amounts: USD 200 and 500, GBP 120 and 300, or AED 735 and 1,835. One Ria US record has no numerical USD 500 fee, rate, margin or total, leaving 195 usable amount quotes from 196 slots. Collection dates run from 4 August to 3 September 2025. The familiar RPW “200” and “500” tier labels come from a historical survey convention; the actual UK and UAE amounts here are in pounds and dirhams, not contemporary dollar equivalents. [1]

For each corridor and amount, the comparison puts every two available service records into a pair. A reversal means the service with the lower explicit fee has a reported total cost more than 0.01 percentage point higher than its partner. Equal-fee pairs are counted separately: they have no fee ranking to reverse. The broad column below compares services within a corridor and amount even if their payment methods, delivery routes or dates differ. The strict column also requires the recorded payment instrument, access point, speed category, receiving method, collection date, receiving-network coverage category and Standard Note to agree. The latter is a match in fields the workbook supplies, not proof that every customer could use either service. [1, 6]

Scroll the table sideways to read all columns.

Exhibit 1. Pair counts from the Q3 2025 RPW workbook, reproduced separately in an independent evidence check. The denominator for each reversal is the number of unequal-fee pairs at that matching level; fee ties are shown against all strict pairs. The methods note locates the seven strict reversal pairs, and the source-row map identifies all 98 underlying service records in the original workbook. A pair can share a service with another pair, and the two amount rows can reuse a service. The counts are neither independent consumer choices nor a population share. [1, 6]

The broad US counts make the opening reversal look common, but many pairings involve different recorded service features or dates. Under the strict match, three reversals remain at each US amount. The UK GBP 120 cell retains one; GBP 300 and both UAE cells retain none. A zero among two UAE unequal-fee pairs carries little weight. Nor is three out of 28 a probability of a migrant choosing badly: the pairs overlap, and the workbook gives neither service take-up nor customer eligibility.

Matching also exposes a second question: what if two services charge the same fee? At USD 200 and USD 500, five strict US pairs have fee ties, and four in each amount have reported totals separated by more than 0.01 percentage point. Each UK amount has two strict fee-tie pairs; within both pairs, the reported totals differ beyond that threshold. Fee alone cannot order those pairs at all. The difference may still be irrelevant to a particular sender if one service is unavailable to them. It is relevant to interpreting what the fee field measures.

The strict pairs share a collection date, but each of the seven reversals still compares different recorded reference exchange rates. Requiring identical reference rates leaves five US unequal-fee pairs at each amount, with no reversal, and no eligible UK or UAE pairs. The seven RPW-convention reversals stand; the workbook establishes none within an identical-reference-rate subset. It also cannot establish that both quotes were available at the same instant. [6]

The examples that keep the conclusion narrow

The fee rule does not always fail. A Walmart2World US-to-India record on 26 August 2025 lists a zero fee and a reported total of −0.53% at both surveyed amounts. Its recorded rate is better than the reference used by RPW, giving a negative margin. That is a source observation, not a promise that a customer could repeatedly gain money from transfers or evidence of a special promotion. Across the full selected set there are 18 zero-fee amount quotes, six negative-margin quotes and two negative-total quotes. Leaving them in the analysis is as important as leaving in the reversals: removing awkward values would manufacture a cleaner story than the source supports. [1]

Amount can change the ordering too. In the UK records, a GBP 120 Wise quote has a GBP 1.31 fee and a 1.09% reported total; an ICICI Bank quote has a GBP 1 fee and a 1.24% total. Both are bank-account transfers initiated online for next-day bank-account receipt, recorded on 20 August. The lower fee loses on the reported total. At GBP 300, the two totals both round to 0.74%, even though the fees differ. This is a comparison at two discrete amounts, not evidence of an exact crossover point, a continuous fee schedule, or equal realised cost at GBP 300. Their recorded reference rates also differ substantially, at about INR 117.526 and 117.136 per pound. [1]

