Four VAT rates, one incentive to misclassify
This is the sixteenth and final country paper in the African Revenue Intelligence Research Series, instantiating ARIM (Paper 1) for Morocco. It analyses a four-rate VAT structure and a sharp 20%-to-35% CIT profit cliff, extending the threshold lineage from Ghana (Paper 5), Uganda (Paper 8), Namibia (Paper 12), and Senegal (Paper 14), completing the 15-country empirical foundation for the comparative synthesis in Paper 17.
The Direction Générale des Impôts (DGI) administers VAT under the General Tax Code at four rates: 20% as the standard, 14% for energy, passenger transport and tourism, 10% for restaurants, banking, insurance and basic foods, and 7% for water, medicines, school transport and newsprint. It is the most rate-differentiated VAT documented in this research series.
Registration is mandatory above MAD 500,000 of turnover for sales or MAD 200,000 for services, with monthly filing above MAD 1 million. The wider the gap between rates, the stronger the pull to declare a 20% supply under a cheaper heading.
Model MA1 therefore sums the gap between each taxpayer’s declared share of sales under each rate and the share typical of its sector. A general retailer with a material share under the 10% banking-and-insurance heading, and no change in what it actually sells, stands out at once.
A cliff, not a slope
Corporate Income Tax applies 20% on profits below MAD 100 million and 35% at or above it. It is a single-point discontinuity rather than a marginal schedule, so a company just under the line has a large, quantifiable reason to manage profit toward it.
Model MA2 scores each company’s proximity to the cliff, then adds a population-level test: is there more taxable profit sitting just below MAD 100 million than a smooth distribution would predict? A cash-expense deductibility cap of MAD 5,000 per day per supplier adds a second, smaller threshold of the same kind.
Data that starts with returns, and a decree with no date
ARIM’s essential tier for Morocco is entirely return-based: category-tagged VAT returns from SIMPL-TVA, corporate returns from SIMPL-IS, and historical audit-adjustment outcomes. Company-registry and special-zone records are desirable, and e-invoicing transaction data is treated as a future addition, not a current input.
Morocco’s e-invoicing regime under Article 145-IX of the General Tax Code has completed a 2025/26 B2B pilot on a four-corner clearance model, but the implementing decree fixing dates, scope and technical rules has not been published. No legally confirmed timeline exists at all, so the transaction domain is switched off by design rather than assumed.
Model MA3 tracks pilot participation and invoice-data structure purely as a deployment-planning input for the DGI. No enforcement action follows from it.
Who controls what: identity and networks
Deterministic matching uses the Identifiant Fiscal, an 8–10 digit code issued by the DGI. Probabilistic Fellegi–Sunter matching then links it to registry, director and shareholder evidence, with particular attention to checking that a claimed Casablanca Finance City or special-zone status matches the actual qualifying record.
The taxpayer graph links companies, directors, shareholders and status holders through directorship, shareholding, related-party and status-verification edges. With no e-invoicing, there are no invoice-level edges yet.
Five active domains, one explainable score
ARIM scores each taxpayer as a weighted sum across its behavioural domains. In Morocco the transaction weight is set to zero pending the decree, which makes this the earliest-stage use of the framework’s modular design.
On a fully synthetic set of 30 taxpayers across eight sectors, the seeded rate-misclassification case was the only one in the top ten with a non-zero rate gap. Cliff proximity alone proved a weak individual signal, which is why the population-level test carries the weight.
Tax behaviour
Rate-category consistency, profit-cliff proximity and cash-expense threshold proximity.
Transaction behaviour
Absent by design (ω₂ = 0) until the e-invoicing decree is published.
Network behaviour
Director and shareholder centrality, plus special-zone status verification.
Sector behaviour
Sector-adjusted margin deviation against peers on the same rate.
Historical compliance
Prior audit-adjustment count and magnitude.
External data
Company-registry cross-reference to verify preferential status.
A flag is a signal, never a finding
Every case file carries domain-tagged reason codes. Morocco adds a new kind: a code that says a profit figure within 2% of MAD 100 million is an individually weak signal, to be read alongside the population-level finding and never as a standalone flag.
Law 09-08, administered by the CNDP, governs personal-data processing, and the research recommends that no enforcement proceed on cliff proximity alone. The five illustrative cases include a false-positive check: a pharmaceutical distributor with most sales at 7% scores near zero because that is its genuine product mix.
The synthetic demonstration shows mechanical coherence only, and no detection or recovery figure is claimed for Morocco until the DGI validates the models on its own data.