Thai financial proxy bank for SROI
Method note · Edition 3
How the Thai financial proxy bank is compiled What gets in, how each value is checked and graded, how far the spot check goes, and how errors are corrected
Summary
The bank collects Thai values that SROI analysts can use as financial proxies: money values that stand for outcomes with no market price of their own (Nicholls et al., 2012, p. 46). Edition 3 holds 328 values in 11 categories, from 137 source documents by 102 publishers.
Every value links to the page that holds the figure and quotes the text with the number, so anyone can find it again. Each names the year it describes, where it applies and how to carry it to another year, and carries a grade from A to D for its source.
This note sets out how values are found, checked, graded and corrected, so that users can judge each value for themselves and repeat the checks.
How to use a value in an analysis is covered in the how-to guide
Scope
The bank holds 273 money values for outcomes in Thailand, and 55 values that are not money values but are used to adjust others: 23 price indices, 18 conversion factors such as exchange rates and emission factors, and 14 percentages such as discount rates and unemployment rates.
The values describe years from 2004 to 2027. Most apply to the whole country (244); 40 apply to a region or a band of provinces, 20 to one province, 3 to a local area and 17 to one facility or provider, and 4 are international.
The bank does not say which outcomes a project should value, how much of an outcome it caused, or whether a value suits a particular group of stakeholders. Those are judgements for each analysis, made with the stakeholders (Social Value International, n.d.).
What gets in
A value enters the bank only when all of the following hold. An automated test checks every value against this list each time the site's code changes.
- The figure is printed in a document or on a page anyone can open, and the record links to that page, not to a home page.
- The text that holds the figure is copied from the source, with the page number where the document has one, and the date the source was opened is recorded.
- The value names the year it describes, its unit (in words and in a form software can read), where it applies, the groups it serves and how to carry it to another year.
- A note in Thai and English says what the value covers, what it leaves out and how to use it.
- A value computed from figures in the source shows its formula and inputs.
What stays out
- A figure whose source page could not be opened again: 11 in this edition, listed with the reason on the bank page.
- A figure not yet confirmed from a page that could be opened, including the original of a figure read from a copy: 129 items, the work list for the next edition.
Search and selection
The search starts with notices, laws and statistics from Thai state bodies, then research and published reports. Market prices are used where no official source exists.
For this edition 468 items were searched for. A figure was found with its source page for 339 of them; 11 were then left out because the page could not be opened again, leaving 328 values (292 read directly from the source and 36 derived from figures in it). The other 129 remain on the search list.
The selection is reported in the manner of the PRISMA 2020 flow diagram for systematic reviews (Page et al., 2021). The bank is not a systematic review: it does not search every database, and it aims to find usable Thai values rather than to collect every estimate of each outcome.
| Publisher | Values |
|---|---|
| Office of Agricultural Economics (OAE) | 34 |
| Bureau of Basic Education Policy and Plan, Office of the Basic Education Commission (OBEC), Ministry of Education | 19 |
| Thailand Greenhouse Gas Management Organization (TGO) | 17 |
| World Bank | 14 |
| Social Security Office, Ministry of Labour | 13 |
| Ministry of Tourism and Sports | 12 |
| National Statistical Office (NSO) | 12 |
| Bank of Thailand (data from the Trade Policy and Strategy Office) | 11 |
94 other publishers supply the remaining 196 values.
Checks
First reading
The compiler opens the page that holds the figure, copies the text with the number and the page number, and records the date checked, the unit, the year and where the value applies.
Second, independent check
A second reviewer who did not do the search downloads every source again and compares the figure, unit, year and grade with the fresh copy. Anything they cannot confirm stays out of the bank.
Automated checks
A test fails when a value lacks any field listed above, when its quote holds no number, when its source is not a secure link to the page with the figure, when a derived value has no formula, or when the CSV download no longer matches the data.
Spot check of a random sample
Someone who did not compile the values checks a random sample against the sources, as set out below.
