AAOIFI Standard 21: the canonical halal stock screen

AAOIFI stands for the Accounting and Auditing Organization for Islamic Financial Institutions. It was founded on 26 February 1990 in Algiers and formally registered as an independent international body in Bahrain on 27 March 1991. It is headquartered in Manama, Bahrain, the city that hosts most of the Gulf Cooperation Council's Islamic-finance infrastructure (the Bahrain Institute of Banking and Finance, the General Council for Islamic Banks and Financial Institutions, several major Islamic banks).

AAOIFI is not a regulator. It is a standard-setting body, much like the International Accounting Standards Board (IASB) is for IFRS or the Financial Accounting Standards Board (FASB) is for US GAAP. Its outputs are voluntary documents - standards on accounting, auditing, governance, ethics, and Shariah - that regulators and central banks then adopt, reference, or use as guidance.

The reach of AAOIFI's standards is wider than most retail investors realise. Bahrain, Oman, Syria, Sudan, Jordan, and the Qatar Financial Centre adopt AAOIFI standards as binding law. Saudi Arabia, the United Arab Emirates (at the federal level), Malaysia, Indonesia, Pakistan, and Kazakhstan use AAOIFI standards as primary guidance even where local Shariah authorities issue their own rulings on top. Most of the world's major Islamic banks (more than 200 institutions across 40-plus jurisdictions) prepare their financial reports in accordance with AAOIFI accounting standards and reference AAOIFI Shariah standards in their product-approval committees.

The Shariah standards are issued by AAOIFI's Shariah Board, a permanent panel of senior scholars representing the four major schools of Sunni jurisprudence (Hanafi, Maliki, Shafi'i, Hanbali). Each standard goes through a multi-year drafting process - position papers, public exposure drafts, scholar review rounds - before final approval. This is why AAOIFI standards carry the weight they do: they are not the ruling of a single scholar, they are the negotiated consensus of a diverse expert panel published only after extensive review.

Shariah Standard No. 21 was published in this process. Its full title in English is “Financial Paper (Shares and Bonds).” In the AAOIFI catalogue it covers the conditions under which a Muslim investor may lawfully purchase, hold, lend, or sell shares of a publicly listed company. It is one of the standards most-frequently referenced by Islamic asset managers, index providers, and retail-facing halal-screening services - because it is the only AAOIFI standard that speaks directly to listed-equity investing.

The first canonical test asks a binary question: is the company's primary line of business inherently impermissible under Islamic commercial law? If yes, no amount of clean financial ratios can rescue the stock - it fails immediately.

The exclusion list codified in Standard 21 covers the following categories: alcohol production, distribution, and retail; tobacco production, distribution, and retail; pork production, processing, and retail; gambling, lotteries, and casinos; conventional banking and lending (any institution whose primary income is interest on loans); conventional insurance (premiums-and-claims structure where the underlying risk pool is invested in interest-bearing assets); adult entertainment and pornography; conventional weapons manufacturing (especially offensive weapons used to harm non-combatants); and any business whose primary revenue depends on a transaction structurally prohibited under Shariah (riba, maysir, gharar).

The test is applied at the level of the consolidated reporting entity. A holding company whose subsidiaries include a small alcohol-related business but whose primary revenue comes from a Shariah-compliant operating segment can still pass test 1 - the question is which activity is primary, not which activity exists. invest-like uses the GICS sector and industry classification as the first-pass signal, then applies named-ticker overrides for cases where the classification understates or overstates the prohibited activity. A few examples: Berkshire Hathaway holds significant exposure to conventional insurance and is therefore excluded structurally despite its diversified holdings; Visa and Mastercard are payment networks, not lenders, so they pass test 1 even though their counterparts (the issuing banks) do not.

The second test is a leverage cap. Under AAOIFI Standard 21, the ratio of total interest-bearing debt to the company's 36-month average market capitalisation must be below 33 percent. The AAOIFI Shariah Board's more recent rulings have favoured a stricter 30 percent ceiling, and invest-like uses the 30 percent threshold as its default to align with the conservative consensus of contemporary scholars.