The services behind these cases illustrate why delivery belongs in the price question. Bank-account receipt can require an account. Cash pickup and a bank transfer may have different access costs, timing and practical reach even if a percentage-cost field ranks them cleanly. A sender who needs cash within an hour may not regard a three-to-five-day route as a substitute. RPW's own SmaRT indicator already screens qualifying services for transparency, speed and accessibility before averaging cheap offers. [2] The comparison here applies a different, narrower exercise: it asks whether the fee ranking survives equality of the recorded service fields, including date. It cannot measure the value of time, failed transfers, actual local pickup access or who qualifies for an offer. Those could change a sender's feasible set without changing the table's arithmetic.

All 98 selected records have transparent = Yes under the workbook's specific definition: the applied exchange rate was supplied to the researcher. This is useful for calculating the RPW margin. It does not certify a fully understandable consumer quote, establish how each fee enters the sender's payment, or rule out a recipient deduction. The comparison contains no advertisement evidence and supports no allegation that any provider hid a charge. [1, 2]

From a reported cost to money received

The RPW percentage is a consistent way to describe these quotes within the workbook. For all 195 complete amount quotes, 100 × explicit fee / stated amount + supplied FX margin reproduces the reported total to within 0.005 percentage point. That is a check on the reporting convention. It is not an identity for the money a recipient actually received. [1, 6]

Receipt depends first on what gets converted. Let P be the stated sender-currency amount, F the fee in that currency, r the provider's receiver-currency units per sender-currency unit, and C any deduction in receiver currency. If the fee is added, the sender pays P + F; P is converted and the recipient gets Pr − C. If the fee comes out of P, the sender pays P; only P − F is converted and the recipient gets (P − F)r − C. These are conditional quote structures, not findings about the providers above. With a fixed all-in budget B and an added fee, receipt would instead be (B − F)r − C. That calculation needs the fee and rate at B, any recipient deduction and the sender's eligibility. Two surveyed amount tiers do not reveal the missing schedule. [3, 6]

The opening USD 2.42 is therefore a normalised cost gap: 1.21% of USD 200 under RPW's convention. A household saving would require a common all-in budget, offers available to the same sender at the same time, known fee treatment and recipient charges. The workbook cannot establish those conditions. Converting each quoted amount to rupees would still use slightly different recorded reference markets.

In these dated records, fee rankings conflict with RPW total-cost rankings even among some services matched on recorded features. Matching excludes many broad pairs, however, and leaves thin UK and UAE comparisons; none of the seven strict reversals has an identical recorded reference rate. The practical question is what reaches the recipient for a given sender's full payment. Answering it requires the fee treatment, conversion amount and rate, any recipient deduction, eligibility, receipt method and timing. The one-cent fee advantage in the opening cannot answer that question.

Source note. Numbered citations correspond to REFERENCES.md: [1] the original RPW Q3 2025 workbook and its field definitions; [2] the Q3 2025 report, including SmaRT and cost decomposition; [3] RPW methodology; [6] the reproducible calculations and original-cell checks. Exact service IDs, workbook rows and comparison definitions are in METHODS.md. The analysis is descriptive and exploratory; it is not a current price comparison.

Methods and original-row locators

The original source paths and private evidence-review paths in this methods note are archival locators in the canonical workspace. The portable private source pack includes copies of the frozen manuscript, original RPW workbook, retrieved research PDFs and empirical code under source/, empirical/ and evidence/; the article itself remains an August–September 2025 quote analysis, not a 2026 price check.