Grades
Each value carries a grade from A to D for its source. The grade rates the source, not the value.
| Grade | Kind of source | What it covers | Values |
|---|---|---|---|
| A | Official statistics or rates set in law | Published by the Thai state body that owns the figure or compiles the official statistics, such as the NSO, the Bank of Thailand, the Wage Committee, NHSO, SSO, TGO or OAE. | 198 |
| B | Research with Thai figures | Peer-reviewed studies, international bodies or Thai research institutes, such as TDRI, PIER and HITAP. | 71 |
| C | Professional and published reports | Reports by NGOs, foundations, professional bodies, or published SROI reports with Thai values. | 1 |
| D | Market prices, news or derived values | One provider's or one facility's price, an official figure read from news or another body's document, or a value derived with assumptions. | 58 |
How borderline cases are graded
- An official figure read from news or from another body's document: D.
- One provider's or one facility's own price: D.
- A value that needs assumptions to derive: D.
- A value worked out by plain arithmetic from one publisher's figures: that publisher's grade.
- Official statistics as published in the statistical tables of the National Statistical Office or the Bank of Thailand: A, even when another agency owns the series. The note names the owner, for example the Trade Policy and Strategy Office for the consumer price index.
- Figures from international databases: B.
What a grade does not tell you
- Whether the value fits your outcome or your stakeholders. Choose values that match the outcome first, then use the grade to choose between them.
- Whether the value is what the outcome is worth. Many grade A values are rates set in law, such as a minimum wage or a ceiling on a claim: floors, ceilings or benchmarks. The note says which.
- How precise the figure is. A grade D value can be right for its own setting, and a grade A value can be out of date, so always check the year.
Valuation methods
Each value records how its source arrived at a money value. The methods follow the families described in the SROI guide (Nicholls et al., 2012, p. 47), with the tools used to adjust values as a group of their own.
| Method | What it measures | Values |
|---|---|---|
| Market price | What a good or service costs to buy, or a rate fixed by law, regulation or a fee schedule. | 159 |
| Conversion factor | An index, exchange rate, emission factor, threshold or rate used to adjust other values; not the value of an outcome. | 47 |
| Replacement cost | What it would cost to achieve the same result another way. The SROI guide counts such market comparisons as usually less credible than proxies already used by credible third parties (Nicholls et al., 2012, p. 51). | 41 |
| Avoided cost | The cost of a service, treatment or loss that the outcome makes unnecessary. An average cost overstates what a small project saves (Nicholls et al., 2012, p. 50). | 37 |
| Derived | Computed from figures in the source, with the formula shown. | 33 |
| Willingness to pay | What people say they would pay to gain something, or accept to give it up, from a stated preference or contingent valuation study (Nicholls et al., 2012, p. 47). | 8 |
| Revealed preference | A value inferred from what people pay for related goods, such as household spending or house prices (Nicholls et al., 2012, p. 47). | 3 |
Derived values
36 values are computed from figures in their source rather than read from it directly, for example a cost per case from a total and a count, or a figure converted to another unit. Each shows its formula and inputs. Table 3 lists 33 of them as derived; the other 3 keep the method of the figures they come from.
24 of them keep the grade of their source, as plain arithmetic on one publisher's figures; the other 12 are graded D under the rules above.
Spot check
60 values drawn at random from edition 2, at least 4 per category, with a recorded seed. Drawn from edition 2 on 2 November 2026 with seed 20261102, so the same draw can be repeated.
A reviewer who did not compile the values opens each sampled value's source and checks the figure, unit, year and grade. Each result is recorded with its date; a value found wrong is corrected in the next edition and listed under corrections.
Checked so far: 0 of 60.
What the check can show: if none of the 60 values is wrong, it supports, at 95% confidence, a share of wrong values in edition 2 below 4.9%; with one wrong value the bound would be 7.7%. These are exact binomial bounds (Clopper & Pearson, 1934). Because the draw keeps at least four values per category, small categories are over-represented, so the bound is a guide for the bank as a whole rather than an exact figure. Values added after edition 2 were not in the draw.
Until the check is complete, the bank is marked under review on its pages and in its data files. The result will be published on the record page of edition 2.
Editions and corrections
Each edition is frozen when it is published. Its CSV and JSON files and the Thai and English PDFs of its record page are kept exactly as they were, with SHA-256 checksums, and the record page reads only those files, so a citation of an edition always leads to the same values.
