{"asof":"2026-09-30","attribution":"Source: Daall Index (index.daall.sa)","licence":"Free to use with attribution","generated":"2026-10-29","data":[{"id":"procurement-leading-indicator","title":"Contract-award data as a leading indicator of government spending, GDP and firm activity: evidence review for Daall Index","summary":"US research finds that contract obligations come before recorded government spending. US defence obligations lead payment-on-delivery, when national accounts record G, by about 2-3 quarters on average, and the early GDP response to defence news shows up in contractor inventories (Briganti and Sellemi, an unpublished UCSD paper estimated mainly on 1951-1988 data; its arXiv version was withdrawn). Highway outlays lag funding by years (Leduc and Wilson). Firm-level studies find that winning contracts raises investment (US: 7-11 cents per dollar on average in the 2016 IMF working paper; 10-13 cents for financially constrained firms in the 2021 JME version) and growth (Brazil: +2.2 pp in the quarter of the win). Award flows are concentrated in a few firms and sectors and dominated by firm-specific swings (Big G). No peer-reviewed or central-bank study was found that uses Etimad or other GCC award-level data.","date":"2026-10-29","model":"claude-opus-5-5","verified_on":"2026-10-30","findings":[{"text":"Ramey (NBER WP 15464, 2009): narrative defence-news shocks and professional forecasts Granger-cause standard VAR spending shocks, so the VAR shocks are dated too late; the news-based multipliers are 0.6-1.1.","grade":"B","status":"verified"},{"text":"Ramey cites Fisher and Peters' multiplier of 1.5, estimated from excess returns on defence-contractor stocks, but notes those stocks did not outperform in WWII or the Korean War (because of an excess-profits tax) and moved early in the late 1970s on foreign arms sales, so equity-based signals depend on the period.","grade":"B","status":"verified"},{"text":"Briganti and Sellemi (UCSD working paper, May 2023 draft; the arXiv version 2212.06073 was withdrawn by the authors): US defence contract obligations lead payment-on-delivery, which is when national accounts record G, by 2-3 quarters on average. This is estimated mainly on 1951-1988 Business Conditions Digest data, with FPDS 2000-2020 as a robustness check. The early GDP rise after defence news comes mainly from contractor inventories, and obligation shocks (1947-2015) Granger-cause VAR-identified spending shocks but not defence news shocks.","grade":"B","status":"corrected"},{"text":"Leduc and Wilson (FRBSF WP 2012-04): state highway outlays often lag well behind federal funding obligations; highway funding shocks raise state GDP on impact and again after 6-8 years, with no permanent effect 10 years out.","grade":"B","status":"verified"},{"text":"Fieldhouse and Mertens (Dallas Fed WP 2316, updated 2026) summarise prior work: highway grants lead outlays and capture the timing of economic effects more precisely (Leduc-Wilson), appropriations give larger and more precise multipliers (Brunet), and defence contract obligations capture inventory and time-to-build dynamics better than outlays (Briganti-Sellemi).","grade":"B","status":"verified"},{"text":"Cox, Muller, Pasten, Schoenle and Weber, 'Big G' (Banco Central de Chile WP 878, 2020; JPE 2024), using the universe of US procurement contracts: spending is concentrated in few firms and sectors, contracts are short-lived, and idiosyncratic (firm-specific) variation dominates fluctuations.","grade":"B","status":"verified"},{"text":"Hebous and Zimmermann (IMF WP 2016/060): each $1 of federal procurement raises a contractor's capital investment by 7-11 cents on average, close to zero for unconstrained firms; the published version (JME 2021) reports 10-13 cents over four quarters for financially constrained firms and no effect for unconstrained ones.","grade":"B","status":"corrected"},{"text":"Ferraz, Finan and Szerman (NBER WP 21219, 2015), using all Brazilian federal procurement auctions 2004-2010: winning at least one contract in a quarter raises firm growth by 2.2 pp that quarter; 93% of the new hires come from unemployment, the informal sector or outside the labour force; effects last beyond the contract term.","grade":"B","status":"verified"}],"implications":["Test the lead directly: whether quarterly award value leads Ministry of Finance spending on goods, services and capital assets, and GASTAT government consumption and investment. Treat the 2-3 quarter award-to-payment lag found for US defence goods (estimated mainly on 1951-1988 data, in an unpublished draft) as a hypothesis, not an expected value, and publish the estimated lag by sector.","Alongside the trailing-365-day level, publish monthly flow and 'surprise' series (awards relative to budget-implied expectations), because the research finds the unanticipated part of spending drives activity and a 365-day window delays turning points.","Use the issuers matched to Tadawul (8.7% of 2025 award value) for event studies from the day an award becomes public (median 5 days from award to first sight for awards dated 7-27 Sep 2026), and test whether awards predict issuer revenue, backlog and capital spending at horizons of 1-4 quarters.","Publish uncapped versions and concentration measures (HHI, top-10 share) for each sector, and test for and adjust any year-end award spikes, because award flows are concentrated and dominated by single large contracts.","Present the GDP link conservatively: offer the index mainly as a sector and firm demand signal; any value for nowcasting GDP still has to be shown out of sample."],"sources":[{"title":"Why Does GDP Move Before G? It's all in the Measurement","url":"https://arxiv.org/abs/2212.06073","type":"working_paper","grade":"B"},{"title":"Why Does GDP Move Before Government Spending? It's all in the Measurement","url":"https://edoardobriganti.github.io/research/paper_draft_MAR2023_whydoesgdp.pdf","type":"working_paper","grade":"B"},{"title":"Identifying