The numerator includes all forms of interest-bearing financing: short-term debt, the current portion of long-term debt, long-term debt, bonds, notes payable, and any other liabilities that explicitly pay interest. Trade payables, deferred revenue, accrued expenses, lease liabilities under IFRS 16 (operating leases), and pension obligations are excluded - the test is about riba-bearing financing specifically, not about all liabilities.

The denominator is the trailing 36-month average market capitalisation. The averaging window is a deliberate design choice. A point-in-time market-cap denominator would let a stock flip between compliant and non-compliant on a single-day price move - if the share price drops 40 percent in a quarter, the same debt level suddenly looks much worse against the smaller market cap, but the underlying business has not become more leveraged in any economically meaningful sense. The 36-month average smooths through cycles. Different index providers use different windows (S&P Shariah uses 24 months; Dow Jones Islamic uses 24 months; AAOIFI's reference recommends 36 months); invest-like uses 36 to track the AAOIFI convention.

Why market cap, and not total assets? The AAOIFI Shariah Board's reasoning is that the equity investor's actual exposure is to the going-concern value of the business as priced by the market, not the historical-cost value as recorded on the balance sheet. A capital-intensive business with large fixed assets at book value but a depressed market cap would look misleadingly compliant under a total-assets denominator. The market-cap denominator forces the test to reflect what the equity investor is actually buying.

The third test caps the fraction of total revenue that may come from impermissible sources. Under AAOIFI Standard 21, the ratio of impermissible income to total income must be below 5 percent. Above 5 percent, the stock fails this test outright. Between 0 and 5 percent, the stock passes, but the investor is expected to purify the corresponding fraction of any dividends received - typically by donating that fraction to charity, removing the impermissible portion from the realised gain.

The impermissible-income sources covered by the test are broad. They include: interest income earned on cash and short-term investments held by the company; revenue from any haram product line that the company operates as a non-primary business (a hotel chain that derives a small fraction of revenue from minibar alcohol sales, an airline that earns a fraction of revenue from in-flight alcohol service, a supermarket chain that derives revenue from pork or alcohol sales); royalty or licensing income from impermissible activities; rental income from properties leased to impermissible tenants.

The challenge with test 3 is data availability. Public companies rarely disclose revenue at the granularity AAOIFI implies. A hotel chain's annual report typically discloses revenue at the segment level (lodging, food and beverage, fees and franchise) but not at the line-item level (food, alcohol, in-room minibar). For these cases, invest-like uses a combination of (a) the company's own disclosed percentages where available and (b) industry-average estimates as a conservative fallback, with named-ticker overrides for the largest cases. A few examples of named overrides: Hilton, Marriott, and Hyatt are flagged as borderline because their food-and-beverage segment historically derives 10-15 percent of revenue from alcohol; Netflix is flagged questionable because of its mature-content exposure, even though the gross revenue fraction is small.

The fourth canonical test caps the proportion of the company's value that consists of liquid financial assets rather than productive operating assets. Under AAOIFI Standard 21, the ratio of (cash plus interest-bearing securities plus receivables) to the 36-month average market capitalisation must be below 70 percent. Some interpretations apply a stricter 30 percent ceiling on the narrower (cash plus receivables) sub-numerator; AAOIFI's own text leaves room for both readings, and the looser 70 percent ceiling is the version most commonly enforced by the major Islamic index providers.

The intuition behind the test is that an equity investor in a Shariah-compliant business should be buying a productive operating business, not a cash pile dressed up as a corporation. If 70 percent or more of the market cap represents financial assets, the investor is effectively buying a money-market fund with a thin operating veneer - and Shariah rules out the investor receiving the underlying interest-bearing returns indirectly through the equity. The test catches cash shells, special-purpose acquisition vehicles before they have deployed capital, and some closed-end funds.

