A rulebook change that made 313 U.S. firms Shariah-eligible overnight moved nothing a matched control could not explain. That single fact is the most useful thing in this paper, and Qadi, Sharma and Medda are straight about it. Their abstract already says the September 2023 DJIM/S&P methodology change produces no robust matched repricing, that the pre-existing inclusions have positive but imprecise matched returns, and that significance arrives only once the pre-event turnover floor is applied. The headline claim they defend is the conditional one: official permission is associated with price effects in a recognised local market among sufficiently tradable securities, and formal eligibility alone is insufficient. So the argument is not whether they oversold the result. It is what a floor-conditioned 1.76pp is worth.

We could not run this ourselves. The price evidence is official Securities Commission Malaysia list events, and we have no Malaysian prices, turnover or list history. No U.S. substitute preserves the official local authority the paper studies.

Seven rulebooks, one common data set

Start with the signal, because it is simple and it is not a return forecast. Take a U.S. common stock, pull lagged Compustat fundamentals, and run them through seven Shariah rulebooks the authors coded themselves: AAOIFI, DJIM, S&P, FTSE/Yasaar, MSCI main, MSCI M-Series and SC Malaysia. All seven test debt, cash and an impermissible-income proxy. Five of the seven also test receivables or liquidity; AAOIFI and SC Malaysia have no receivables screen in the implemented layer. The denominators differ across rulebooks: total assets for some, 24- or 36-month average market cap for others. That produces an eligibility vector per security-month.

Two numbers come out of the vector. Disagreement, defined as 4e(1-e) on the share of rulebooks that pass the firm, peaks when the standards split evenly. And boundary proximity: for each rulebook, the distance from the binding ratio to its threshold, scaled by the threshold, minimised across standards, then pushed through exp(-5D). The scale parameter of 5 is fixed by hand, and the paper says so, calling the transform a monitoring index rather than a transition probability.

Why this might carry information. The authors formalise permitted investor mass, the capital formally allowed to hold the stock, as unconstrained money plus the mandate capital of each approving rulebook. In their mean-variance benchmark, required returns fall as that mass rises. A boundary crossing changes the vector, changes the mass, and can force rebalancing. Note the honest bit: historical mandate-capital weights are unobserved, so permitted investor mass is never estimated. Equal rulebook weights stand in, and the paper refuses to call that number investor mass.

What came back

Route A is the U.S. panel: 1,342,606 security-months, 13,188 securities, January 1999 to December 2024. Route B is the price evidence. It uses 25 official Securities Commission Malaysia semi-annual compliant-securities PDFs, November 2013 to November 2025. Events are defined at stock-code level before matching to Compustat Global. Day 0 is the release date printed on the cover page, or the first trading day after it.

The monitoring test works. In a logit on 300,000 stratified security-months with year effects and two-way clustering, disagreement carries 1.1609 (z=39.75) and proximity 3.2384 (z=30.61), with larger firms less likely to flip (-0.0557, z=-8.48). The authors flag that the outcome is a screen-implied transition computed from the same boundaries as the regressor, so part of this is mechanical. Read it as a turnover and compliance warning light.

The pricing side is where the paper earns its keep by failing. The Fama-MacBeth slope on standardised eligibility share is 0.0013 (t=4.13) with size, value, momentum and reversal controls. Add profitability, ROA, asset growth and capex and it drops to 0.0005 (t=1.62). Add the screening ratios themselves and it is -0.0001 (t=-0.35). Quality plus the screening ratios themselves absorb it. Disagreement survives at 0.0007 (t=2.54), which the paper calls a residual characteristic association rather than a priced state.