Government Spending Shocks: It's All in the Timing","url":"https://www.nber.org/system/files/working_papers/w15464/w15464.pdf","type":"working_paper","grade":"B"},{"title":"Roads to Prosperity or Bridges to Nowhere? Theory and Evidence on the Impact of Public Infrastructure Investment","url":"https://www.frbsf.org/wp-content/uploads/wp12-04bk.pdf","type":"working_paper","grade":"B"},{"title":"A Narrative Analysis of Federal Appropriations for Research and Development","url":"https://www.dallasfed.org/-/media/documents/research/papers/2023/wp2316r1.pdf","type":"working_paper","grade":"B"},{"title":"Big G","url":"https://www.bcentral.cl/en/web/banco-central/content/-/details/working-papers-n-878","type":"working_paper","grade":"B"},{"title":"Can Government Demand Stimulate Private Investment? Evidence from U.S. Federal Procurement","url":"https://www.imf.org/en/publications/wp/issues/2016/12/31/can-government-demand-stimulate-private-investment-evidence-from-u-s-43784","type":"working_paper","grade":"B"},{"title":"Procuring Firm Growth: The Effects of Government Purchases on Firm Dynamics","url":"https://www.nber.org/system/files/working_papers/w21219/w21219.pdf","type":"working_paper","grade":"B"}],"verification":{"checked":8,"corrected":2,"removed":0}},{"id":"awards-and-equities","title":"Government contract awards and listed-firm outcomes: evidence for the Daall Index","summary":"Studies of contract-award announcements find small short-window effects: about +0.5% on day -1 for US-dominated corporate-government contracts (1990-2000, working paper); about +1.2% (market model; 1.4% under robustness models) for Saudi NREP renewable awards in a single 42-observation study whose text and tables disagree (Belgacem & Louhichi 2026); and, per an unverified 2026 SSRN abstract, -0.2% to -0.7% in China, with losses at high-valuation and private suppliers. Reactions rise with contract length and, for government contracts, only weakly with size relative to assets. The small-firm effect is found for inter-corporate contracts only, and the Saudi study finds larger reactions for domestic firms and larger projects. Price movement before the media date is likely partly a reporting lag. Dependence on government demand mainly makes firms more resilient and helps constrained firms invest: US contractors held up much better in 2008-09 (Goldman 2020), constrained firms invest 10-13 cents per procurement dollar (Hebous & Zimmermann, JME 2021), and Brazilian winners grew faster for longer (Ferraz et al. 2015). Gulf evidence is very thin.","date":"2026-10-29","model":"claude-opus-5-5","verified_on":"2026-10-30","findings":[{"text":"Elayan, Pukthuanthong & Roll (working-paper draft, July 2005): across 1,963 corporate-government contract announcements (mostly US federal and local; Dow Jones, 1990-2000, contracts over $1m), the day -1 abnormal return was +0.54%, versus +1.13% for inter-corporate contracts.","grade":"B","status":"corrected"},{"text":"Elayan et al.: for government contracts, announcement returns rise with contract term and with the number of competitors, and only weakly (10% level) with contract size relative to total assets; firm size is not significant there (smaller firms gain more only in inter-corporate contracts). The leakage evidence (positive returns on days -5 to -2) comes from the inter-corporate sample; in the government sample only day -2 is significantly positive (+0.18%).","grade":"B","status":"corrected"},{"text":"Belgacem & Louhichi (JRFM 2026): 42 firm-events (26 Saudi-listed, 16 foreign-listed sponsors) across 21 Saudi NREP solar and wind awards, 2017-2024. The market-model CAAR over days -1 to +1 was +1.21% (BMP t=2.56). Robustness checks give +1.42% (market-adjusted, t=2.31) and +1.39% (CAPM, t=2.27). Returns were higher for Saudi-listed (domestic) sponsors (0.94, p=0.021 in the text; 0.98, p=0.023 in Table 4) and rose with log project capacity (0.37, p=0.048 in the text; 0.43, p=0.031 in Table 4). Text and tables disagree on several numbers.","grade":"A","status":"corrected"},{"text":"Goldman (RFS 2020): by 2009, US federal contractors had 15% higher market capitalization, 18% higher capital expenditures and 26% more bank credit than similar firms, because procurement contracts insulated them from the 2008-09 crisis.","grade":"A","status":"verified"},{"text":"Hebous & Zimmermann (IMF WP 2016/060): each $1 of US federal procurement raises firm capital investment by 7-11 cents on average, more for financially constrained firms and close to zero for unconstrained ones. The published version (JME 2021) reports 10-13 cents over four quarters for constrained firms and no effect for unconstrained firms.","grade":"B","status":"corrected"},{"text":"Ferraz, Finan & Szerman (NBER w21219, 2015): in Brazil (2004-2010), winning at least one federal contract in a quarter raises firm (employment) growth by 2.2 percentage points that quarter; the effect lasts beyond the contract, partly because winners take part in and win more later auctions.","grade":"B","status":"verified"}],"implications":["Run event studies from the earliest public timestamp, meaning whichever comes first of the Tadawul disclosure and the date the award first appears on Etimad, not the award date. Keep point-in-time snapshots so backtests cannot use information before it was published, and report price movement before the event, since leakage is common.","Test an issuer-level surprise measure (award value relative to trailing revenue or market value, spread over the contract's length) on uncapped values, since sector caps remove some of the largest awards. Evidence that relative size matters for government contracts is weak (t = 1.77 in Elayan et al.), so this is a hypothesis to test, not a finding.","Test, rather than assume, whether issuer results differ by valuation and by ownership (PIF or state-linked versus private); the Chinese evidence for such a split could not be read in full.","Build a government-dependence