The CSV is described column by column in a data dictionary and as a Frictionless Tabular Data Package (Frictionless Data, n.d.).
Value codes are never reused. A value taken out keeps its code on the list of values left out.
An error found after publication is never fixed in a frozen file. It is listed under corrections with the date it was found, and fixed in a later edition.
Method changes by edition
Edition 2, published 3 November 2026 · Edition 2 record
- Every value had a second check by a reviewer who did not do the search, comparing the figure, unit, year and grade with a fresh download of the source.
- Stricter grading: an official figure read from news or another body's document, one provider's or one facility's own price, and a value that needs assumptions are graded D; plain arithmetic from one publisher's figures keeps that grade.
- Each value names the groups it serves, so the bank can be filtered by who a project helps.
- Released under CC BY 4.0 from 2 November 2026.
- A reviewer spot-checks a random sample of 60 values against their sources; until that is done the edition is marked under review.
Edition 3, published 11 November 2026 · Edition 3 record
- All 278 values carried over from edition 2 were checked again against fresh downloads of their sources: a machine check confirmed that the quoted text and the figure are still in the document, and a researcher re-read the source wherever the machine could not confirm them. 250 were confirmed and 28 fixed: 11 errors, listed under corrections after publication, and 17 fixes to wording or source references that leave the value as it was. Each value's last check date is in its verified field. Quotes in which PDF text extraction had dropped or moved Thai tone marks and vowels were restored from the printed page.
- A newer year of a value is added as a new row, and the older row points to it. A published value is changed only to correct an error. Every change of value, year, source or grade, and every withdrawal, is listed in the edition's change log, and every error found in an earlier edition, including a description that misstated what a value measures, is listed under corrections after publication. Wording fixes that change neither a value nor its meaning are not listed.
- Grade A also covers official statistics as published in the statistical tables of the National Statistical Office and the Bank of Thailand, even when another agency owns the series; the note names the owner. Other copies of an official figure remain grade D.
Corrections found after publication
- AG-27 in edition 2, found 11 Nov 2026: Graded A, but the cost is the Office of the Cane and Sugar Board's figure read from the Office of Agricultural Economics' compendium, which the rubric grades D. (fixed in edition 3)
- AG-32 in edition 2, found 11 Nov 2026: Graded A, but the cost is the broiler production and marketing working group's figure read from the Office of Agricultural Economics' compendium, which the rubric grades D. (fixed in edition 3)
- EM-12 in edition 2, found 11 Nov 2026: Graded A, but the value combines two Social Security Office pages and assumes the new wage ceiling applies to this benefit, which the rubric grades D. (fixed in edition 3)
- EM-13 in edition 2, found 11 Nov 2026: Graded A, but the value combines two Social Security Office pages and assumes the new wage ceiling applies to this benefit, which the rubric grades D. (fixed in edition 3)
- EM-07 in edition 2, found 11 Nov 2026: The proxy read "formal workers", but the survey's average covers employees only, and the English outcome did not say what the Thai one says. (fixed in edition 3)
- IN-14 in edition 2, found 11 Nov 2026: The note described the Office of Industrial Economics' value added as sales less all selling and administrative costs, but the Office adds back sales and administrative staff pay, rent, bad debts and exchange losses. (fixed in edition 3)
- IN-15 in edition 2, found 11 Nov 2026: The note described the Office of Industrial Economics' value added as sales less all selling and administrative costs, but the Office adds back sales and administrative staff pay, rent, bad debts and exchange losses. (fixed in edition 3)
- IN-16 in edition 2, found 11 Nov 2026: The note described the Office of Industrial Economics' value added as sales less all selling and administrative costs, but the Office adds back sales and administrative staff pay, rent, bad debts and exchange losses. (fixed in edition 3)
- EN-09 in edition 2, found 11 Nov 2026: Dated 2025, with a note saying the totals cover fiscal years 2016 to 2025, but the open data runs into fiscal year 2026; the note's price range also started at THB 18 instead of 15. (fixed in edition 3)
- EN-10 in edition 2, found 11 Nov 2026: Dated 2025, but the open data it comes from runs into fiscal year 2026. (fixed in edition 3)
- EN-11 in edition 2, found 11 Nov 2026: Dated 2025, but the open data it comes from runs into fiscal year 2026. (fixed in edition 3)
Limits
- Values describe a year and a place. 75 describe years before 2021; bring them to your project's year before use, and use the value for your project's area where the bank has one.