Source and unit

Controlling original: World Bank Remittance Prices Worldwide (RPW) full workbook, sheet Dataset (from Q2 2016), period=2025_3Q, destination IND, origins USA, GBR, ARE; SHA-256 f1d7265b1856d61812c158d3d03fae686dbff58735deabe016a6bfe23482d2a8. The three corridors were selected before extraction and are not a random India-remittance sample. The unit is one service record, with a quote for each of two stated amounts where numerical fields are present. There are 46, 28 and 24 records; 195 usable numerical amount quotes of 196 slots. Ria ID 1212255609, Excel row 203609, lacks the USD 500 fee, provider rate, FX margin and total. Its blank fields are missing values. Each selected record says transparent=Yes, which the workbook legend defines narrowly as disclosure of the applied rate to the researcher. These are collection snapshots dated 4 August–3 September 2025, not current offers or customer transactions.

The retained original workbook is at execution/editor/reviews/A30-EVIDENCE-001/isolated-reproduction/raw/rpw_dataset_2011_2025_q3.xlsx, copied without content changes from the original A30 pilot. The original-cell verification is execution/editor/reviews/A30-EVIDENCE-001/original_cells.json; independent checks are .../INDEPENDENT_CHECKS.json. The evidence review's immutable manifest has SHA-256 82b7dbc0b773848bbd041fba5e3b6d4e6ff332d9a4e898052cc7403ebca3d76b. The corrected drafting contract has SHA-256 ec05862376a66df2cef510bb9d148437a81526001ee2b92393a527f7727e36dd. See REFERENCES.md for source URLs, exact pages and sections.

Cost and pair definition

For stated sender-currency amount P, explicit sender-currency fee F and supplied margin M in percent, the RPW reported total is 100F/P + M, subject to display rounding. All 195 usable records reconcile to within 0.005 percentage point. This equality defines the reported dataset cost; it does not fix whether any individual service adds or deducts the fee, what the full sender budget is, or whether a recipient-side charge exists. The article's USD 2.42 illustration equals USD 200 × (2.48 − 1.27)/100; it is an RPW-normalised sender-currency difference, not observed payout or a same-budget saving.

Within each corridor and amount, enumerate unordered service pairs with complete fee and reported total. A rank reversal requires unequal fees and a higher reported total for the lower-fee service by more than 0.01 percentage point. The threshold separates reported totals at the displayed precision. Pairs with equal fees form a separate tie category, whether or not their reported totals differ. No pair count is a count of independent people or transactions; a service appears in many pairs and in two amount cells.

The original comparison ladder has four levels: (0) same corridor and amount; (1) also same payment instrument, access point, speed category and receiving method; (2) also same collection date; (3) also same receiving-network coverage category and Standard Note. The article's “broad” is level 0 and “strict recorded match” is level 3. Standard Note null and empty string are treated as one unrecorded-note state. Of 98 service records, 94 have no substantive Standard Note; the other four include one send-limit note and three account-required notes. Equal unrecorded state is not evidence of equal eligibility. Note2 is blank in all 98. The core matched service/date/coverage fields are complete. The methods and thresholds were refined in exploratory analysis; this is not a preregistered test.

Exhibit 1 is reproduced in VISUAL_DATA.csv. broad_unequal_pairs and strict_unequal_pairs are reversal denominators. strict_all_pairs includes fee ties. The source calculation is execution/editor/reviews/A30-EVIDENCE-001/isolated-reproduction/comparison_counts.csv, independently verified against original cells in INDEPENDENT_CHECKS.json (all_24_count_rows_reproduced=true). The six rows are distinct corridor/amount cells; no pooled rate is warranted. The broad reversal counts are 99/999, 127/959, 48/295, 29/295, 10/217, 6/217. The strict counts are 3/28, 3/28, 1/5, 0/5, 0/2, 0/2, with total strict pairs 33, 33, 7, 7, 2, 2. Strict fee ties are 5, 5, 2, 2, 0, 0, of which 4, 4, 2, 2, 0, 0 have total-cost separation >0.01pp.

Original rows for reader examples and the seven strict reversals

All IDs below refer to the original workbook sheet and Excel row stated above, not a copied CSV line number. The pair file in the independent review is isolated-reproduction/strict_reversals.csv; original values are in original_cells.json and isolated-reproduction/selected_quotes_q32025.csv.