ratio for each Tadawul issuer and test whether it predicts resilience during oil-price and budget downturns, capital spending, and credit access, rather than only short-term returns.","Audit name matching against commercial registration numbers, subsidiaries and consortia/JVs, since only 8.7% of 2025 value is currently matched. Exclude direct purchases (1.3% of value) from event studies, and use large samples because expected effects are only around 0.5-1.5%."],"sources":[{"title":"The Valuation Effect and Determinants of Corporate Contracting","url":"https://www.anderson.ucla.edu/documents/areas/fac/finance/20-05.pdf","type":"working_paper","grade":"B"},{"title":"To Reward or Not to Reward? Stock Market Reaction to Renewable-Energy Project Awards","url":"https://doi.org/10.3390/jrfm19020139","type":"peer_reviewed","grade":"A"},{"title":"Government as Customer of Last Resort: The Stabilizing Effects of Government Purchases on Firms","url":"https://wrap.warwick.ac.uk/id/eprint/169461/","type":"peer_reviewed","grade":"A"},{"title":"Can Government Demand Stimulate Private Investment? Evidence from U.S. Federal Procurement","url":"https://www.imf.org/en/publications/wp/issues/2016/12/31/can-government-demand-stimulate-private-investment-evidence-from-u-s-43784","type":"working_paper","grade":"B"},{"title":"Procuring Firm Growth: The Effects of Government Purchases on Firm Dynamics","url":"https://www.nber.org/papers/w21219","type":"working_paper","grade":"B"},{"title":"When Public Demand becomes Bad News: Government Procurement and Stock Market Reactions","url":"https://papers.ssrn.com/sol3/papers.cfm?abstract_id=7573758","type":"working_paper","grade":"B"}],"verification":{"checked":7,"corrected":4,"removed":1}},{"id":"index-methodology","title":"Building and governing a Saudi government-demand index from Etimad awards: practices worth borrowing","summary":"Conference Board composites weight components by inverse volatility, revise only a six-month window between annual January recomputes, and rebase to 2016=100. S&P Global PMIs are diffusion indices centred on 50 with fixed headline weights, where raw survey data are not revised but seasonal factors may be. For Daall's award-value index, the main design risks are concentration in very large contracts, the lag of a trailing-365-day sum (centred about six months back, an arithmetic property not taken from these sources), moving Hijri holidays and upward revisions from late-appearing awards. The Census Genhol utility can build Ramadan/Eid regressors (with overall-mean centring for holidays that cycle through the year). The IOSCO Principles (FR07/13, July 2013), now verified from the primary text, provide a governance framework to follow voluntarily. A vintage archive is a sound practice, but the Philadelphia Fed real-time data set is not among the cited sources. Evidence on Hijri effects is limited to Pakistani monetary series (abstract checked), with nothing found for Saudi procurement, and the equity providers' capping rulebooks were not reviewed.","date":"2026-10-29","model":"claude-opus-5-5","verified_on":"2026-10-30","findings":[{"text":"The Conference Board weights each composite-index component by the inverse of the standard deviation of its monthly changes (normalised levels for diffusion indexes and spreads), with the factors scaled to sum to one. When a component is unavailable, the others' factors are recomputed that month to sum to one, and missing data are otherwise imputed autoregressively (Conference Board, Calculating the Composite Indexes).","grade":"B","status":"verified"},{"text":"The Conference Board builds index levels with a symmetric percent-change formula, rebases them to a 2016 average of 100, updates only the latest and previous six months between annual recomputes, and recomputes the standardisation factors and full history once a year, usually in January (Conference Board, Calculating the Composite Indexes).","grade":"B","status":"verified"},{"text":"S&P Global's PMI diffusion index is % 'higher' plus half of % 'unchanged' (0-100, with 50 as no change) and is then seasonally adjusted. The headline PMI weights New Orders 30%, Output 25%, Employment 20%, inverted Suppliers' Delivery Times 15% and Stocks of Purchases 10% (S&P Global US Manufacturing PMI release, 1 April 2026).","grade":"B","status":"verified"},{"text":"S&P Global states that underlying PMI survey data are not revised after publication. Seasonal adjustment factors may be revised from time to time, which changes the seasonally adjusted series (S&P Global US Manufacturing PMI release, 1 April 2026).","grade":"B","status":"verified"},{"text":"The US Census Bureau's Genhol utility (28 August 2020) generates user-defined moving-holiday regressors for X-13ARIMA-SEATS, with separate before (from as early as 42 days ahead), interim and after (up to 49 days) windows. For holidays like Ramadan that move through all calendar months, it recommends centring on the overall mean rather than calendar-month means, and its own Idul Fitri example does this.","grade":"B","status":"corrected"},{"text":"A State Bank of Pakistan / MPRA working paper (Bukhari, Jalil & Rao, 2011) finds that Ramadan and Zilhaj (Dhul-Hijja) raise currency holdings and lower deposits in Pakistani monthly data, and warns that ignoring such seasonal factors causes omitted-variable bias and poor forecasts (verified from the abstract).","grade":"B","status":"verified"},{"text":"IOSCO's Principles for Financial Benchmarks (Final Report FR07/13, July 2013) group 19 principles under governance, quality of the benchmark, quality of the methodology and accountability, to be applied in proportion to the size and risks of each benchmark and administrator. Principle 7 (Data Sufficiency) requires anchoring in an active market of observable arm's-length transactions, but not that every determination be built solely from transaction data.","grade":"B","status":"verified"}],"implications":["Publish a