- Most money values are prices and costs (233 of 273). They show what a good or service costs, which is not always what a change is worth to the people who experience it. Values should reflect what matters to stakeholders (Social Value International, n.d.), so test the proxies that matter most with them.
- An average cost per case overstates what a small project saves, because fixed costs do not fall when there are a few fewer cases (Nicholls et al., 2012, p. 50).
- 17 values come from one facility or provider and may not hold elsewhere.
- Quotes stay in the language of the source. The notes are NXT's, in Thai and English.
Known gaps are listed under scope and limits on the bank page
SROI principles
Social Value International sets out eight principles for accounting for social value (Social Value International, n.d.). Four bear directly on financial proxies.
Value the things that matter (principle 3)
Value is the relative importance of different outcomes, informed by stakeholders' preferences, and financial proxies are one way to express it (Social Value International, n.d.). The bank offers documented starting points; whether a value reflects what matters to your stakeholders is for your analysis to test.
Do not overclaim (principle 5)
Only the value an activity is responsible for creating should be claimed (Social Value International, n.d.). The bank gives a value per unit of outcome; deadweight, displacement, attribution and drop-off are applied in the analysis (Nicholls et al., 2012, p. 62), as the how-to guide explains.
Be transparent (principle 6)
Each decision should be explained and documented, including the sources of information (Social Value International, n.d.). Every value carries its source, the quoted text, its year and its grade, so a report can describe each proxy and its source as the SROI guide's report format asks (Nicholls et al., 2012, p. 83).
Verify the result (principle 7)
Any account of value involves judgement, so it needs appropriate independent assurance (Social Value International, n.d.). The second check, the spot check and the frozen editions let anyone verify each value. A review by researchers outside NXT, with no part in compiling the bank, would add to these checks; the record of each edition will say if one has taken place.
References
Works cited on this page. Each was opened and read at the pages cited.
- Clopper, C. J., & Pearson, E. S. (1934). The use of confidence or fiducial limits illustrated in the case of the binomial. Biometrika, 26(4), 404–413. https://doi.org/10.1093/biomet/26.4.404 Opened 11 Nov 2026.
- Frictionless Data. (n.d.). Tabular data package. Data Package (v1) specifications. https://specs.frictionlessdata.io/tabular-data-package/ Opened 11 Nov 2026.
- Nicholls, J., Lawlor, E., Neitzert, E., & Goodspeed, T. (2012). A guide to social return on investment (Rev. ed.). The SROI Network. https://www.socialvalueint.org/s/The-SROI-Guide-2012.pdf Opened 11 Nov 2026.
- Page, M. J., McKenzie, J. E., Bossuyt, P. M., Boutron, I., Hoffmann, T. C., Mulrow, C. D., Shamseer, L., Tetzlaff, J. M., Akl, E. A., Brennan, S. E., Chou, R., Glanville, J., Grimshaw, J. M., Hróbjartsson, A., Lalu, M. M., Li, T., Loder, E. W., Mayo-Wilson, E., McDonald, S., . . . Moher, D. (2021). The PRISMA 2020 statement: An updated guideline for reporting systematic reviews. BMJ, 372, Article n71. https://doi.org/10.1136/bmj.n71 Opened 11 Nov 2026.
- Social Value International. (n.d.). The principles of social value. https://www.socialvalueint.org/principles Opened 11 Nov 2026.
How to cite
Cite the edition of the bank you used, not this page:
NXT Consulting Group. (2026). Thai financial proxy bank for SROI (Edition 3) [Data set]. https://www.nxtconsultinggroup.com/en/sroi/financial-proxies/edition-3
Other formats (APA, Thai, Chicago, BibTeX, BibLaTeX, RIS) are in the citation section of the bank