Scroll the table sideways to read all columns.

The opening US USD 200 comparison is Western Union 1212255647/row 203647 versus Ria 1212255565/row 203565: USD 1.99 versus USD 2 fee, 1.48% versus 0.27% FX margin, 2.48% versus 1.27% total; 20 August 2025; debit card, Internet, less than one hour, bank-account receipt, high coverage, no substantive Standard Note. Each record has its own reference FX rate: 87.0675 and 87.0725 INR/USD, respectively. The UK GBP 120/300 comparison is Wise 1212252123/row 200123 and ICICI Bank 1212252765/row 200765; same recorded date/method/speed/coverage, but reference rates 117.525529949999 and 117.13565 INR/GBP. The zero-fee/negative-total counterexample is Walmart2World 1212256131/row 204131 on 26 August 2025. All three examples are source observations, with no inference about promotion, actual acceptance or recipient cash.

Sensitivities, countercase and limits

All seven strict reversals remain using a >0.02pp gap, excluding three flagged Xoom USD 200 records, restricting to different firms, and matching pickup location and Note2. The extra pickup/Note2 restriction reduces UAE unequal-fee pairs from two to one per amount, still with no reversal. Recomputing FX margins from the displayed rate fields preserves the signs of the seven reversals. Three Xoom USD 200 records (IDs 1212255387, 1212255390, 1212255391) have supplied 2.26% margins but displayed-rate-implied 2.2515316%. The residual of about 0.0084684pp is not explained by ordinary rounding of the four-decimal displayed provider rate; its cause is unknown. Source values remain unchanged.

Requiring exactly equal recorded reference FX rates within the strict match leaves five US unequal-fee pairs per amount and zero reversals; UK and UAE have no eligible pairs. All seven strict reversal pairs use different reference rates. Thus a strict match on recorded service/date fields is not a same-instant, identical-market-benchmark price experiment. Other unobserved conditions include within-day timing, customer eligibility, fee incidence, card issuer fees, downstream recipient deductions, reliability and actual geographic reach. No causal effect of funding or receipt method is isolated. The selected services are unweighted; RPW's stated 80% market-coverage aim is not a verified weight for these three corridors. No confidence interval, consumer probability, national transfer burden, realised payout or provider-wide recommendation follows.

The conditional receipt identities in the article define P as the stated amount in sender currency, F as the explicit fee in that currency, r as receiver-currency per sender-currency exchange rate and C as recipient deduction in receiver currency: fee added, sender outlay P+F, receipt Pr−C; fee deducted from included P, sender outlay P, receipt (P−F)r−C. For a common budget B with an added fee, receipt would be (B−F)r−C, conditional on the applicable fee/rate at that budget. None is claimed as an observed provider payout. The two amount tiers do not identify the continuous fee/rate schedule. RPW reported total and actual receipt answer different questions.

Exhibit 1 source counts

Download VISUAL_DATA.csv

The six rows below reproduce the frozen VISUAL_DATA.csv. Broad matching means the same corridor and stated amount; strict matching additionally requires the recorded payment, access, speed, receipt, date, coverage and Standard Note fields to agree. Reversal denominators count unequal-fee pairs. All-pair denominators include fee ties. The units are overlapping service pairs, not independent senders.

Scroll the table sideways to read all columns.

Source: World Bank RPW Q3 2025 full workbook, sheet Dataset (from Q2 2016), period 2025_3Q, destination IND; original workbook SHA-256 f1d7265b1856d61812c158d3d03fae686dbff58735deabe016a6bfe23482d2a8. Counts were independently reconciled to original cells. Collection dates: 4 August–3 September 2025. A zero of two strict UAE unequal-fee pairs is a small recorded comparison, not a population estimate.

References and source versions

The numerical article claims use the original World Bank workbook and exact row locators in METHODS.md. Search summaries and literature abstracts are not numerical authorities here. PDF page numbers count the cover as page 1.