monthly seasonally adjusted DGI alongside the trailing-365-day headline, using X-13ARIMA-SEATS with Genhol-built Ramadan, Eid al-Fitr and Eid al-Adha variables centred on the whole-sample mean (as the US Census Bureau advises for holidays that move through every month), and keep each effect only if it tests significant.","Re-examine the cap: at the DGI level the 10% cap has never bound (the largest award since 2023 moves the index by 4.4 points). Keep sector caps set from data available at the time, publish capped and uncapped series with the number and value of capped contracts, and test spreading multi-year awards over their contract terms. This is a design proposal; no source in this note tests it.","Add count-based and value-based diffusion indices (share of contracting entities or sectors with year-on-year growth in 90-day awards, centred on 50), since direct purchases were 66.5% of 2025 awards by count but 1.3% by value (Daall measurement).","Build a point-in-time vintage archive, which is not yet in place: keep each award's first-seen date (recorded since 30 Jul 2026, near-complete from September 2026), store every published release, publish revision tables and completeness ratios, and fix a revision window with one annual January benchmark.","Publish a methodology mapped to the IOSCO principles, covering governance and oversight, minimum data-sufficiency thresholds per sector, change and consultation procedures, an audit trail and a complaints channel, plus a periodic external review."],"sources":[{"title":"Calculating the Composite Indexes","url":"https://www.conference-board.org/data/bci/index.cfm?id=2154","type":"index_provider","grade":"B"},{"title":"S&P Global US Manufacturing PMI News Release (March 2026 data; embargoed until 0945 EDT 1 April 2026)","url":"https://www.pmi.spglobal.com/Public/Home/PressRelease/c75dd7dd17c94989afb36aad2fbf7f0b","type":"index_provider","grade":"B"},{"title":"Genhol - a utility that generates user-defined moving holiday regressors for X-13ARIMA-SEATS","url":"https://www2.census.gov/software/x-13arima-seats/win-genhol/download/genholdoc.pdf","type":"official_document","grade":"B"},{"title":"Detection and Forecasting of Islamic Calendar Effects in Time Series Data: Revisited","url":"https://mpra.ub.uni-muenchen.de/31124/","type":"working_paper","grade":"B"},{"title":"IOSCO's Final Report on Principles for Financial Benchmarks (not opened)","url":"https://centerforfinancialstability.org/wp/2013/07/18/ioscos-final-report-on-principles-for-financial-benchmarks/","type":"other","grade":"C"},{"title":"Principles for Financial Benchmarks - Final Report (FR07/13)","url":"https://www.scribd.com/document/951227722/IOSCOPD415","type":"official_document","grade":"B"}],"verification":{"checked":7,"corrected":1,"removed":0}},{"id":"saudi-official-sources","title":"Benchmarking a Saudi procurement-award index: official and specialist sources, frequency, lag and fit","summary":"No official Saudi series measures government procurement commitments. Etimad-based awards can be validated against MoF quarterly cash outlays (use of goods and services SAR 316bn and capital spending SAR 169bn in 2025; the full-year report came 54 days after year end), the Contractors Authority's periodic project-award report (cadence and coverage not documented), and MEED Projects/Kamco data for the wider projects market (Saudi awards $84.5bn, about SAR 317bn, in MEED's initial 2025 count, since revised to about $102bn). None of these sources gives Etimad's own award volume; measured by Daall from the award record it was SAR 84-109bn a year in 2022-2025, VAT included. Two rule changes create structural breaks: the regional HQ procurement rule (1 Jan 2024) and the new procurement law (effective 2 Jan 2027), which raises the direct-purchase ceiling tenfold to SAR 1m. Etimad publication timing, SAMA procurement data and which PIF entities fall outside Etimad were not verified from primary documents.","date":"2026-10-29","model":"claude-opus-5-5","verified_on":"2026-10-30","findings":[{"text":"MoF Quarterly Budget Performance Report Q3 FY2025: use of goods and services was SAR 64.6bn, 73.6bn and 84.0bn in Q1-Q3 2025 (SAR 222.2bn for nine months). Per Arab News, citing the Q4 report, Q4 2025 capital spending was SAR 50.9bn vs 43.1bn a year earlier (+18%). Per Argaam, MoF released full-year actuals on 23 Feb 2026 (54 days after year end).","grade":"A","status":"verified"},{"text":"GASTAT's Q4 2025 GDP release, published 9 Mar 2026 (68 days after year end), shows real government final consumption fell 3.5% and gross fixed capital formation fell 1.7% in 2025. As a real national-accounts aggregate that includes the public wage bill, it is only a loose macro check on award values.","grade":"A","status":"verified"},{"text":"New Government Tenders and Procurement Law (published 4 Sep 2026, effective 2 Jan 2027; Baker McKenzie; Royal Decree M/76 per CMS). The direct-purchase ceiling rises from SAR 100,000 to SAR 1m (Art. 32). The limit up to which an agency head may delegate procurement decisions rises from SAR 10m to 50m (Art. 51). MoF contract review falls from 15 to 4 working days (Art. 57). An agency with unpaid contractor dues that ignores an MoF notice may not issue new award decisions, subject to exceptions (Art. 91).","grade":"C","status":"corrected"},{"text":"Clyde & Co: Council of Ministers Resolution 245 (26 Nov 2019), the Regulations on Preference for Local Content and Local SMEs and Companies Listed on the Capital Market, sets a Mandatory List of National Products. For national products not on that list, competing foreign products are evaluated as if priced 10% higher. Baker McKenzie notes that the 2026 law requires new Preference Regulations.","grade":"C","status":"verified"},{"text":"Regional HQ rule (in force since 1 Jan 2024; Arab News, Aurne): contracts up to SAR 1m are exempt. A non-RHQ multinational's offer can be accepted only if it is the sole technically compliant bid, or if it