For ordinary operating companies the test is rarely binding. A typical Tech, Healthcare, Energy, or Industrials business has a liquid-assets ratio well below 70 percent because the bulk of its value sits in customer relationships, brand, intellectual property, plant, and inventory rather than cash. The test most frequently binds on small-cap companies that have recently raised capital and not yet deployed it, on holding companies with large minority stakes, and on certain financial-services adjacent businesses (asset managers, brokerages) where the test 1 primary-business exclusion has already done the work.

Every ticker in the 12,500-stock universe is run through the same deterministic implementation at each quarterly refresh. The logic lives in a single file (src/lib/scoring/halal.ts) so the screen is one place, audited as one unit, and never diverges between the landing page and the per-stock pages.

Test 1 runs on a structured exclusion list keyed off GICS sector, industry, and a curated set of named-ticker overrides. The exclusion list itself is committed to the repository, version-controlled, and reviewed publicly so any user can audit which industries are excluded and why. When a stock matches one of the exclusion categories, the function returns non_compliant with a reason code identifying the category (industry_alcohol, industry_gambling, industry_conventional_finance, industry_pork, etc.). When the stock's industry falls into a borderline category (insurance broker, financial-technology infrastructure that does not extend credit, certain real-estate operators), the function returns questionable rather than auto-failing, surfacing the case for human review.

Test 2 runs on the live fundamentals data. The numerator is the long_term_debt field from the balance sheet (which in our data model captures the total interest-bearing debt including current portion). The denominator is the 36-month average market cap, computed by averaging the trailing 36 monthly closing market caps from the price history. The 30 percent threshold is the configured default; the function exposes the actual ratio (not just the pass/fail bit) so the per-stock page can show the exact number, e.g. “debt ratio: 14.2 percent - clear pass” or “debt ratio: 31.8 percent - just over the 30 percent line.”

Test 3 is the test most affected by data gaps. Where the company discloses revenue at the segment level with enough detail to compute the impermissible fraction directly, invest-like uses the disclosed figure. Where the company does not, invest-like falls back to a curated table of named-ticker overrides for the cases where industry context strongly implies impermissible revenue (hotels, airlines, mixed-retail chains). When neither disclosed segment data nor a named override applies, the test is documented as “applied where data permits” rather than silently skipped - the methodology page at /methodology/halal/ enumerates the cases where test 3 does and does not fire.

Test 4 is similarly applied conservatively. The receivables figure used as the narrow numerator is the standard accounts-receivable line from the balance sheet. Cash includes cash plus short-term investments. The wider numerator (cash plus interest-bearing securities plus receivables) uses the long-term investments line for the interest-bearing-securities term where the data is available. Where the data is incomplete, the function returns insufficient_data rather than guessing - in v1 of the screen, insufficient_data resolves to the same UI outcome as questionable, prompting the user to apply manual judgement.

All four test outputs (pass/fail + numerical value + reason codes) are persisted on the stocks table at refresh time, indexed for query, and re-rendered into the static halal pages with daily revalidation. There is no “black box” in the screen - every per-stock verdict is the deterministic output of the same auditable function, and the per-stock page at /halal/[ticker]/ exposes the underlying numbers so the user can verify the verdict for themselves.

The full implementation of the screen has been documented in a working paper deposited at Zenodo (DOI 10.5281/zenodo.20393706). The paper describes the data sources, the threshold defaults, the named-ticker overrides, the limitation notes (especially around test 3 data availability), and the validation harness that compares the invest-like verdicts against published Islamic-index inclusion lists.

The working paper is open-access, citable from any peer-reviewed venue, and stable - the DOI guarantees the methodology a researcher cites today will resolve to the same document a year from now even if the running code evolves. For users who want to audit the screen against their own dataset, the methodology page at /methodology/halal/ documents every threshold, every override, and every edge case in production.

For the avoidance of doubt: invest-like is not staffed by Islamic scholars, the screen is a structured filter rather than a fatwa, and any user whose local scholar or madhhab uses different thresholds should treat the screen as a starting point rather than a final ruling. The reason to publish the methodology this transparently is precisely so a scholar can examine it, identify where it diverges from their interpretation, and make an informed adjustment for their own follower base.