[1] World Bank, Remittance Prices Worldwide, full Q3 2025 workbook, official data catalog and XLSX resource. Sheet Dataset (from Q2 2016), period=2025_3Q, destination IND, origins USA/GBR/ARE, original row and service ID as in METHODS.md. Legend tab: transparent, date, inter lcu bank fx, cc(X) lcu fee, cc(X) fx margin, cc(X) total cost %. Retained original SHA-256 f1d7265b1856d61812c158d3d03fae686dbff58735deabe016a6bfe23482d2a8. This controls all case numbers, field definitions, dates and computed exhibit counts.

[2] World Bank Group, Remittance Prices Worldwide, Issue 54, September 2025, Q3 report, retained original SHA-256 a41fd59a81dbdf1a60873c60506ae473135e8d97d07d788952ff133d828a9662. PDF pp. 8–10: alternative averages and SmaRT's three cheapest qualifying services, with transparency, speed and access criteria; pp. 16–17, Figure 13: fee/FX-margin cost composition; p. 19, Box 2: total-cost components. PDF p. 1 has an anomalous stale Issue 49 header; Issue 54 identity appears from p. 2. Duplicated/stale p. 3 overview text is not used for any claim.

[3] World Bank RPW, Methodology, captured in execution/programmes/10/a30/session01/raw/methodology_browser.html on 30 September 2026 per BROWSER_RECEIPT.json. Sections “Transfer fee” and “Exchange Rate Margin” explain the components and possible untracked recipient charges; “Data Collection” identifies snapshot collection and possible recollection; “Firms Data” states the market-coverage selection aim. The captured page and workbook Legend control definitions. No realised market share for these services is inferred.

[4] Thorsten Beck and María Soledad Martínez Pería, What Explains the Cost of Remittances? An Examination across 119 Country Corridors, World Bank Policy Research Working Paper 5072, October 2009; retained original SHA-256 c836b8df8f6940ef60bf8c8e9a70d62efdfaf11eab492766b92f41d696288ea5. PDF pp. 5–8 describe corridor/provider cross-sectional data and India-bound corridor-average variation; pp. 13–16 give cost association results. Context for existing cost research, not evidence for A30's Q3 2025 pair counts or a causal interpretation. A later 2011 journal article has a different title; this citation identifies the inspected 2009 working paper.

[5] Kangni Kpodar and Patrick Amir Imam, How Do Transaction Costs Influence Remittances?, IMF Working Paper WP/22/218, November 2022; retained original SHA-256 6a9b470c1a13fd532bd5f1a72b99ed7dd339c94491786b5ae11c164833be660d. PDF pp. 9–10 describe RPW country-average costs and remittance-flow data; pp. 22–23 define the disclosure-based transparency proxy. Context only. No flow elasticity is transferred to India or to these service quotes.

[6] Indiconomics A30 calculations, methods and original-row locators, exhibit data and all-service row map. Independent original-cell calculation record: execution/editor/reviews/A30-EVIDENCE-001/INDEPENDENT_CHECKS.json; independently authored checks: independent_extract.py and independent_check.py in that directory. The original World Bank workbook in reference 1 controls the source observations. The independent review verifies the calculations; agreement with the producer is not evidence. No household outcome is observed.

The official data catalog and methodology links were rechecked in the independent evidence review on 30 September 2026 (execution/editor/reviews/A30-EVIDENCE-001/FRESHNESS_AND_SOURCE_CHECK.md). Economic quote dates remain August–September 2025. A source update timestamp does not turn them into current consumer prices.

Source-row map

Download SOURCE_ROWS.csv

The 98 selected service records below preserve the frozen SOURCE_ROWS.csv identifiers and original Excel row locators. transparent=Yes is the workbook's narrow applied-rate-disclosure field; it does not certify the consumer quote or recipient payout. The original workbook and empirical code are retained privately in the source pack.

Scroll the table sideways to read all columns.