ranks best technically and is at least 25% cheaper than the second-best bid. Since Nov 2025 contracting entities request exemptions through Etimad.","grade":"C","status":"verified"},{"text":"Saudi Contractors Authority (via Arab News/Al-Eqtisadiah, 13 Jul 2026): about 117,000 registered contractors. A recent Authority report counted 25 projects worth over SAR 29.5bn awarded in June 2026, second in value to May (over SAR 30bn). The report's cadence, release date and project-size cut-off are not stated.","grade":"C","status":"corrected"},{"text":"MEED Projects (initial data, Jan 2026): Saudi contract awards fell from a record $164bn in 2024 to $84.5bn in 2025. Kamco, using MEED data, shows a 72.5% y/y fall in Q2 2025 to $9.8bn. MoF's Q2 FY2025 report (not in the original list) shows capital spending down 38.6% in the same quarter. A Skift summary of an Oct 2026 MEED webinar gives $102bn for 2025, implying an upward revision of about 21%.","grade":"C","status":"corrected"},{"text":"PIF FY2024 annual report (via CNBC/AGBI): giga-project investments fell 12.4% to SAR 211bn (6% of assets vs 8%), with little giga-project detail and no project-level write-down disclosure. MEED (via Skift, Oct 2026) puts giga-project awards at $31.7bn (2023), $26.7bn (2024), $13.6bn (2025) and $10.6bn so far in 2026. Cumulative giga-project awards of about $117bn are roughly 17% of the $700bn awarded in Saudi Arabia over the past decade.","grade":"C","status":"corrected"}],"implications":["Publish a quarterly reconciliation of award value against MoF use of goods and services and capital spending, with coverage ratios and lead/lag correlations. State that the index covers awards published on Etimad; which bodies (PIF, giga-projects, Aramco) fall outside has not yet been established from primary documents.","Flag a structural break at 2 Jan 2027: track direct-purchase share by count and value (now 66% and 1.3%) in the SAR 100k-1m band, and test whether Etimad still publishes direct purchases fully once the ceiling rises tenfold.","Add a break flag at Jan 2024 for the regional HQ rule and track foreign or non-HQ winners' share by sector, especially for contracts above SAR 1m.","Track the listed-issuer share (8.7% of 2025 value) against the LCGPA preference for listed companies and against the new Preference Regulations once issued, and record mandatory-list and local-content fields where Etimad exposes them.","Use the Contractors Authority project-award report and MEED Projects (recording each data vintage, because MEED revises its counts) as external construction benchmarks; the Authority report's cadence and coverage are not documented. Use the new 4-working-day MoF review and Art. 91 to watch for changes in the award-to-publication lag (median 5 days from award to first sight for awards dated 7-27 Sep 2026)."],"sources":[{"title":"Saudi Arabia: Key Changes Under New Government Procurement Law","url":"https://www.bakermckenzie.com/en/insight/publications/2026/10/saudi-arabia-key-changes-under-new-government-procurement-law","type":"other","grade":"C"},{"title":"Local Content in Saudi Public Procurement: Key Considerations and the Broader Localisation Landscape","url":"https://www.clydeco.com/en/insights/2026/09/local-content-in-saudi","type":"other","grade":"C"},{"title":"Contract awards decline in 2025","url":"https://www.meed.com/contract-awards-decline-in-2025","type":"industry_report","grade":"C"},{"title":"Saudi Giga Project Spending Shrinks. Hotel Developers Keep Building","url":"https://skift.com/2026/10/28/saudi-giga-project-spending-shrinks-hotel-developers-keep-building/","type":"news","grade":"C"},{"title":"GCC Projects Market Update (Q2-2025)","url":"https://www.kamcoinvest.com/sites/default/files/research/pdf/GCC%20Projects%20Market%20Update%20-%20Q2-2025_Eng.pdf","type":"industry_report","grade":"C"},{"title":"Quarterly Budget Performance Report, Q3 of FY 2025 (1446/1447H) - English translation of the official Arabic version","url":"https://www.mof.gov.sa/en/financialreport/2025/Documents/Q3%202025%20-%20En.pdf","type":"official_statistics","grade":"A"},{"title":"Saudi Arabia's budget deficit widens to $25.3bn in Q4 2025 as spending rises","url":"https://www.arabnews.pk/business/saudi-arabias-budget-deficit-widens-to-253bn-in-q4-2025-as-spending-rises-2634216","type":"news","grade":"C"},{"title":"2025 actual budget: Revenues hit SAR 1.1T, spending at SAR 1.38T","url":"https://www.argaam.com/en/article/articledetail/id/1882586","type":"news","grade":"C"},{"title":"Gross Domestic Product for the Fourth Quarter of 2025","url":"https://www.stats.gov.sa/documents/d/guest/gdp-q4_2024-en-pdf","type":"official_statistics","grade":"A"},{"title":"GASTAT: Saudi Economy Achieves 4.5% Growth in 2025","url":"https://www.spa.gov.sa/en/N2532192","type":"news","grade":"C"},{"title":"Saudi Arabia allows contracting exceptions for firms without regional HQ","url":"https://www.arabnews.com/node/2633989/business-economy","type":"news","grade":"C"},{"title":"Saudi RHQ Program 2026 - Procurement Rule and Exemptions (aurne.org insight)","url":"https://aurne.org/insights/saudi-rhq-program-government-procurement-exemptions-2026","type":"vendor","grade":"C"},{"title":"Saudi Arabia has up to 117k contractors, small firms being qualified for major projects","url":"https://www.arabnews.com/node/2650710/business-economy","type":"news","grade":"C"},{"title":"What's behind the $8 billion hit to Saudi Arabia's massive wealth fund (URL slug: saudi-arabia-pif-fund-sees-8-billion-writedown-in-megaprojects)","url":"https://www.cnbc.com/2025/08/14/saudi-arabia-pif-fund-sees-8-billion-writedown-in-megaprojects.html","type":"news","grade":"C"},{"title":"PIF: Giga-projects and some unanswered questions","url":"https://www.agbi.com/analysis/giga-projects/2025/08/pif-giga-projects-and-some-unanswered-questions/","type":"news","grade":"C"}],"verification":{"checked":8,"corrected":4,"removed":0}},{"id":"competition-integrity","title":"Measurable