Then the clean U.S. shock. In September 2023, DJIM and S&P dropped their cash and receivables screens, mechanically admitting 313 firms, 8.00% of the 3,912 firms in the event groups, 233 of them under both families. Unmatched announcement CARs look real: 0.750pp over [0,1] (t=2.31), 1.210pp over [0,3] (t=2.63). Against matched still-ineligible controls they are 0.368pp (t=0.89) and 0.643pp (t=1.09). September placebos in non-event years give +2.088pp at [0,3] in 2020 (t=2.62) and -2.670pp at [0,5] in 2022 (t=-3.99). The paper says the 2020 and 2022 failures further limit the event design, and that the price evidence is too fragile to support an unconditional inclusion-premium interpretation.

Why did Malaysia differ?

The official-list sample funnels hard, and the significance arrives at a specific step:

The abstract concedes this ordering. It reports that the inclusions already trading before the preceding review have positive but imprecise matched returns, and it concedes that a joint 20-day pre-event test rejects. What it offers against that are the checks it lists. Leave-one-list-date-out keeps [0,10] between 1.29 and 1.98pp. First-inclusion-only on 237 securities gives 1.688pp (p_wild=0.018). A three-month mid-review timing placebo on the same matched sets is null (0.841pp, p_wild=0.473). Nearby turnover cutoffs reproduce the estimate (quartile 1.84pp, tercile 1.62pp, median 2.53pp) while market-cap cutoffs do not (0.87pp, p_wild=0.174).

Two results inside the paper still argue against reading this as constrained-buyer demand pressure. Within inclusions there is no dose-response in turnover: the slope is 0.086 for [0,10] (p_wild=0.925). The earlier positive interaction came entirely from a strongly negative exclusion slope of -3.543 (p_wild=0.010). And requiring the three controls to share the treated firm's two-digit SIC leaves 266 events at 0.869pp with p_wild=0.273. The estimate needs a pre-event turnover floor to reach significance and loses it under same-industry controls, which is not what a clean constrained-buyer demand shock looks like.

The mechanism diagnostics come up empty, and the authors report it. Shariah-sensitive portfolio ownership rises 0.022pp across 288 treated inclusions (p_wild=0.051) against 0.000pp for controls, but the difference is 0.021pp with p_wild=0.360. Predicted constrained pressure forecasts the ownership change (0.059, p_wild=0.037) and its CAR coefficient is negative (-0.006, p_wild=0.264). Listed ETFs are not the buyer either: capacity pressure loads 0.000 on CAR[0,10] (p_wild=0.993), and no list date shows positive aggregate share creations over [0,10]. The price path also refuses to behave like same-day repricing of a public list, with [0,1] at -0.233pp and the effect accumulating after day 3.

Inference rests on 24 non-baseline list-date clusters. The remedy on offer is a 999-replication wild-cluster bootstrap by list date, which the authors call intentionally conservative, alongside two-way date-security clustering. The cluster count itself does not change, and they say 24 clusters remain a material limitation.

No transaction costs are charged anywhere in the paper, and the authors say as much themselves. They write that a tradable early-warning strategy would require out-of-sample forecasts of official rather than researcher-emulated classifications, a decision rule fixed before each review, implementation lags, transaction costs, and investable capacity. Capacity is the part worth staring at. Across the 378 liquidity-qualified inclusion events with ETF coverage, a one-percent allocation from the core Malaysian Shariah ETFs equals 2.81 times the median treated stock's pre-event average daily trading value.

What a desk can use

Current eligibility tells a constrained mandate what it may hold. Disagreement and boundary distance rank which holdings are likeliest to flip label next month under the emulated rulebooks, which is a monitoring signal rather than an official reclassification, and the ranking is partly mechanical by the authors' own account. Authority-specific exposure matters more than the count of approving rulebooks, since the equal-weighted count prices as quality once the screening ratios are controlled for. Tradability decides whether an official change costs anything to implement.

What the study does not deliver is a pre-review trade. Official inclusions and researcher-emulated labels are different objects. The boundary measures were validated against the second. The paper states outright that it does not claim these measures constitute a profitable strategy. I would change my mind if the same 1.76pp appeared with industry-matched controls and a positive turnover slope inside inclusions. Until then the defensible reading is theirs: a conditional price association in one market under one regulator, and permission alone buying nothing.