procurement-integrity indicators (single bidding, direct awards, bidder counts, concentration, price outliers): definitions, index potential and biases for an Etimad-based index family","summary":"Practice from the EU Single Market Scoreboard, World Bank ProACT, the Open Contracting Partnership red-flags guide and Fazekas-style corruption risk indices converges on a few computable indicators: single-bid rate in competitive procedures, non-competitive (direct-award) share, bidder counts, supplier/buyer concentration, and price outliers within item categories. The Scoreboard and ProACT publish these as recurring aggregated indicators, and their documentation warns that indicators simplify reality, depend on country and market structure and data coverage, and signal risk rather than wrongdoing. Validation evidence is real but imperfect. Hungarian CRI scores track firm profitability, contract-value-to-estimate ratios, political links and tax-haven registration. In Italian roadwork contracts, OLS/LASSO/Ridge predictions of police-flagged winners misclassify about 26% of cases as false positives and 17% as false negatives (Decarolis & Giorgiantonio 2022). For Daall Index the binding constraints are Etimad field availability (bid counts, item-level prices, supplier IDs) and the structural break from Saudi Arabia's new procurement law. A law-firm summary says it was approved 5 Aug 2026, effective 120 days after Gazette publication, and raises the direct-purchase threshold tenfold (SAR 100k to 1m) while widening eligible categories.","date":"2026-10-29","model":"claude-opus-5-5","verified_on":"2026-10-30","findings":[{"text":"EU Single Market Scoreboard (2021 edition): 12 indicators each add +1 (satisfactory) or -1 (unsatisfactory) by default, with single bidder, no calls for bids and publication rate triple-weighted and indicators 7-12 one-third weighted. Single bidder (excluding framework agreements and direct awards) is green at <=10% and red above 20%, and ranged from 2% (Iceland) to 51% (Poland).","grade":"B","status":"verified"},{"text":"Same Scoreboard: 'no calls for bids' (share of procedures negotiated without any call for bids) is green at <=5% and red at >=10%. Decision speed (mean days from bid deadline to award, open procedures excluding framework agreements) is green at <=120 days. The Commission says the indicators 'simplify reality' and that corruption is omitted or covered only indirectly.","grade":"B","status":"verified"},{"text":"Fazekas, Toth & King (2016, European Journal on Criminal Policy and Research 22(3)) build a contract-level corruption risk index from Hungarian e-procurement red flags linked to restricted competition and recurrent awards to the same company, aggregable to organisations, sectors, regions and countries. Higher-scoring firms show higher profitability, higher contract-value-to-estimate ratios, more politician management or ownership, and more tax-haven registration.","grade":"A","status":"verified"},{"text":"World Bank ProACT technical report (Al-Shaibani, Fazekas, Regos & Toth, June 2022): integrity indicators are scored 0-100 (100 = no risk, 0 = red flag, with a 50 mild level or continuous values for some). Their country-specific thresholds come from OLS regressions, with indicators chosen for their link to single bidding and supplier dependence on one buyer. Competition indicators are bidder counts, new market entry, supplier market share and non-local supplier. Listed limitations include small-value contracts exempt from publication, portal changes and changes in classification codes.","grade":"B","status":"corrected"},{"text":"OCP's 'Red flags in public procurement' (2024) describes 73 OCDS-mapped flags. They include: a single bid in an open/selective procedure (R018); tender values (R016) or line-item unit values (R017) beyond Q3+1.5*IQR within an item category; a supplier's share of a buyer's contracts above a user threshold such as 40% or an IQR fence (R040); market HHI above 1,800 or rising, with a 200-point change cited as significant (R051); more than one direct award just below the threshold (R049); and summed direct awards to one supplier exceeding it (R055).","grade":"B","status":"corrected"},{"text":"Decarolis & Giorgiantonio (2022, EPJ Data Science 11:16) predict whether a contract's winner is a police-flagged firm from red flags in 12,786 Italian roadwork contracts (2009-2015). OLS, LASSO and Ridge misclassify about 26% of cases as false positives and 17% as false negatives. Random forests improve accuracy, and indicators the anticorruption authority does not track add accuracy only in the random forest.","grade":"A","status":"corrected"},{"text":"Best, Hjort & Szakonyi (NBER w23350, rev. 2022; AER 2023), using 16 million Russian public purchases, attribute 39% of the variation in prices paid for narrowly defined items to the individual bureaucrats and organisations managing procurement. (Inference, not from the paper: buyer-level price indices carry signal if items are tightly controlled.)","grade":"B","status":"verified"},{"text":"Clyde & Co (1 Oct 2026), reading the MoF summary of Saudi Arabia's new Government Tenders and Procurement Law (approved by the Council of Ministers 5 Aug 2026): the direct-purchase threshold rises from SAR 100,000 to SAR 1 million and its eligible categories widen, permitted contract variation rises from 10% to 20%, and MoF pre-signature review falls from up to 15 to up to 4 working days. The law takes effect 120 days after Official Gazette publication, with the full text then still unpublished.","grade":"C","status":"verified"}],"implications":["Audit Etimad fields before launching: publish a data-coverage sub-index (share of awards with bidder count, supplier ID, item classification, estimated value), modelled on the Scoreboard's missing-data indicators and ProACT's 0/100 transparency indicators; only compute single-bid and bidder-count indices if bid counts are reliably present.","Publish a Non-Competitive Award Index both by count and by value (currently 66% vs 1.3%), mark the break when the new procurement law takes effect, and add a threshold-bunching monitor (density of awards within 1-5% below SAR 100k and SAR 1m) instead of reading level changes as behaviour.","Compute sector concentration (HHI, top-5 share, distinct suppliers, new-entrant share) for each Tadawul industry group on UNCAPPED award values, after resolving supplier names to commercial-registration identifiers; report OCP/DOJ bands (1,800 and a +200 change) and minimum-market-size filters to avoid small-market artefacts.","Treat price outliers as an experimental product: only within narrow item categories with unit prices, using Q3+1.5*IQR fences and buyer fixed effects; if unit prices are unavailable, use award value relative to estimated value instead, and label all outputs as risk/efficiency signals, not corruption measures.","Publish methodology, revision policy (records arrive late: median 5 days from award to first sight for awards dated 7-27 Sep 2026) and a validation plan, for example correlating flags with contract changes after award or with outcomes at matched listed issuers; OECD MAPS-type benchmarks were not reviewed in this note."],"sources":[{"title":"Single Market Scoreboard - Public Procurement","url":"https://ec.europa.eu/internal_market/scoreboard/_docs/2021/12/policy-areas/public-procurement_en.pdf","type":"official_document","grade":"B"},{"title":"An Objective Corruption Risk Index Using Public Procurement Data","url":"https://link.springer.com/article/10.1007/s10610-016-9308-z","type":"peer_reviewed","grade":"A"},{"title":"Red flags in public procurement: A guide to using data to detect and mitigate risks","url":"https://www.open-contracting.org/wp-content/uploads/2024/12/OCP2024-RedFlagProcurement-1.pdf","type":"other","grade":"C"},{"title":"open-contracting/cardinal-rs: Measure red flags and procurement indicators using OCDS data","url":"https://github.com/open-contracting/cardinal-rs","type":"other","grade":"C"},{"title":"ProACT: Using data to strengthen the efficiency and integrity of public procurement","url":"https://blogs.worldbank.org/en/governance/proact-using-data-strengthen-efficiency-and-integrity-public-procurement","type":"other","grade":"C"},{"title":"Procurement AntiCorruption and Transparency Platform (ProACT): Technical Report","url":"https://www.procurementintegrity.org/assets/about/ProACT_methods_paper_20220809_final.pdf","type":"official_document","grade":"B"},{"title":"Corruption red flags in public procurement: new evidence from Italian calls for tenders","url":"https://link.springer.com/article/10.1140/epjds/s13688-022-00325-x","type":"peer_reviewed","grade":"A"},{"title":"Individuals and Organizations as Sources of State Effectiveness","url":"https://www.nber.org/system/files/working_papers/w23350/w23350.pdf","type":"working_paper","grade":"B"},{"title":"Saudi Arabia's new Government Tenders and Procurement Law: An Overview of the Key Reforms","url":"https://www.clydeco.com/en/insights/2026/10/saudi-arabia-new-government-tenders-procurement","type":"vendor","grade":"C"}],"verification":{"checked":8,"corrected":3,"removed":0}},{"id":"arabic-entity-resolution","title":"Matching Arabic-script supplier names to tickers and LEIs: methods and how accuracy is measured","summary":"In OpenSanctions Pairs (an Oxford/OpenSanctions arXiv preprint, 2026, not peer reviewed), off-the-shelf LLMs scored 98.95% F1 (GPT-4o, zero-shot) and 98.23% (DeepSeek-R1-Distill-Qwen-14B) against 91.33% for the production rule matcher. These scores come from label-stratified samples of at most 10,000 pairs drawn from the 755,540-pair corpus, and the test sets differ by model. The absolute scores are inflated by a third of pairs that are trivially positive, and the paper tests LLMs alone, not a combined rules-plus-LLM pipeline. Cross-script transliteration (Latin/Cyrillic/Arabic) is one of two LLM failure modes the paper names, but Arabic is only 0.10% of its names. OpenSanctions' own matcher documentation reports higher cross-script error for Arabic/Farsi. Published Arabic name-matching studies (2006, 2014) cover person names only; I found no company-name benchmark. LSEG PermID publishes a 0-1 match score with Excellent/Good/Possible bands, and GLEIF publishes certified identifier-to-LEI files, but neither publishes precision or recall. Daall should therefore build and publish its own value-weighted evaluation; the 8.7% listed-issuer share is an internal figure not checked here.","date":"2026-10-29","model":"claude-opus-5-5","verified_on":"2026-10-30","findings":[{"text":"OpenSanctions Pairs (Smith et al., arXiv preprint 2026, not peer reviewed) releases 755,540 analyst-labelled pairs from 293 sources, but scores models on label-stratified samples of at most 10,000 pairs. On those samples zero-shot GPT-4o reached 98.95% F1 and an optimised DeepSeek-R1-Distill-Qwen-14B 98.23% (on 3,680 pairs), against 91.33% for the nomenklatura RegressionV1 rule matcher (on 8,000 pairs). v2 notes that a third of the corpus is trivially positive auto-merge pairs that inflate absolute F1, and it no longer claims parity with human labelling consistency, which v1 asserted without measuring.","grade":"B","status":"corrected"},{"text":"In OpenSanctions Pairs v1 the rule matcher at threshold 0.15 put 14.1% of all test pairs in false positives and 0.4% in false negatives (about 62% of true non-matches flagged). The paper names two unquantified LLM failure modes, cross-script transliteration (Latin/Cyrillic/Arabic) and off-by-one dates or identifiers. v1 suggests, without testing it, a high-recall rule filter followed by LLM re-ranking. Arabic-script names are only 0.10% of the benchmark's entity captions (v2).","grade":"B","status":"corrected"},{"text":"OpenSanctions' documentation for its own matcher says that fuzzy comparison across writing systems raises error rates for scripts including Arabic/Farsi, that Soundex/Metaphone do not support non-alphabet (abugida, logographic, syllabic) scripts, and that company matching is 'particularly vulnerable' to misspelt legal-form words (e.g. 'Lymited').","grade":"C","status":"verified"},{"text":"Census working paper CES 18-46 (Cuffe & Goldschlag 2018, MAMBA) standardises names (including dropping suffixes such as CO and INC), scores candidate pairs with about 20 string comparators and classifies them with a random forest trained on about 2,000 clerically reviewed pairs. The forest reaches cross-validated precision above 95%; a logistic regression on the same 1,900 pairs scores P 0.915 / R 0.926 / F1 0.920. On 700 hand-coded pairs, false negatives were 1.42% of pairs against 31.44% for SAS DQ, and false positives 12.51% against 2.70%; the authors note this test favours MAMBA.","grade":"B","status":"corrected"},{"text":"Peeters, Steiner & Bizer (EDBT 2025) find that zero-shot GPT-4 beats fine-tuned RoBERTa/Ditto on 3 of 4 English product benchmarks (by 2.7 to 4.7 F1) and trails them by 0.3 to 4.5 F1 on the other three of their six datasets. It is far more robust on unseen entities, and the best prompt has to be chosen per model and dataset.","grade":"A","status":"verified"},{"text":"Arabic name-matching evidence is old and limited to person names. Freeman, Condon & Ackerman (HLT-NAACL 2006) added Arabic-English character-equivalence classes to Levenshtein matching. On a 577-Arabic / 968-English news-name set, recall went from 57% (baseline at a 0.725 threshold) to about 67% (at 0.85), and to 82% with all enhancements. El-Shishtawy (IJCLCLP 2014), on synthetic distorted Arabic person names, raised the worst-case accuracy from 69% (Soft TF-IDF) to about 80%.","grade":"A","status":"corrected"},{"text":"GLEIF's ISO 20275 ELF code list v1.6 (19 Feb 2026) lists more than 3,600 legal forms across more than 200 jurisdictions. It has 9 active Saudi forms, e.g. 78VH شركة مساهمة and TTE8 شركه ذات مسؤولية محدودة, none with an abbreviation recorded (checked in the v1.6 CSV).","grade":"B","status":"verified"},{"text":"The PermID APIs User Guide (Refinitiv, Apr 2020) says Record Matching returns a 0-1 Match Score that 'indicates the probability that the match is correct', which users can use to pick matches for manual review, labelled Excellent >=0.9, Good >=0.6, Possible >=0.1. GLEIF runs a free LEI-mapping certification and publishes BIC-, ISIN-, MIC-, OpenCorporates- (and QCC-, GEM-) to-LEI files and S&P CIQ-to-LEI pairs via LEI search.","grade":"C","status":"verified"}],"implications":["Build a gold-standard set: a stratified, value-weighted sample of Etimad suppliers, each labelled by two reviewers. Publish precision and recall for ticker mapping by count and by SAR value, with confidence intervals, so the 8.7% listed share comes with an estimate of missed issuer value.","Use identifiers before names: commercial registration or Unified National Number where they can be obtained, LEI records (GLEIF lists the Saudi Commercial Register, RA000513, as a registration authority), and GLEIF ISIN-to-LEI files to reach tickers. Use fuzzy name matching only as a fallback, and write down how subsidiaries roll up to listed parents.","Run a layered pipeline: Arabic spelling normalisation; split legal-form and activity words (ELF Saudi codes plus a local variant list) into separate fields instead of deleting them; frequency-weighted word and character comparison; then a probabilistic or classifier score. Send mid-confidence pairs to LLM adjudication and all high-value awards to human review.","Ship a match-confidence band and match-method field with each mapped award, in the style of PermID's levels. Fix the version of the matching algorithm and log restatements when supplier mappings change.","Favour precision in issuer mapping, since false positives directly inflate listed-company exposure, and review misses on names written in both Arabic and Latin script, a failure mode one preprint names but does not quantify."],"sources":[{"title":"OpenSanctions Pairs: Large-Scale Entity Matching with LLMs (v1); v2 retitled in the PDF 'OpenSanctions Pairs: A Large-Scale Dataset for Pairwise Entity Matching'","url":"https://arxiv.org/html/2603.11051v1","type":"working_paper","grade":"B"},{"title":"Tuning the matching algorithm - OpenSanctions","url":"https://www.opensanctions.org/docs/api/tuning/","type":"vendor","grade":"C"},{"title":"Squeezing More Out of Your Data: Business Record Linkage with Python","url":"https://www2.census.gov/ces/wp/2018/CES-WP-18-46.pdf","type":"working_paper","grade":"B"},{"title":"Entity Matching using Large Language Models","url":"https://arxiv.org/pdf/2310.11244","type":"peer_reviewed","grade":"A"},{"title":"LEI Mapping - LEI Data - GLEIF","url":"https://www.gleif.org/en/lei-data/lei-mapping","type":"official_document","grade":"B"},{"title":"PermID APIs User Guide (Apr 2020; 'Date of Issue: Apr 2019')","url":"https://developers.lseg.com/content/dam/devportal/api-families/open-permid/permid-entity-search/documentation/permid-apis-user-guide-apr-2020.pdf","type":"vendor","grade":"C"},{"title":"ISO 20275: Entity Legal Forms Code List (v1.6, 2026-02-19)","url":"https://www.gleif.org/en/lei-data/code-lists/iso-20275-entity-legal-forms-code-list","type":"official_document","grade":"B","added_in_verification":true},{"title":"Cross Linguistic Name Matching in English and Arabic: A 'One to Many Mapping' Extension of the Levenshtein Edit Distance Algorithm","url":"https://aclanthology.org/N06-1060/","type":"peer_reviewed","grade":"A","added_in_verification":true},{"title":"Linking Databases using Matched Arabic Names","url":"https://aclanthology.org/O14-2003/","type":"peer_reviewed","grade":"A","added_in_verification":true},{"title":"GLEIF Registration Authorities List v1.9","url":"https://www.gleif.org/en/lei-data/code-lists/gleif-registration-authorities-list","type":"official_document","grade":"B","added_in_verification":true}],"verification":{"checked":8,"corrected":4,"removed":0}}]}